We need to tell you about what happened four days ago.

The Reuters Institute published their 2026 trends report. They surveyed 280 media leaders from 51 countries in November and December. Google search traffic to publishers fell 33% globally in the year ending November 2025. In the US? 38%.

But here’s the part that made me sit up: Publishers expect another 43% decline over the next three years. One in five said they expect to lose more than 75% of their search traffic. When asked about 2026 strategy, most said they’re putting less effort into traditional Google search.

Not pivoting. Not adapting. Less effort.

That’s not fear. That’s capitulation. The industry just admitted the old playbook is finished.

Business Insider cut 21% of staff after losing 55% of organic traffic. The New York Times watched search drop from 44% to 37% of their traffic. DMG Media (MailOnline, Metro) saw click-through rates fall up to 89% on certain searches.

The commodity TOFU factory isn’t closing. It’s already closed. The smoke is still clearing.

Picture This

You’re a publisher. You built your business on a simple bargain: produce good informational content, rank it, get clicks, make money. It worked for a decade.

Google rolled out AI Overviews in May 2024. By October, they showed up in over 42% of searches. When Pew Research tracked 68,000 real queries in July 2025, users clicked results 8% of the time when AI summaries appeared. Without AI summaries? 15%.

That’s a 46.7% drop in clicks. Not for bad content. Not for thin content. For any content that gets an AI Overview.

Zero-click searches went from 56% to 69% between May 2024 and May 2025. Think about that. Seven out of ten searches now end without anyone clicking anything.

Then December Hit

Google pushed a core update in December 2025. Publishers reported “catastrophic” losses. Some went from steady Discover traffic to zero overnight. Not “down 20%.” Zero. Traffic declines ranged from 12% to 100% depending on the site.

Chegg (the learning platform) saw a 49% decline in traffic between January 2024 and January 2025. They filed an antitrust lawsuit in February 2025 alleging Google trained AI on their content, then used that AI to compete with them directly.

Google’s response? They’re tightening spam policies. Explicitly targeting “scaled content abuse”. Massive volumes of pages created primarily to game rankings. When everyone can mass-produce TOFU with AI, Google either raises the bar or drowns in garbage.

They chose to raise the bar. Way up.

Two Things Are True At Once

Supply collapsed

It’s stupidly cheap to produce infinite “good enough” explanations now. AI-assisted writing is a major share of new publishing. The web isn’t saturated. It’s drowning.

Demand evaporated

Users don’t click anymore. They get synthesis in the AI Overview, they ask ChatGPT, they stop earlier because the answer feels complete. Although, Digiday reported in December 2025 that all AI platforms combined account for 1% of publisher traffic. ChatGPT sends 87.4% of that 1%. It’s not enough to offset Google losses. Not even close.

But you’re fighting on three fronts now:

• Classic search

• AI Overviews on Google

• Chat interfaces (ChatGPT, Gemini, Perplexity)

And you can’t assume any of them will send clicks.

TOFU keywords behave like “answer endpoints” now. Some drive clicks. You just can’t bank on it.

So What Actually Survives?

Back to that Reuters report.

Original investigations and on-the-ground reporting ranked at the top. Service journalism and evergreen content? Least important.

That’s executives repositioning their entire operations based on data, not theory.

The shift isn’t TOFU to MOFU. It’s commodity TOFU to defensible TOFU.

1. Experience TOFU

First-hand tests. Before/after. Teardowns. “We tried it so you don’t have to.” Aligns with Google’s E-E-A-T framework on experience, but more importantly, it’s just hard to fake at scale with AI.

2. Data TOFU.

Original research. Benchmarks. Live indexes. You either did the study or you didn’t. Plus it’s the easiest thing for AI to cite. You win even when users don’t click.

3. Community TOFU.

Threads. Annotated case studies. Public teardowns. When your TOFU lives in community spaces, you’re not dependent on Google’s algorithm at all.

4. Tool TOFU.

This is where it gets interesting.

Research from Brixon Group in May 2025 found ROI calculators hit 65-78% completion rates. Conversion rates of 25-35%. Compare that to regular landing pages.

B2B decision-makers spend an average of 4 minutes 27 seconds on interactive calculators. More than three times typical dwell time on static content.

A marketing automation provider saw 218% higher conversion rates with AI-powered ROI calculators versus static content. A B2B SaaS company increased conversion rate from SQL to Opportunity by 34% and cut their sales cycle by 21 days just by integrating calculator data into sales process.

If the answer requires interaction, the SERP can’t steal it. You have to use the tool. Reading about a calculator doesn’t help you calculate.

The Job Changed

Old TOFU goal was rank, click, retarget.

New TOFU goal is different.

• You shape mental models so your framework is in their head before they’re ready to buy. You do preference formation so your approach feels obvious before they evaluate alternatives. You create distribution assets that travel on their own. You become the “source chunk” AI tools reference.

• You design for “SERP-as-feed” not “SERP-as-portal.” Assume most won’t click. Win value without the click. Publish frameworks people quote. Structure content so AI extracts clean chunks.

• Don’t scale content. Scale insight production. Google explicitly targets scaled content abuse. Doesn’t matter if AI or humans make it. Use AI for research, outlining, repurposing. But gate output on novel input. Proprietary data. Real examples. Expert review. Actual POV.

Your KPIs need updating too. Lower clicks mean you measure brand search lift. Assisted conversions. Share of conversation across answer surfaces. Newsletter capture. Community growth. Your “organic sessions” dashboard understates influence by a bigger margin every quarter.

The Real Question

HubSpot lost 70-80% of traffic in weeks. They had one of the best SEO teams in the world.

NerdWallet built their entire business on organic search. Brutal quarters.

280 media executives from 51 countries just told Reuters they’re walking away from traditional search. Not pivoting. Walking away.

Here’s what nobody’s saying out loud: most of you reading this are still building the same TOFU you were building two years ago. Same keyword exports. Same content briefs. Same “optimize for volume” approach.

You saw HubSpot’s numbers. You read about Business Insider cutting staff. You nodded at the Reuters data.

And tomorrow you’ll open the same keyword tool and map content to the same metrics that stopped working in May 2024.

The people who survive this aren’t the ones waiting for “best practices” to emerge. They’re the ones who looked at interactive calculators converting at 25-35% and said “we’re building that next week.” Who started publishing original research nobody else has. Who stopped chasing rankings for “what is X” and started building tools people actually need to use.

You don’t need permission. You need to decide whether you’re rebuilding or just rearranging deck chairs while your traffic drops another 33% this year.

The data came out four days ago. What are you doing on this upcoming Monday?

Signing out
Pankaj & Vaishali

IKEA launched in Japan in 1974. Their furniture was beautiful, affordable, exactly what had worked across Europe.

Twelve years later, they pulled out completely.

The problem wasn’t quality. Wasn’t price. It was size. IKEA’s standardized dimensions (perfect for Northern European apartments) were comically oversized for Japanese homes.

Three-seat sofas that couldn’t fit through doorways. And most Japanese consumers didn’t own cars, so they were expected to haul these massive flat-packs home on packed subway trains.

IKEA resisted adapting. They figured customers would adjust.

They didn’t.

Twenty years later, IKEA returns to Japan with a completely different approach. They curated only products that fit Japanese apartments. Shortened sofas. Added delivery and assembly services. Designed showrooms around typical Japanese living spaces.

Today? Multiple stores across Japan. A localization success story.

So why am I telling you this furniture story in an SEO newsletter?

Because the exact same dynamics are playing out right now with AI-powered international SEO. The technology makes it trivially easy to launch in new markets. Does absolutely nothing to help you win in those markets.

You can spin up translated pages in 47 languages by Friday. Doesn’t mean anyone will trust them, rank them, or convert on them.

Let me show you where this is all heading.

Before we get into the weeds, let’s nail down what “international SEO” actually means (because half the confusion comes from people using the same term to mean completely different things).

International SEO is really two jobs running in parallel:

First job: Make search engines understand which version of a page is for which audience. Language + country/region. This is the technical layer (site structure, indexing signals, canonicalization, hreflang tags).

All the machinery that keeps search engines from showing your French-Canadian page to users in France (or vice versa).

Second job: Make humans in each market feel like the page was built specifically for them. Not just language. Cultural norms, trust signals, UX expectations, the unspoken rules that separate “this brand gets me” from “this is clearly a foreign company trying too hard.”

AI touches both jobs. What’s fascinating (and where teams are getting destroyed) is watching how those two responsibilities blend together in ways that create entirely new failure modes you wouldn’t see coming.

Two things changed the game in last one decade:

Neural machine translation made “good enough at scale” realistic starting around 2016. But the real breakthrough came in 2017 with the transformer architecture, which fundamentally changed how translation models work.

Since then, progress accelerated fast. Google’s PaLM 2 in 2023 demonstrated that large language models could outperform traditional translation systems. In 2024, this enabled Google Translate’s biggest expansion ever (110 new languages), reaching 614 million more speakers.

But translation quality still varies wildly. Portuguese to English? Excellent. Idiomatic Japanese to German? Still challenging. Recent research shows that claims of “human parity” need serious caveats.

What this means: AI makes it cheap to spin up thousands of localized pages fast. Doesn’t make them correct, culturally appropriate, or aligned with actual search intent.

The second shift?

Google now bridges language gaps automatically. They’ll translate snippets on the fly. Sounds helpful until a user in Madrid lands on your auto-translated English page, bounces because the tone feels off, and your rankings slide.

There’s a foundation layer AI can help with (or completely obliterate). Think of it as the physics layer.

First: correct locale signaling. Google’s documentation is explicit about hreflang annotations. AI can generate these at scale. Beautiful in theory.

In practice? Auto-generate the wrong mapping and you create indexation chaos. I’ve seen teams launch thousands of AI-generated pages where 30% had broken hreflang. Google ignored the whole setup. Months of work, zero ranking benefit.

Second: language tagging standards. Web standards push hard on BCP 47 language tags. AI workflows constantly produce mismatched labels. I’ve watched an AI tool assign “zh-CN” to traditional Chinese content meant for Hong Kong (should be “zh-HK”). Small mistake, weeks to fix.

So where does AI actually help in International SEO?

Market selection: AI can synthesize demand proxies, competitor footprints, content gaps, and cost models. When done right, it beats “we should probably do Germany because... Germany?”

ActiveCampaign did this when expanding into French, Italian, German, and Spanish markets. They identified which keywords had real search volume per market, which topics needed education versus direct pitches.

Over eight months: 392 pages across four languages, 83,000 additional clicks, 4.1 million more impressions.

That’s market-informed expansion, not translation.

Keyword and intent modeling: Here’s where most teams die. Literal translation fails because search intent isn’t the same across cultures.

ActiveCampaign discovered Italian users needed more educational content before purchase. German users obsessed over data privacy. Spanish markets responded to completely different pain points.

They couldn’t translate their English keyword strategy. They rebuilt it per market with native specialists. Result: 73% year-over-year traffic increase to their Spanish blog, 111% rise in Spanish Help Center usage.

That’s the difference between “we translated it” and “we understood the market.”

Content localization operates on a ladder:

• Translation = meaning transfer

• Localization = meaning + locale conventions (currency, date formats, payment methods)

• Transcreation = meaning + persuasion adapted (tone, humor, cultural references)

A CTA like “Grab yours now!” works in the US but sounds unprofessional in Germany. In Japan, it can feel disrespectful.

AI speeds drafts dramatically. But “quality” is multi-dimensional. Linguistic fluency doesn’t equal cultural resonance doesn’t equal SEO outcomes. You can have perfectly fluent content that ranks terribly or converts poorly.

Cultural adaptation is where teams under-invest and AI both helps and harms.

Nielsen Norman Group breaks this down: high-context cultures (Japan, Middle East) prefer indirect cues. Low-context cultures (US, Germany) want explicit information.

When IKEA returned to Japan, they didn’t just resize furniture. They redesigned the entire experience. Swedish cafés became destinations. Menu items adapted (matcha lattes alongside meatballs). Plant-based options aligned with local sustainability values.

They understood complete cultural context, not just product dimensions.

AI can generate culturally appropriate variants and flag risky phrases. But it can also produce generic content that feels flat or miss delicate social norms.

One indie game developer launched on Steam Japan with machine translation. The title became “クソダンジョン” (”Sh*tty Dungeon”). Brand-destroying failure overnight.

And localization isn’t just text. It’s product UX. Engagement signals depend on localized checkout flows, local payment methods, shipping estimates that match expectations.

What happens when you deploy AI-driven international SEO at scale?

Speed-to-market becomes realistic. But volume is easy. Differentiation isn’t.

I’ve seen an eComm brand launch with 7,000+ AI-translated product pages and get zero traction because none felt native. It’s like showing up to a business meeting in another country wearing technically appropriate clothing that’s just... slightly off.

Everyone notices. Nobody says anything. You don’t get the deal.

In markets like China (Baidu), South Korea (Naver), and Japan (Yahoo Japan), local search engines have different ranking factors. You need local hosting, local backlinks, structured data that reflects regional expressions of authority.

AI can identify gaps but can’t be a substitute for building relationships.

Policy and governance keep this from turning into a penalty factory.

Google maintains spam policies and Bing explicitly discusses “automatically generated content.” Both watch for massive content production without quality checks.

The tension: AI is phenomenal for scale. Search engines are wary of scale without value. If your process is “AI translates everything, we publish immediately,” you’re building a penalty waiting to happen.

Professional hreflang audits routinely find improper implementation causes ranking drops. Most aren’t caught until traffic crashes.

DeepL’s 2024 survey found 96% reported positive ROI from localization, 65% reported 3× ROI. But that’s proper localization with human oversight, not machine translation dumped onto pages.

Most teams approach this as cost-cutting. “We can translate faster and cheaper!” Then wonder why conversions never materialize.

The actual play: use AI for mechanical work (hreflang generation, translation drafts, technical QA) so humans can focus on cultural adaptation, intent alignment, trust building.

Think about it: AI is incredible at the “what” and “how much.” Terrible at the “why” and “so what.” You need both.

If your international SEO strategy starts and ends with “we used AI to translate everything,” you didn’t build an international presence. You built 10 versions of the same site that all fail to convert in different languages.

IKEA learned this twice. First failure taught them standardization doesn’t work everywhere. Second attempt taught them to balance global brand consistency with local market adaptation.

Twenty years between attempts. You can learn it faster. Question is whether you’ll learn it the easy way by paying attention to what actually drives results in each market, or the expensive way by launching fast, failing hard, and fixing it later when you’ve already burned trust.

Human oversight isn’t optional, or not just nice-to-have. It’s the entire point.

Because you’re not trying to fool algorithms into ranking pages. You’re trying to build trust with humans in markets you don’t fully understand yet. AI can help you move faster. Can’t help you move smarter unless you’re steering it with actual market knowledge and cultural understanding.

That part still requires humans. Probably always will.

See ya next week..

Signing off
Pankaj & Vaishali

You know what’s breaking right now? The entire concept of “ranking.”

Not slowly. Not “oh we’ll adapt over time.” It’s already broken, and most SEO teams haven’t noticed yet because their dashboards still show pretty green lines going up.

Here’s the problem: you’re measuring success by position #3 for a keyword, but there is no single #3 anymore. There are hundreds of them. Maybe thousands. Each one tailored to a different user, a different context, a different conversation history.

Ask ChatGPT or Perplexity for B2B marketing tools for a small business right now. Then ask your colleague to do the same thing. Different answers. Not slightly different but materially different brands, different reasoning, different “winners.”

Same question. Same model. Totally different outputs.

So when you report “we rank #3,” what does that even mean? #3 for who? In what context? With what user history? The metric you’re optimizing for doesn’t correspond to any single reality anymore.

And that’s just the start of the problem.

This week we’re breaking down why AI answers differ between users, what’s actually personalization versus normal system chaos (they’re not the same thing), and what this means for how you need to think about SEO in 2026.

It gets technical in spots, but every bit of this matters for your day-to-day work.

Stop Confusing Two Completely Different Things

When people see different AI answers, they usually just shrug and say “oh, personalization.” But that’s lazy thinking. There are actually two separate phenomena happening here, and mixing them up will wreck your strategy.

True Personalization (The System Knows You)

This is when the AI intentionally adapts based on signals about you specifically.

OpenAI lets you set custom instructions like tone, format, response style. You tell it “be concise” or “use bullet points” and it remembers. Perplexity goes further, letting you fill out an actual profile with your bio, interests, even your location for better local results. The systems are literally conditioning outputs based on what they know about you as an individual user.

And then there’s memory. If you’ve enabled it (and honestly, most people have without realizing it), ChatGPT is remembering facts across conversations. “I run a SaaS agency.” “I’m vegetarian.” “I prefer short answers.” All that context stacks up and influences future responses.

This is real personalization. The system is using data about you to change what it says.

System Variability That Just Looks Like Personalization

But here’s what most people miss: even with zero user profile, answers can still differ wildly.

The same prompt can generate different outputs because of non-determinism in how these models run inference. Researchers have documented this—you can feed identical inputs and get materially different results just from normal system-level factors.

Then there’s retrieval. In RAG systems (Retrieval-Augmented Generation, the thing that lets AI pull from the web), what documents get fetched can vary by time, region, how the index is feeling that day, or backend changes nobody tells you about. Unless the system is explicitly using your profile to personalize retrieval, this variability has nothing to do with you. It’s just the system being inconsistent.

And don’t even get me started on model updates, A/B tests, and APIs. You and I could be on completely different model versions right now and we’d never know. Vendors ship changes constantly. Two users in different experimental cohorts see different outputs, and that gets blamed on “personalization” when it’s really just the platform testing new stuff.

Also, you may don’t know but what you get from apps like Claude or ChatGPT could be totally different from the results you get when you ask the exact same question using their API. That’s a crucial thing many people might overlook.

So, what you’re actually experiencing most of the time? A messy blend of some deliberate personalization plus a lot of ordinary system chaos.

How Personalization Actually Happens (When It Does)

Let me walk through the four ways these systems genuinely personalize, because understanding the mechanics changes how you think about optimization.

1. Your Style Preferences Change What Gets Surfaced

Most platforms let you set how you want responses formatted. Seems simple, right? But this doesn’t just change presentation, it also changes content selection too, because the model starts prioritizing different trade-offs based on your settings.

If you’ve told ChatGPT you want concise answers, it’s going to lean toward sources that are easier to summarize. If you prefer detailed explanations, it’ll pull from longer-form content. Same question, different emphasis, different sources highlighted.

For SEO, this means visibility isn’t just about being relevant, it’s about being packagable in multiple formats. Dense academic papers might win for some users. Snackable comparison guides might win for others. You need both.

2. Memory Creates Brand Familiarity Loops

The memory feature is where things get interesting. And by interesting, I mean potentially problematic for brands trying to break in.

When the system remembers context across sessions, it can develop preferences based on what worked before. If a user accepted a recommendation last time, that brand gets weighted more heavily next time. If they engaged with certain sources, those sources become the default.

OpenAI’s own documentation talks about this as a feature, which means: less repetition, more relevant answers, better user experience. And sure, that’s true from the user’s perspective. But from a brand visibility perspective ? you’re looking at path dependency.

Get in early, stay in. Get left out early, fight like hell to break back in.

The privacy angle here is real too. More stored context means more attack surface for data leakage and prompt injection. Research on LLM security consistently flags this risk. But platforms are pushing forward anyway because the UX benefits outweigh the concerns for most users.

3. Location and Context Are Filtering You Out

Even without a long-term profile, the system is making assumptions based on contextual signals.

Google’s been doing this forever and they’ll straight up tell you that search results vary based on language, location, and other factors. Perplexity does it too, using your location to improve accuracy for things like weather and local business queries.

For SEO, this makes local and geo-specific optimization way more fragmented. The “near me” effect gets amplified in AI answers. If you’re not showing up in the right regional context, you’re invisible to entire user segments even if you’d technically be the best answer.

4. Personalized Retrieval Is The Endgame

This is the big one. The mechanism that actually breaks traditional SEO measurement.

What if the AI retrieves different documents because of who you are? Not just different presentation of the same sources, but fundamentally different sources based on your history, preferences, past clicks, previously accepted answers.

Research on personalized LLMs shows this is absolutely possible. You can use embeddings of user profiles, long-term memory stores, or personalized ranking algorithms to tailor which documents even get considered before the model generates an answer.

If—when—answer engines go this direction at scale, “ranking” stops being a single leaderboard you can check. It becomes thousands of micro-leaderboards. One for each persona. One for each context. One for each user journey pattern.

You can’t just ask “where do we rank?” anymore. The question becomes “which cohorts see us, and which don’t?”

That’s a fundamentally different optimization problem.

Why Everything Won’t Be Hyper-Personalized (Probably)

Before you spiral into panic mode, let’s talk about the forces pushing against extreme personalization. Because it’s not a foregone conclusion that every answer becomes completely unique to each user.

The case for personalization is obvious. Better UX, less repetition, more relevant answers, assistants that feel like they actually know you. OpenAI positions memory and customization as quality-of-life features, and they’re right.

But the case against is strong too.

Privacy and security get exponentially harder with more personalization. Every bit of stored context increases the blast radius when something goes wrong like prompt injection, data leaks, inference attacks. Security researchers are sounding alarms specifically about RAG and agent systems that maintain persistent user state.

Then there’s the fairness problem. Heavy personalization can create filter bubbles that reduce viewpoint diversity and reinforce existing beliefs. Great for confirmation bias, terrible for actually learning anything new or encountering alternate perspectives.

And reproducibility becomes nearly impossible. If answers vary too much user-to-user, how do you audit for correctness? How do you verify claims? How do you hold the system accountable when everyone’s seeing something different?

My bet? We’ll see selective personalization in areas where it clearly improves UX—workflow optimization, format preferences, remembered context but with guardrails in sensitive domains where too much variability creates problems.

Not everything personalized. Not nothing personalized. Somewhere strategic in the middle.

What Changes In Your Actual SEO Work

Enough theory. Let’s talk about what this means for the work you’re doing Monday morning.

You Can’t Measure The Old Way Anymore

Traditional SEO assumes you can check a relatively stable SERP and track your position. That assumption just died.

If two users get materially different recommendations for the same query, what does “average rank” even mean? Not much. Maybe nothing.

You need cohort-based visibility instead. Track by persona, geography, industry, funnel stage. Build test accounts that represent your key audience segments and monitor what each cohort sees over time.

Your SEO reporting is about to shift from “we rank #3 for this keyword” to “we appear in 18% of AI answers for mid-market SaaS buyers in financial services.”

Our prediction: We might witness the closure of various companies that track LLM prompts aka modern keyword tracking tools.

That’s a massive change in how you communicate value to executives. And yeah, it’s going to be an uncomfortable conversation the first time you have it. But it’s the only way to measure what’s actually happening.

Keywords Are Out, Entities and Scenarios Are In

Personalized systems don’t want generic pages that try to be everything to everyone. They want content that cleanly matches a specific context.

Think about how this plays out. The AI knows “I’m a Series A SaaS marketer in fintech with a team of five.” It’s looking for content that speaks directly to that scenario. A generic “marketing tools” page doesn’t cut it anymore.

What wins is crystal-clear entity positioning—who you are, what category you own, what you’re best at. No vagueness. And scenario-based content—”best for small teams,” “enterprise compliance guide,” “under $50K budget” comparison pages.

The content also needs to be retrieval-friendly. Structured, specific, easy to snippet without heavy rewriting. If the model has to work hard to extract value from your page, it’ll just cite someone else.

This is exactly why vague positioning gets crushed in AI search. The model needs a crisp “bucket fit” to know when you’re relevant and when you’re not. Our last issue was exactly on this “SEO x Positioning.”

Brand Compounds Faster (For Better And Worse)

In personalized environments, the AI develops trust priors based on user behavior.

Did your recommendation work last time? You get weighted more heavily next time. Do you show up consistently across sources? The model starts treating you as more authoritative. Does your positioning stay consistent everywhere? That makes you easier to categorize and recommend.

This creates momentum. Once you’re in, you’re more likely to stay in. The user gets familiar with your brand, the AI treats you as a known good option, and the loop reinforces itself.

The flip side? Breaking in gets harder. If you weren’t cited early, you’re fighting uphill against brands that already have that trust prior established.

Brand, distribution, and consistency aren’t buzzwords anymore. They’re literally how personalization algorithms decide who to keep recommending and who to filter out.

You Need Coverage, Not Just One Perfect Page

Stop trying to build one canonical page that ranks for everything. That’s not how personalized systems work.

You need content across the slices that personalization uses. Industries, roles, tool stacks, regions, constraints. “For enterprise” versus “for startups.” “With compliance requirements” versus “move fast and break things.” “North America” versus “EU data residency.”

Alternatives and comparisons too, so you’re eligible when the user’s context implies they’re evaluating trade-offs. If you only have “why we’re great” content, you disappear from consideration sets.

Yeah, this is more work upfront. But it’s how you stay visible across different contexts instead of only appearing for one narrow slice of users and hoping for the best.

Local and Vertical SEO Just Became Your Secret Weapon

If location and context personalization increases, and all signs point to yes then “best option near me” or “in my country” or “for my compliance regime” becomes a default filter.

Google already does this. Their own documentation talks about how personalization and localization affect results. AI answer engines are going the same direction.

For local businesses and vertical-specific tools, this is actually good news. You don’t need to outrank global giants if you can dominate the personalized context that actually matches your market.

A regional agency in Cleveland doesn’t need to beat the big names in New York. They need to own “Cleveland” plus their vertical. A compliance-focused tool doesn’t need to beat generic project management software. They need to own “HIPAA” or “SOC 2” as qualifiers.

Personalization makes the playing field more fragmented. That’s a threat if you’re trying to be everything to everyone. It’s an opportunity if you’re willing to own a specific slice really well.

The Reality You’re Actually Dealing With

So here’s where we land.

Yes, different users get different LLM answers because of real personalization—memory features, profile settings, contextual signals. OpenAI documents this, Perplexity documents this, it’s happening.

But also yes, a ton of what looks like personalization is just normal system variability. Non-determinism, retrieval inconsistencies, A/B tests you don’t know you’re in, API calls, model updates that ship without announcement.

For SEO, the net effect is the same: you’re moving from single-SERP optimization to probabilistic, cohort-based visibility. Entity clarity, scenario coverage, and brand trust signals determine whether you even get considered in the first place.

The old playbook was “rank #1 for the keyword.” The new playbook is “appear in enough of the right contexts for the right personas that your brand becomes the default answer.”

Is that harder? Yeah, absolutely. More variables, more complexity, less certainty, messier reporting.

But it’s also more defensible once you nail it. Personalization creates momentum. Early wins compound. The brands that figure this out now build advantages that are hard to disrupt later.

So the question isn’t whether personalization is coming to AI search. It’s already here. The question is whether you’re adjusting your strategy while there’s still time to build position, or whether you’re going to keep optimizing for a game that doesn’t exist anymore.

Your call.

Signing Out,
Pankaj & Vaishali.

Two months ago, a friend texted me a screenshot. Her marketing agency didn’t show up at all in ChatGPT’s recommendations for “B2B marketing help.” Her competitor (half the size and charging 30% less) was in the top three suggestions.

The only meaningful difference: the competitor’s entire web presence screamed “SaaS lifecycle marketing for Series A startups.” My friend’s agency? “Strategic marketing solutions for growing businesses.”

That vagueness cost her a $40k retainer. The prospect never knew she existed.

Positioning isn’t marketing polish anymore. It’s infrastructure. The algorithm deciding whether you exist doesn’t care about your versatility or your impressive client roster. It wants one answer: What bucket do you go in?

Get it wrong and you disappear.

And disappearing has a price tag now.

ChatGPT mentions brands 3.2 times more often than it cites them with links. When someone asks for recommendations in your category, the AI either knows what you are and mentions you, or it doesn’t and you’re invisible. Just gone.

The brands getting mentioned? Their AI search traffic converts at 4.4 times the rate of traditional organic traffic. Higher intent, clearer fit, better outcomes. The brands not getting mentioned are losing deals before the conversation starts, watching competitors win customers who never knew alternatives existed.

Here’s the part that should terrify generalists: fewer than 25% of the most-mentioned brands are also the most-sourced. Being everywhere doesn’t guarantee visibility. Being clearly classified as the answer to a specific question does.

So what does positioning actually look like when it works?

The Brands That Got This Right

Google became the verb for search. Not “the leading search engine” or “the most popular way to find information online.” Just search itself. When something’s confusing, people say “Google it.” Not “search for it on a search engine.”

That’s what sharp positioning does. It collapses the gap between category and brand until they’re the same thing.

Kleenex owns facial tissue. Band-Aid owns adhesive bandages. Xerox owned photocopying so hard that people said “xerox this” while standing at a Canon machine. The category disappeared into the brand name.

Software follows the same pattern. Salesforce doesn’t sell business software. They sell CRM. Slack doesn’t sell enterprise communication platforms. They sell team chat. Zoom doesn’t sell video conferencing solutions. They sell video calls. Each one claimed specific territory and defended it until asking for the category meant asking for the brand.

The companies that hedged? Dead or dying.

Yahoo tried being search, email, news, portal, everything. Google killed them by just being search. Evernote tried being notes, tasks, collaboration, web clipping, documents. Notion beat them by being a workspace. Skype tried being calls, video, chat, collaboration. Zoom focused on video and won.

Specific beats general. Every time.

Now here’s where this gets worse for most businesses.

The Classification Game You’re Already Losing

Picture this: an AI organizing a storage unit. It wants neat rows, clear labels. Hand it a box labeled “Sometimes tools, sometimes sporting goods, sometimes kitchen supplies depending on the season,” and it freezes. Can’t classify it, so it shoves it in the back corner.

That’s what happens when you tell the algorithm you do everything.

Web design, paid media, enterprise consulting, local retail support all on the same homepage. You’re not demonstrating range. You’re creating classification paralysis.

HubSpot owned “inbound marketing software” for years. Marketo owned “marketing automation.” Both specific, both clear. Then came the “marketing platforms” trying to be everything at once. Most died. The survivors narrowed down. Mailchimp focused on email marketing for small business. ActiveCampaign doubled down on marketing automation for SMBs. ConvertKit sharpened to email for creators.

Same pattern everywhere. Zoom did video calls, killed Skype doing calls plus video plus chat plus collaboration. Dropbox did file sync, beat Box doing storage plus collaboration plus workflow plus content management. Intercom did customer messaging, beat Zendesk trying to be support and sales and messaging and knowledge base.

This gets worse in a fragmented search world.

Your customers don’t just search Google anymore. They ask ChatGPT, check Reddit, scroll TikTok, try Perplexity. Each platform classifies you independently. Each one decides if you fit the query.

Vague positioning that barely worked on one platform fails catastrophically across five. Google might have given you credit for versatility through backlinks and domain authority. ChatGPT doesn’t care about your domain authority. Reddit doesn’t care about your backlinks. They care what you are, and if they can’t tell, they recommend someone who made it obvious.

Generalists used to win by being the one-stop shop. In 2026, generalists lose by being unclassifiable.

Let me show you exactly what this looks like in practice.

Why “We Do Everything” Translates to “We Are Nothing”

Listing every capability feels smart. You built those skills. Cutting them feels like leaving money on the table.

Here’s what actually happens.

Someone searches for help with a specific problem. Infrastructure management for a software shop. Lifecycle email for a DTC brand. They describe it to an AI, and the AI filters through businesses that match.

“We help organizations scale” triggers a red flag. Scale how? Scale what? Which industry? The AI has 47 other options that gave clear answers. You don’t make the list.

Test this yourself right now.

Open ChatGPT and ask: “What are the best CRM tools for small businesses?” Watch Salesforce, HubSpot, and Pipedrive appear immediately. Now ask: “What are the best business software tools?” Watch it struggle, ask clarifying questions, hedge with generic categories.

Try it with your own category. Ask for the specific thing you do. Do you show up? Now ask for a vague version of what you do. Does anyone show up clearly, or does the AI flounder trying to narrow the question?

That floundering is what happens to your business when you position vaguely. The AI can’t slot you anywhere specific, so it doesn’t recommend you at all.

Specificity isn’t limiting. It’s the filter that gets you into the conversation. “We manage IT infrastructure for software development firms” tells the algorithm exactly what box you belong in. When someone asks for that thing, you show up.

This creates a compound effect. The more consistently you signal one specific thing, the stronger your classification becomes. The AI stops hedging. You become the primary recommendation.

Here’s the part that breaks most people’s brains: narrower positioning generates MORE total traffic, not less.

A software development agency had 23 blog posts targeting keyword variations around “API documentation.” Each post competed with the others. None ranked above position 8. They consolidated into one canonical page about API documentation with supporting content exploring adjacent entities like API design and developer onboarding.

That single page started ranking positions 1-3 for over 40 related queries within 60 days.

The narrow focus didn’t limit them. It concentrated all their authority signals into one clear entity that Google could confidently recommend for everything related to that topic.

Vagueness destroys that confidence. Every time you add “and we also do this completely different thing,” you dilute the signal. Should it classify you as an infrastructure shop or a consulting firm? It can’t answer, so it doesn’t recommend you.

The businesses winning aren’t the most versatile. They’re the most boringly clear about one specific thing they do for one specific type of client.

OK, the part you actually need to understand.

The Entity Optimization Framework

Entity-based SEO undersells what’s happening. You’re teaching AI systems who you are using the same signals humans use: consistency, specificity, repetition.

An entity is a distinct, well-defined thing. A restaurant has cuisine type, price range, location, hours. A software company has product category, target market, technology stack, pricing model. The clearer those attributes, the easier the entity is to classify and recommend.

Building a strong entity means controlling those attributes everywhere they appear. Website, LinkedIn, Google Business Profile, team profiles, case studies, guest posts, podcast appearances.

Here’s where most brands fall apart.

Website says “enterprise software consulting.” LinkedIn says “digital transformation partners.” Founder’s bio says “technology strategist and advisor.” The AI encounters this inconsistency and has to choose which signal to trust, or it averages them into mush.

Start with your core category. Not “marketing” or “consulting” but the specific subcategory you operate in. “Lifecycle email for DTC brands selling physical products.” “Growth marketing for mobile gaming companies.” “IT infrastructure for venture-backed SaaS startups.”

If you can’t define it that specifically, you don’t have a category. You have a vague description.

Define your target explicitly. Not “businesses” but the actual profile. Industry, size, stage, business model. Payroll software for restaurants with 10-50 employees is a different entity than payroll for tech startups with remote teams. Same product category, different entities.

Watch how the winners do this.

Gusto says “payroll, benefits, and HR for small businesses.” Rippling says “workforce systems for fast-growing companies.” Same basic product, completely different positioning. Both winning their lane.

Carta doesn’t do financial software. They do equity management for private companies. Brex doesn’t do corporate cards. They do financial services for startups. Stripe doesn’t do payment processing. They do payments infrastructure for internet businesses.

Each one picked specific category plus specific target. Then they hammered that positioning everywhere. Homepage, case studies, API docs, founder interviews, support pages.

The repetition isn’t lazy copywriting. It’s how entity classification actually works.

Now reinforce that category and target everywhere someone might encounter you. Homepage headline, meta descriptions, team bios, case studies, social profiles, email signatures. Not word-for-word copy, that’s lazy. The same core positioning expressed consistently.

The AI builds confidence. Same category on your website, confirmed in LinkedIn, reinforced by your case studies, echoed in what you write about. Every signal strengthens the classification. After enough exposure, the system stops questioning what you are.

That knowledge gets you recommended. When someone asks for what you do, the AI doesn’t hesitate. You move from maybe to obviously.

Now comes the hard part.

Subscribe now

The Painful Part: Choosing What to Cut

The hardest conversation isn’t about what to emphasize. It’s about what to delete.

Nobody wants to hear their versatility is killing them. Everyone has that story about the weird client project that worked out great, proof they can adapt.

Those stories are true, and they don’t matter.

The algorithm doesn’t care what you could theoretically handle. It cares what you fundamentally are. If 80% of your messaging points to one thing but 20% points to five other things, you’ve added enough noise to confuse the classification.

“But we really do serve enterprise and small business.”

Pick one to position around. You can still take the other work when it shows up, but stop advertising both. Position for enterprise, get enterprise leads. Take small business through referrals if you want, but don’t split your entity profile courting both.

Same for service breadth. Strategy, execution, training, all genuinely good. Choose the one that defines you, that represents the majority of revenue or where you want to grow. Push the others into supporting roles, not the headline.

The fear is “If I narrow, I’ll lose opportunities.”

The reality is opposite. Narrowing makes you visible for opportunities that match. Right now you’re invisible for everything because you’re too broad to classify.

Watch what happens when you tighten positioning.

The leads that come in actually fit. They found you looking for exactly what you do. Qualification becomes easier. Sales cycles shorten. Project fit improves.

The wrong opportunities stop coming, the ones that would have been bad fits anyway. That’s not a loss. That’s filtering working in your favor.

The Boring Clarity That Actually Wins

None of this is sexy. Narrowing your positioning, reinforcing it everywhere, saying the same specific thing until the algorithm classifies you. It feels limiting compared to “we can do anything.”

But expansive doesn’t win anymore. Specific wins. Boring wins.

The brand that plants a flag in one clear territory beats the brand trying to claim the whole map.

The timeline isn’t instant, but it’s faster than most SEO.

Entity consolidation shows initial improvements in 30-60 days as authority signals concentrate. Broader classification that impacts AI mentions and knowledge graph positioning takes 90-180 days as systems verify your entity relationships through sustained consistency.

Brands with highly fragmented positioning see the fastest gains. Going from 23 scattered pages to one canonical entity page produces measurable results within two months. Brands starting from scratch need longer, but the compounding effect kicks in hard once classification solidifies.

Your competitors who figured this out six months ago are already showing up in AI recommendations while you’re completely absent, wondering why ChatGPT doesn’t understand how versatile you are.

The algorithm understands perfectly. You’re too versatile to classify, which means too versatile to recommend.

Pick your lane. Make it narrow enough that the AI knows exactly what box you belong in. Stay in that lane until the system stops questioning whether you belong there.

The brands that refuse this, that cling to positioning as generalists, are about to learn an expensive lesson. In a world where AI controls discovery, being everything to everyone means being invisible to the algorithm.

And invisible means irrelevant.

Pick your lane. Make it boring and specific. Do it this week.

Meanwhile, see you on next thursday.

Signing off
Pankaj & Vaishali

For 20+ years, search was a muscle memory.

Something pops into your head, your fingers hit Google, ten blue links show up. Click one. Done.

Not sexy. Not exciting. Just... there. Like gravity.

That predictability built entire careers. We spent two decades perfecting one loop: rank, click, convert, repeat next month.

Then people stopped restricting themselves to Google only.

I watched my 21-year-old cousin buy sneakers last week. She started on Google, sure, but only to narrow down to 2 or 3 pairs. Then she jumped to YouTube to watch her favorite creator review them.

From there, Pinterest to see how they actually looked on real people in real outfits. ChatGPT to ask which model had better arch support. Reddit to find “honest opinions” and “I’ve worn these for 6 months” threads.

Only after bouncing through five different places did she land on a brand website. And by then? The decision was already made.

What she was doing was search. Just not the search we spent two decades optimizing for.

The funnel didn’t collapse. It scattered.

This Is What Search Looks Like Now

Here’s what happened while we were all busy cranking out content briefs and chasing keyword difficulty scores.

People figured out that different platforms are better at different jobs.

“Explain this concept to me” → ChatGPT, Claude, Perplexity
“Is this thing actually legit?” → Reddit, review sites, communities
“Show me how to do this” → YouTube
“Where do I buy it?” → Amazon, marketplace apps
“Just handle it for me” → Agents (coming faster than you think)

Daniel Kahneman spent years documenting this in cognitive psychology research. People instinctively pick whatever path requires the least mental effort.

And the internet just reorganized itself around that principle, and nobody needed to announce it.

You know, nearly 45% of Gen Z now start searches on TikTok or Instagram instead of Google. Another 40% actively choose them over Google depending on what they’re looking for.

Not because Google got worse (they still own the 89% share in search). Because trust became easier to find somewhere else.

Why Every Platform Wants to Be a Search Engine Now

Picture this from a platform’s perspective.

For decades, Instagram and YouTube and Reddit were doing all the work of creating intent. Someone would see a product, get interested, then leave to Google it. They’d click around, maybe buy, and never come back to thank the platform that sent them.

Platforms generated the demand. Google captured the transaction.

That was fine when search felt like one place. It stopped making sense the moment ChatGPT hit 100 million users in two months and showed people that answers don’t need ten blue links.

Google even issued a Code Red in December 2022. That’s not speculation, that actually happened.

But AI wasn’t the only wake-up call. The creator economy exploded during COVID. Lockdowns pushed everyone online. Time spent on platforms doubled. Content views per post more than doubled. Creators got reach they’d never seen before and just kept producing.

The creator economy sits somewhere between $150 billion and $210 billion right now. Projections have it hitting $525 billion to over $1 trillion by 2030.

So users started asking questions directly inside the platforms while watching creators. And the platforms noticed search was leaking value right through their fingers.

Subscribe now

After that, here’s what went down:

Reddit built native search and AI-assisted answers that pull from years of discussion threads and hand you a single response.

YouTube doubled down on making video the primary way people learn, evaluate, and make decisions. Not just entertainment. Discovery.

TikTok turned into a recommendation engine where every swipe is a query and every video is an answer. Half the time it leads straight to checkout without ever leaving the app.

Meta’s building its own AI search engine to cut dependence on Google and Microsoft. Although, they already replaced the search bar across their apps with conversational search baked into the social feed.

Every platform reached the same conclusion on their own: If users are already searching here, why the hell are we sending them somewhere else to finish the job?

Google stopped being the default for search.

What This Actually Means For SEO (The Part You Care About)

OK, so what the hell does this mean for your actual job?

The hard part isn’t learning new tactics. It’s wrapping your head around the fact that visibility works completely differently now.

A few years back, SEO was beautifully simple. Rank on Google, get traffic, convert. Done. That worked because everything happened in one place.

Not anymore.

Here’s what actually happens now:

Someone’s scrolling YouTube and sees your brand mentioned in a review. Two days later, they’re on Reddit and spot a thread echoing the same take you’ve been pushing. Then they ask ChatGPT a question and your exact framing shows up in the answer.

By the time they land on your site, they’re not comparing meta descriptions. They’re thinking, “OK, these people actually know what they’re talking about.”

That’s the whole game now.

People don’t remember *where* they first saw you. They remember *how often* they saw you saying the same damn thing across different places they already trust.

And here’s the weird part: that repetition doesn’t feel like marketing to them. It feels like expertise. Like you’re just... everywhere. In a good way.

It means, authority stopped living in your website domain.

It’s in the mentions, the citations, the times a creator name-drops you, the language people borrow when they explain your category. It builds every time your voice stays consistent no matter where someone bumps into it.

One viral post won’t do it. One big rankings win won’t either.

What actually works is showing up with the same point of view, over and over, in the places people already hang out. Until it stops being new information and becomes “yeah, obviously.”

And this is where our dashboards start lying to our faces.

When search scatters across five platforms, clicks become optional. Sometimes the answer resolves before anyone visits your site. Sometimes someone decides to trust you long before they ever open your homepage.

That’s what “zero-click” really means. Not that Google stole your traffic (though sure, that too). It means we’re still measuring 2018 user behavior while everyone moved on to 2025.

Traffic didn’t disappear. It just stopped walking through the front door we’ve been watching for 20+ years.

So yeah, SEO evolves. Gets bigger, not smaller. Less about owning the top spot. More about being the name people recognize and trust when they’re finally ready to make a decision.

Wherever the hell that decision happens.

The Uncomfortable Reality of Search Fragmentation

Look, fragmentation isn’t the crisis everyone’s making it out to be.

It’s actually a filter. And honestly? A pretty good one.

It filters for clarity. For people who actually know their stuff, not just people who figured out how to game one algorithm. For ideas strong enough to survive getting repeated fifty times without sounding stupid.

The brands winning right now are the ones that feel familiar by the time someone’s ready to buy.

Search stopped being a place you visit. It’s a state people arrive at after bouncing through five different apps and platforms.

So the only question that matters anymore:

Do we show up in those moments when someone’s mind clicks from “maybe” to “yeah, let’s do this”?

That’s it. That’s the strategy.

Signing off
Pankaj & Vaishali

A few weeks back, I watched someone point their phone at a menu and ask out loud: “What’s the healthiest dish here for a Type-2 Diabetic?”

No typing. No website visit. Just an answer in seconds.

That was my first real encounter with voice+vision search in the wild. And here’s what hit me: this isn’t some 2030 concept we get to plan for. It’s shipping now inside Google AI Mode, ChatGPT’s multimodal interface, Meta’s Ray-Ban glasses, and the next wave of smart wearables and appliances coming in 2026.

The shift is brutal in its simplicity: your surroundings become the search box.

Walking down the street? See shoes you like. Point. Ask. Get alternatives under your budget. The funnel just collapsed from Awareness → Consideration → Purchase to See → Ask → Buy.

If you’re still building SEO strategies around keyword tools and SERP features, you’re optimizing for a world that’s already fading. Let me show you what’s actually happening right now.

The Technology Is Live, Not Coming

Google’s AI Mode accepts voice and image inputs together. You can upload a photo, ask a question about it, and get a structured answer powered by Gemini. Their “Live Search” feature goes further—share your camera feed in real time and ask questions about what you’re looking at.

This is production software. Millions of users are interacting with it today.

Meta’s Ray-Ban glasses already let you look at an object and ask a question. The AI sees what you see, hears what you say, and responds. No screen. No typing. Just existence + curiosity.

OpenAI’s multimodal models combine text, image, and voice in a single interface. You can send a picture, describe what you want to know about it through speech, and get synthesis across formats.

The foundation isn’t theoretical anymore. It’s deployed.

The Money Trail Tells the Real Story

If you really want to know where the world is going, look at who’s writing the cheques.

The venture world has quietly moved billions into technologies that sit exactly at the intersection of voice, vision, and real-world AI interaction.

Not hype cycles. Not speculative research. But companies building the hardware and software that make “point and ask” the default interface of daily life.

Humane AI Pin raised over $230M to build a screenless AI assistant - a wearable device with a camera, voice interface, and no display. The product struggled, but the funding signal is clear: investors believe a camera+voice+AI stack will replace chunks of how we search.

Brilliant Labs pulled in $6M+ to build glasses where “the camera is the new cursor.” You look. The AI understands. An agent responds. They’re not building AR for gaming—they’re building AR for knowing.

Rewind (now Limitless) raised $10M for a wearable pendant that records your day, then becomes a searchable archive of your lived experience, an extended memory.

Meta has poured tens of millions into glasses that see, hear, and interpret context. Their Ray-Ban Meta glasses are the first consumer-grade multimodal search device in the real world. And it’s not tech obsessives using them—it’s regular people.

Rabbit raised $30M to build voice-first agentic interfaces. Right now it’s audio-only, but their roadmap is obvious: cameras are coming, multimodal understanding is coming, real-world context is coming.

Luma AI raised $900M+ to teach models how to understand 3D spaces, not just flat images. If you want a computer to answer questions about the real world, it has to perceive the real world as objects, shapes, and context.

Then there are the appliance giants like Samsung, LG, Bosch, Dyson are quietly embedding cameras, sensors, and voice modules into ovens, washing machines, and refrigerators.

The Pattern Behind All This Money

Zoom out and every one of these bets lands in the same place: a three-layer stack that lets machines perceive the world, understand what you want, and act on it.

Layer 1: Devices that see

(glasses, appliances, 3D capture, cameras everywhere)

Layer 2: Models that understand

(multimodal AI, computer vision, speech models)

Layer 3: Agents that act

(Anthropic Agents, OpenAI agents, Google Gemini ecosystem)

Perception. Understanding. Execution. That’s the full stack for voice+vision search. VCs are funding the infrastructure that replaces the search landscape.

You may ask, Is that a search?

Of course it is. It’s a search without browser.

What This Actually Means for SEO

When search becomes something that happens through voice and vision, the relationship between “website → search engine → user” breaks. There’s a new middle layer now: the AI agent. And it behaves very differently from a human visitor.

Agents don’t browse. They don’t scroll. They don’t skim your hero section to get the vibe. They extract. They compare. They decide.

That forces SEO into a new shape.

1. Your Data Becomes the Real Homepage

In this world, the first impression of your brand doesn’t come from your homepage design. It comes from your data.

An agent doesn’t zoom out to admire your layout. It goes straight into the bones of your information: the numbers, the specifications, the metadata, the ingredients, the labels, the reviews, the structured facts.

Because an agent has one job: don’t get the user the wrong answer.

So it needs your data to be clean, unambiguous, and machine-readable. In a world of voice+vision search, your content isn’t what introduces your brand. Your clarity does.

2. Your Brand Becomes the Ranking Signal

When queries become conversational and contextual, keywords lose power.

A camera feed and a spoken question don’t translate neatly into “best walking shoes for a 60yr old man suffering with arthritis.” They translate into: “Who can I trust to answer this correctly?”

That’s not a keyword question. It’s an entity question.

Search engines, and now agents are no longer matching text. They’re evaluating identity:

• Is this brand who it claims to be?

• Is its information consistent everywhere?

• Do trusted sources talk about it?

• Does it have a clear footprint across the web?

It’s the quiet shift from “optimizing pages” to strengthening your presence as an entity.

3. Trust and Structure Become the New Moat

The more machines participate in the search-and-recommend process, the more the cost of being wrong increases. And when the cost of being wrong increases, the bar for recommendation rises.

This is why in the voice+vision era:

• Accuracy beats cleverness

• Structure beats style

• Consistency beats volume

• Truth beats theatrics

It’s not that creativity dies but it’s alone can’t carry you anymore.

Agents need to be confident that your data is correct, not just roughly correct, not “good enough to get the click,” but reliably true. The brands that can offer that level of clarity will be chosen again and again.

The Quiet Evolution of SEO

When you put these three shifts together: Data as the homepage, entities over keywords, trust as the moat—you end up with a new definition of SEO:

SEO becomes the discipline of making your brand the safest, clearest, most defensible answer a machine can recommend.

Not the most optimized page. Not the most articles. Not the largest keyword cluster footprint.

Just the most answerable brand, and that is where the next decade of organic visibility begins.

What Happens Next

We’re stepping into a world where the internet becomes invisible and intelligence becomes ambient. We won’t “search” the way we used to. We’ll simply ask. And the answers will find us.

Voice+vision search is not the future of search. It’s the first interface of the post-search era where the best answer wins, the best entity is chosen, and the most structured brand becomes the default recommendation.

The technology is live. The capital is deployed. The behavior is forming.

The only question left is whether you’re building for the search box... or for the world itself?

Signing out,

Pankaj & Vaishali

I spent the past month buried in research on agentic AI in ecommerce. Not the chatbot stuff everyone’s been hyping for two years but the actual autonomous agents that plan, compare, and buy on behalf of users across multiple sites without human hand-holding at each step.

The shift isn’t theoretical anymore. OpenAI and Stripe launched a protocol that lets AI agents complete purchases inside ChatGPT. Google, Mastercard, PayPal, and Adobe backed a competing standard. Salesforce shipped their Agentforce platform and McKinsey projects agents could orchestrate $3 to $5 trillion in global retail by 2030.

That’s not a consultant speak. That’s infrastructure getting built right now while most ecommerce teams are still arguing about whether to add a chatbot to their footer. Let’s understand this:

Agentic AI Isn’t a Chatbot

Agentic AI in ecommerce means autonomous agents that handle multi-step workflows: discover products across sites, compare options based on preferences, check inventory and delivery, then execute a purchase or hand you a shortlist.

The chatbots you’ve been testing answer questions. These take action.

Picture this: you need camping gear under $500. Today, you open six tabs, filter by price on each site, compare specs, check reviews, verify shipping times, then finally check out on one. Tomorrow, you tell your agent “build me a camping kit under $500, prioritize weight and weather rating,” and it handles the rest.

Gartner and Salesforce estimate roughly one-third of enterprises will deploy agentic AI by 2028, up from nearly zero today. Your competitors are building this into their roadmaps right now.

How Shopping Actually Changes

Shopping moves from “users navigate to your site” to “agents orchestrate outcomes across multiple providers.”

Your website traffic won’t just decline but a significant chunk will disappear entirely. Customers increasingly interact through conversational interfaces, and many transactions happen without anyone visiting your website. What the industry calls “zero-click” or “invisible” commerce.

Here’s what that looks like in practice.

Commerce shifts from vertical destinations to horizontal, intent-driven ecosystems. Instead of going to Amazon for household stuff and your brand site for specialty items, a single agent handles groceries, fashion, bills, and travel across dozens of providers based on one shopping list or set of preferences.

McKinsey’s $3 to $5 trillion projection by 2030 means a huge fraction of online transactions get mediated by AI agents rather than direct human browsing. If you’re running an ecommerce business and that number doesn’t make you rethink your acquisition strategy, you’re not paying attention.

The Infrastructure Got Built Fast

A few 2024-2025 launches moved this from concept to operational.

Agentic Commerce Protocol (ACP) by OpenAI and Stripe: an open standard letting AI agents discover products, interact with merchant APIs, and complete payments inside ChatGPT. “Buy it in ChatGPT” went from demo to actual checkout flow.

Google AP2 standard: backed by Mastercard, PayPal, American Express, Adobe, and Alibaba. Enables cryptographically signed mandates and auditable agent actions, solving the “how do we trust a machine to spend money” problem at scale.

Enterprise platforms: Salesforce Agentforce, commercetools’ Agentic Commerce, BigCommerce agentic platforms. Packaged capabilities for agent-driven discovery, pricing, merchandising, and support without rebuilding your entire stack.

The infrastructure layer is done. Standardized discovery, high-quality APIs, identity and payment rails, policy guardrails—agents can now transact at scale without anyone rebuilding ecommerce from scratch.

How Merchants Will Use Agents

Retailers are deploying agents as “digital workers” embedded directly in operations. Let’s see how:

Product discovery and personalization

Agents power conversational search. Someone types “I’m going camping in the northwest, build me a kit under $40,000” and the agent assembles options based on weather data, user history, and real-time inventory.

They personalize recommendations and bundles in real time, improving add-to-cart rates and shortening browse time. Dynamic assembly based on intent, not “customers who bought this also bought that.”

Dynamic pricing and inventory

Agents continuously forecast demand, detect slow-moving SKUs, and auto-adjust prices or promotions to balance margin and sell-through. No spreadsheet updates, no weekly pricing meetings.

Inventory agents predict stockouts and trigger reorders or stock transfers across locations with minimal human intervention. The system notices a pattern forming and acts before you run out.

Customer service automation

Agentic service agents connect directly to order systems and policies. They check order status, modify deliveries, issue refunds within constraints, trigger follow-up journeys, a complete end-to-end resolution.

Agentforce case studies show large efficiency gains. Support workloads handled by hybrid agent-human teams at a fraction of previous cost, faster resolution, higher satisfaction. Agents handle the straightforward 80 percent, escalate the complex 20 percent.

B2B procurement

In B2B ecommerce, agents handle quote-to-order flows, contract pricing, approvals, and reorders while respecting permissions, net terms, and audit trails. Complex procurement becomes as frictionless as consumer one-click, but within enterprise controls.

That’s the sell side. The buy side is where this gets uncomfortable.

Consumer Agents Coming for Your Traffic

Personal AI assistants are acting as shopping concierges. AI shopping agents in ChatGPT or other tools research, rank, and shortlist options across multiple retailers, check delivery and returns, then let users confirm or delegate full checkout.

As trust grows, these agents will automate repetitive purchases: groceries, pet food, household supplies, subscription renewals. Users set preferences and budget once, agents handle execution.

The data backs this up.

McKinsey, Adobe, and multiple research polls show a majority of consumers either already use AI for shopping or expect to within 12 months. Not early adopters. Mainstream intent.

But here’s the catch: trust is uneven. Consumers trust retailer-branded or bank-branded agents more than generic assistants. Many aren’t comfortable with fully autonomous checkout for everything yet. “Yet” is doing heavy lifting in that sentence.

SEO 2.0 aka Human + Agent Optimization

When agents do the searching and filtering, the battleground shifts. You’re no longer ranking in human-facing search. You’re being preferred in agent decision logic.

Your product data must be clean, structured, and semantically rich so agents can parse features, constraints, and trade-offs. Sloppy data that humans can interpret through context? Agents skip it.

You need agent-ready APIs and feeds exposing real-time price, stock, delivery, and policy details. If ChatGPT or Perplexity can’t pull your inventory status programmatically, you’re invisible to their recommendations.

Think of it as a new discipline.

You must make your offers machine-readable and competitively attractive so agents select you in multi-variable optimizations: price, speed, reliability, reviews, sustainability, whatever constraints users set.

Traditional SEO optimized for one variable at a time and assumed humans would click through to decide. Agent optimization means competing on the full bundle because the agent is making the decision, not presenting options.

Which raises an uncomfortable question.

Who Owns the Customer

If consumers primarily interact with a general assistant like ChatGPT, Google, or anything else, that assistant becomes the primary interface and gatekeeper. Search engine dependency on steroids, but more concentrated because one assistant handles everything.

Launch your own branded agents on your site and app to maintain a direct relationship, while still participating in third-party agent ecosystems for reach. You can’t afford only one channel.

Even if you solve the ownership problem, there’s the retention problem and which leads us to the next one.

Loyalty

Agents optimize for utility metrics: price, ratings, availability, fit with constraints. They don’t care about emotional attachment.

Loyalty shifts from “I love this brand” to “my agent consistently finds the best combination for me.” Brands move toward subscriptions, replenishment programs, and data-driven personalization that lock in agent-friendly recurring behavior.

Your existing loyalty infrastructure probably won’t cut it.

Your loyalty programs must speak to agents as much as humans. Clear value rules, predictable rewards, APIs that let agents evaluate benefits and redemption. If your points program requires reading fine print, agents route around it.

And then there’s the part everyone’s quietly worried about but not addressing.

The Risk Everyone’s Ignoring

Agentic commerce introduces new fraud and security problems that traditional systems weren’t built for.

Fraud at machine speed: Malicious agents exploit promo codes, test edge cases in returns policies, or perform account takeovers faster than human attackers. You need agent-aware fraud and bot-defense systems, not just CAPTCHA.

Accountability gaps: If an agent makes a poor or unauthorized decision—books wrong travel dates, makes a risky purchase—who’s responsible? User, platform, or merchant? Regulation hasn’t caught up. Everyone’s building fast, the legal frameworks are lagging.

The situation is already complex, and privacy concerns complicate it further.

Data privacy and consent: Agents need deep access to behavioral and financial data. One data breach involving an agent’s purchasing history and you’ve got a nightmare scenario.

Consumer confidence in AI-mediated payments is still forming. Industry groups say adoption hinges on robust identity, explainability, and dispute-resolution mechanisms. Translation: nobody’s figured out the trust layer yet.

So where does this actually go?

2025-2030

Agent-to-site is here. Agents acting as super users that browse sites and APIs on behalf of humans are live in travel, retail, and subscriptions.

The next phase is already taking shape. The Agent-to-agent (A2A).

Where consumer agents negotiating directly with merchant agents while humans mostly setting goals and constraints, grows rapidly later this decade. That means:

• Commerce becomes more invisible.

• Most recurring and low-consideration purchases become hands-off.

• Human attention gets reserved for configuration, edge cases, and high-stakes choices.

So, What to Do Now

Invest in agent-ready infrastructure: APIs, structured data, payments, fraud controls. Waiting until agents dominate a category means you’re already sidelined.

Treat agents as a new customer class. Design products, data, pricing, product descriptions with vivid details so that both humans and their AI representatives find attractive and easy to work with.

Start with the basics.

If your product data is a mess, your APIs are slow, or your inventory feed updates once a day, fix that before you worry about the next ChatGPT feature launch. Agents won’t wait.

What are you seeing in your space? Have you started building your SEO strategy for agent-driven commerce? Hit reply and let me know.

Signing out, Pankaj

In the last issue, we talked about the shifts already shaping how people search. This week, we’re going one level deeper that shapes the forces underneath those shifts. These are the patterns that don’t show up in keyword tools but quietly decide who gets seen and who gets ignored.

We’ll look at model bias, proprietary data, agents, community signals, structured meaning, and why TOFU content no longer works the way it used to.

This is the part of SEO most people don’t talk about, but it’s where the real leverage sits.

Let’s continue.

8. Proprietary Data: The New Ranking Moat

The internet is drowning right now. Rewritten articles everywhere. Recycled keyword guides that all say the same thing. ChatGPT-polished posts that sound professional but say nothing new. Because everything sounds polished, polished content is worth almost nothing anymore.

What still matters? Data. Real data that nobody else has.

I had a call last month with a SaaS founder losing his mind. His content team cranked out three blog posts a week. Clean writing. Good structure. Zero traction. Meanwhile, a competitor with ten total blog posts dominated every AI answer in their space.

The difference?

That competitor published two massive benchmark reports with real usage data from 900 customers. Every answer Perplexity and ChatGPT gave cited that report. Models trust data because it is hard to fake.

Most brands are already generating way too much content, and it is mostly garbage dressed up with decent grammar. Only a few brands can generate internal insights, customer patterns, workflow statistics, real benchmarks. This is why proprietary data cuts through.

Models can summarize a thousand generic how-to posts, but they cannot invent your specific failure rates, your conversion data, your patterns across real websites.

Look at who actually dominates: Ahrefs because of data, HubSpot because of benchmark reports, Shopify because of ecommerce insights. These brands win because everyone references them, and models pick up on that fast. Proprietary data is your moat.

9. Agents: Your Future Users Won’t Always Be Human

SEO has been built around one assumption for decades: a human types a query, reads results, and clicks a link. That assumption is dying fast.

We are moving into a world where agents (personal assistants, shopping bots, research tools) do not browse like humans. They operate. They extract information, complete the task, move on. If your site hides details behind accordions, modals, pop-ups, or “contact sales” buttons, an agent will skip you. Agents cannot tolerate messy navigation. They reward websites that are boringly clear.

When a human searches “best CRM for a 5-person team under $100,” they read, compare, scan reviews, open tabs. Agents filter, match, and output. They look for team size, pricing, features, region. If your page does not provide these specifically, you do not pass the filter.

You are not just losing to better competitors. You are also losing to incomplete information. The pages that win now have structure, not just storytelling.

10. Trusted/Gated Communities

Search engines are not just changing what they show. They are changing who they trust.

Google spent 25 years leaning on websites. Blogs, news sites, content hubs, affiliate networks. But the modern search engines shifted attention somewhere else: real conversations in communities.

Not SEO pages. Not polished guides. Communities.

Go search “best CRM for small teams under $100 per month” in ChatGPT or Perplexity. You may see a few brands that barely rank on traditional Google results. Why? Those brands get recommended constantly on Reddit, Quora, in Discord groups, Instagram captions, on LinkedIn threads. A Reddit thread with 40 comments from actual users carries more weight than a 4,000-word blog post stuffed with keywords.

Because people speak honestly in communities and models use it like a reality filter. Your brand needs to live where conversations happen. Not through spam. Through answering questions, sharing real insights, becoming “the helpful one” in threads.

If people in a community recognize your name, AI models will too. Communities create digital footprints that LLMs treat as high-trust data.

11. Structured Data: The Machine Language of Search

When LLMs or a browsing agent needs to answer “project management tools under $50 per user,” it scans for structured information. If your pricing lives in paragraphs like “our flexible plans start at competitive rates, contact us to learn more,” it gets nothing. It moves to competitors who put pricing in clean tables with columns for tier name, price, user limit, and features. Same goes for specs, eligibility, availability, integrations. Machines need clarity, not prose.

This is bigger than schema markup. Schema helps, but the foundation is how you structure your actual content. Use tables for comparisons. Use clear headings that match what people search. Use consistent labels (not “Plans” on one page and “Pricing Options” on another). List features as bullets, not buried in paragraphs. Show integrations in a grid, not mentioned casually in blog posts.

Humans can decode messy pages. Machines cannot. Pages that machines can easily parse, get cited and selected. Pages they cannot parse get skipped.

12. Multimodal Search

Search used to be a text-only ritual. You typed a question. You scanned links. You clicked one that looked promising. Sadly, that ritual is dying.

We are moving from typing to showing, from asking to demonstrating, from describing to revealing. When something breaks at home, most people do not bother typing “Why is there water dripping under my sink pipe near the U-joint fitting?” They hold up their phone, take a picture or record a quick video, and ask “What is wrong with this?” This is a huge shift.

SEO was built on user language: keywords, variations, synonyms, intents. But when users point a camera instead of typing a sentence, there is no keyword. There is no query. There is only context. And models love context because images and videos contain condition, scene, materials, colors, errors, movement, environmental clues. No keyword research can replicate that richness.

Multimodal search compresses the user journey. Traditionally users moved like this: see problem, describe problem, search, click result, interpret, fix. Multimodal removes half the steps: see, show, fix. The distance between intent and action gets shorter. This is why it is so disruptive. Fewer clicks. Fewer interpretations. Fewer irrelevant pages.

The future belongs to brands that also speak visually, not just textually.

13. AI Model Bias

AI bias does not show up with a warning label. It just works quietly in the background, picking winners and losers based on patterns it has seen before.

SEO historically rewarded quality: better content and links led to higher rankings, suggesting a meritocratic system. AI search operates differently. It chooses answers based on probability and familiarity, not quality.

If your brand barely exists across the surfaces models crawl (websites, communities, reviews, social conversations, documentation), the model has no idea you exist. The probability of you being the correct answer is too low. This is not personal. It is statistics.

Models rely on reinforcement. If 50 sources mention “Ahrefs is the best tool for backlink analysis,” the model locks that in. It does not verify whether Ahrefs is actually the best. It just sees consistency and rolls with it.

The internet works like a high school cafeteria. Whoever gets mentioned the most becomes the popular kid, and AI amplifies that automatically. If people talk about you, you get visibility. If nobody does, you are noise.

LLMs also suppress brands that look inconsistent, unclear, or contradictory. If you show one identity on your website, another on LinkedIn, another on Reddit, and another in directories, the model becomes uncertain. Uncertainty equals invisibility. AI chooses answers it understands clearly.

A brand with clear, repeated positioning will almost always outrank a brand with better content but unclear identity.

Your job is to shift probability in your favor. Not with tricks. With clarity, consistency, repetition, and presence across surfaces. This is visibility engineering.

14. The Death of TOFU Content: And What Replaces It

Just a few years back, the backbone of SEO was “How to” articles. “What is” definitions. “Ultimate guides.” “Top 10 tools.” They worked because search engines needed pages to answer basic questions, but that world does not exist anymore.

AI systems now generate clear, instant answers for almost every basic question. They do it faster, cleaner, and more directly than any blog ever could.

When someone searches “What is technical SEO?” or “How to remove duplicates in Excel?” they do not want a 3,000-word article. They want a direct answer. AI gives it to them instantly. This kills the incentive to click, and without clicks, TOFU becomes a dead asset.

LLMs operate differently from search engines. Traditional Google had to choose one winner for “Best CRM tools.” LLMs do something else: summarize 50 articles, blend their ideas, remove repetition, produce one clean answer. This destroys the advantage of broad TOFU content. Even if you publish the best article, the model folds your work into a summary and credits nobody.

The only content that survives summarization is content that cannot be summarized: experience, data, opinions, failures, frameworks, original thinking, contrarian takes, human nuance.

What replaces TOFU? Firsthand experience (what you tested, what failed, what you learned). Proprietary insights (your data, benchmarks, analyses). Strong opinions refined from working in the trenches. Narratives and mental models that simplify the world. AI cannot fabricate lived expertise. Your clarity becomes the signal.

This wraps up the second half of this series. If the first issue was about the shifts happening on the surface, this one was about the deeper currents that will quietly redefine how brands earn visibility.

Agents, bias, proprietary data, communities, structured meaning, multimodal search, and the collapse of TOFU. These are not trends. They are long arcs. And the brands that understand them early will build an advantage that compounds over years, not months.

Next week, we’ll switch gears. Instead of talking about the future, we’ll talk about execution - how to apply all of this in a practical, day-to-day SEO workflow so you’re not just aware of the changes, but actually moving with them.

See you in the next issue.

Signing out, Pankaj

SEO is entering its most dramatic shift since the early 2000s. Not because “search is dying,” but because the way people look for answers is changing in front of us. The old playbooks that worked for a decade are starting to lose power, and the gaps are widening between brands that adapt and brands that don’t.

This newsletter is my attempt to help you stay on the right side of that gap.

In this first issue, I want to break down the changes that actually matter. Not the recycled predictions, not the buzzwords, but the real signals that are already shaping how visibility, trust, and discovery work today. Think of this as a field report from the frontlines, written for people who are serious about staying relevant in the next era of search.

Let’s start with the seven shifts that are quietly rewriting the rules. These are the foundations. Everything else builds on top of them.

Let’s get into it.

1. Personalization in AI and LLMs

For the first time in internet history, two people can ask the same question and get two completely different answers without any search history, cookies, or personalization settings. Earlier, search engines relied on consistent answers for the same keywords, often tracked using rank trackers. But now, the game is changing.

Although, Google has always done some level of personalization (search history, location, language settings, etc.). But LLMs push this much further. This change is fundamentally driven by how Large Language Models (LLMs) operate because the answers are shaped by who’s asking, how they’re asking, and what the system already “knows” about them.

When I ask an AI model “Best CRM for a SaaS startup?” and you ask the same, the model taps into:

• Our chat histories

• Our phrasing

• Our previous interactions

• The model’s own internal randomness

• The data sources it has implicitly “learned” from

This personalization stems not from “thinking,” but from advanced pattern recognition. As detailed in a paper like Apple’s “The Illusion of Thinking,” these models are primarily designed to predict the next word based on established patterns.

So, All of this adds up to:

The question “What ranks for X?” becomes:
“For which persona, in which context, inside which ecosystem, with which settings?”

The results? Personalized answers.

2. International SEO: Language & Culture

I want to talk about another huge change. For the last 25-30 years, if you wanted to reach people in other countries, you had one main job: translation. You had to take your English words and turn them into Spanish, Japanese, or German. It was hard work.

But today, that barrier is falling down. AI translates better than we ever could. Models like GPT-5, Gemini 3.0, Claude 4.5, and LLaMA 4 can already:

• Translate text with near-native fluency

• Detect and rewrite formality vs informality

• Rewrite tone (professional, friendly, emotional, authoritative)

Now, if a user asks a question in Arabic or Norwegian, the AI will give them a perfect answer in their own language even if the original information was written in English. The AI does the translation instantly.

Does that mean AI translation will do its job better than a human translator? No, Not at all. I mean, obviously you can scale your SEO with multilingual version of your site but the real moat now is nailing the culture nuances.

LLMs detect sentiment but it cannot replicate cultural beliefs, taboos, geo-political tensions, historical sensitivities, locally lived experience, and regional pride into content. The cultural alignment of your brand should be there to win the narratives, audience trust, and get cited in multi-lingual searches.

In short, LLMs may remove the language barrier in scaling content generation but without human involvement you cannot produce regional relevant expertise with cultural richness.

3. Voice Search is Growing

For the last ten-fifteen years, voice search has been disappointing. It was a layer on top of traditional search. Mostly used for weather updates and reminders. We used Ok Google, Siri, Alexa, and to some extent, a few wearable devices. They were useful, but weren’t smart. All were robotic.

That is changing right now. In 2026, voice search will not just be a tool. It will be one of the main ways for people to find answers. Here are a few stats that worth paying attention to:

• 41% of US adults use voice search daily

• 50% of consumers like using voice search to discover local businesses

• Voice search assistants are getting smarter with an average accuracy rate of 93% for answering search queries

Earlier, “Voice SEO” meant only long-tail keywords, FAQ Schema, conversational-style content, but now it’s more than that. We’re seeing AI-powered wearable devices like sunglasses, watches, bracelets, etc that answers your question without typing a single word.

The scary part? No more clicks. They will get the answer and move on.

Voice + Vision gets even wilder. Voice assistants are starting to “see.” You can point your phone camera at an ergonomic chair parts, “How can I assemble this chair?” The AI will look, listen, and tell you how to do it. It is a conversation with a machine that has eyes and ears. Google’s Project Astra is another leading prototype in this space.

Voice Search Will Collapse TOFU, MOFU, and BOFU Into One Step. For example, “Book me a cleaner for tomorrow morning under X budget.”.

It’s one-shot, intent → action.

The winner will be the structured, scannable content, with positive sentiment of users. Voice (and vision) search will become the new operating system of search.

4. Google is No Longer The Only Gatekeeper

I want you to think about how you used the internet this week.

Did you ask ChatGPT a question? Did you look for a recipe on TikTok? Did you check Reddit to see if a product was actually good? Did you look up a person on LinkedIn?

If you did any of these things, you are part of a massive change.

For 20+ years, “searching” meant Google. But, now we’ve entered a multi-engine, multi-interface, multi-platform discovery ecosystem.

Google is still the heavyweight, but it’s no longer the monopoly.

The Internet search has broken into pieces.

Although Google still holds around 89–90% of the global search engine market on the open web, but the share has dipped below the historical 90%+ line for the first time in a decade.

In the old days, you had to follow Google’s rules to be seen. Now, you have more freedom.

This is scary for some people, but I want you to see the opportunity.

The SEO mindset shifts from “ranking on Google” to being present across the internet’s ecosystem.

5. Positioning is The New Title

This is the section where most brands lose the game without even realizing it. Most people think positioning messages are just a marketing slogan, something to print on a business card.

Traditional search is only looking for keywords. If you wrote the word “pizza” enough times, Google showed you to people hungry for pizza. But AI asks a deeper question. It does not just ask what you say, it asks “Who are you?”

Imagine the AI is trying to organize a very messy room. It wants to put every business into a clear, labeled bucket. This is where the trouble starts for generalists.

If you tell the algorithm, “I do everything! I design websites, I manage ads, I consult large corporations, and I help small retailers,” the system becomes confused. It can’t determine your proper classification. Your profile becomes indistinct, and when an algorithm is uncertain, it simply disregards you. If you attempt to appeal to everyone, the system perceives you as appealing to no one.

To win in 2026, you must be sharp. You need to be boringly clear. Instead of saying, “We help organizations scale,” you should state, “We manage the IT infrastructure for software development firms.” When you are this specific, the algorithm is satisfied. It confirms, “Perfect! I know exactly what you do.” It places you in the correct category.

Pick a lane. Stay in it. Tell the internet exactly who you are, over and over again.

In its completeness, we SEO’s call it entity-optimization. Making your brand/product a distinct, well-defined entity with clear attributes, category, and use-cases (entity-based SEO.).

6. Product-led SEO

Gone are the days when SEO success was primarily driven by publishing 4000+ word “ultimate” and “how-to” guides, solely focused on hitting primary keywords and generously scattering secondary ones. Yeah, that playbook is pretty much toast.

For years, the game was simple: write content, rank on Google, get clicks, convert visitors. Companies like HubSpot and Ahrefs crushed it with this approach.

But that relied on people actually clicking through to websites and reading articles. AI search doesn’t work like that anymore. People are directly looking for answers and solutions in the shortest possible manner, and AI models trust brands that show real-world usage, not just brands with lots of content.

For example, if I ask ChatGPT for the “best sunscreen in winter for dry skin in Miami’s weather,” it will likely only suggest products that match at least some of those criteria, such as “dry skin,” “Miami’s weather,” and “winter.” This process involves checking the ingredient list for compatibility with the specified search parameters before presenting the results.

Now, your 4000+ word “Best ultimate guide for Sunscreen” with random product suggestions won’t be of much use.

You can’t fake a publicly available product’s activity at scale.

7. Real-time Search (or Content Freshness)

SEO used to be straightforward. Write solid content, build some links, and wait for rankings to climb. That playbook still works, but AI search engines have changed the game a bit.

Both SEMrush, and Ahrefs, shows in their studies that recency aka content freshness is a major ranking factor in LLMs. Tools like ChatGPT, and Perplexity are constantly scanning the live web. Research shows that content depth, readability, and freshness matter more than traditional SEO metrics like traffic and backlinks when it comes to securing AI mentions and citations.

There are now two timelines running at once: your comprehensive evergreen content that stays relevant for years, and your quick-response content that jumps on what’s happening right now. Miss that second part, and you’re invisible when people need answers most.

LLMs appearing to heavy weigh freshness as a key factor when selecting sources, particularly for time-sensitive topics. In a recent research, 90% of the cited content in AI answers is only 9 months old on average.
Imagine someone asks ChatGPT: “Google’s latest update tanked my traffic. What happened and what should I do?”

The AI will look for recent analysis from known experts, updated guides mentioning that specific update, explanations from trusted sources, and people who’ve made sense of previous updates. It won’t surface 5 year old Google Algo guide.

Closing the First Edition

We’ve covered the first seven shifts shaping the future of SEO. If there is one message in all of this, it is that search is not disappearing. It is just changing faster than most people expect, and the people who adjust early will have a real advantage.

We are only halfway through the full picture.

Next week, we will get into the parts of SEO that most people still overlook. Things like how model bias can quietly push brands out of visibility, why proprietary data becomes a real ranking moat, how agents will change user journeys, and why trusted communities are starting to matter more than high authority links.

See you next week for Part 2.

Signing out, Pankaj