When users stop clicking links, can your content still be seen?

A few days ago, a friend complained to me.

He said he'd spent a fortune building his company website and had finally fought his way onto the first page of Google, into the top three. And the result? A customer came to him and asked, "How's your product?" He shot back, "Didn't you look at our website?" The customer said, "I asked ChatGPT. It recommended another company."

He was stunned.

I froze for a second too. And then I realized something: search is going through a fundamental change.

And the vast majority of people doing content and marketing haven't reacted yet.

Search has changed — completely

Think back. What did your last Google search look like?

You'd type in two or three words, hit enter, and get back a page of blue links. You'd click the first one, take a look — nope, not it — back out, click the second one. Back and forth a few times until you finally found your answer.

That was search for the past twenty years.

And now?

You open ChatGPT and ask it: "I run a small e-commerce team under ten people with a tight budget — which email marketing tool should I pick?"

It doesn't hand you a list of links.

It just gives you the answer. It pulls together features, pricing, and user reviews from several products, serves it up in a paragraph, even builds a comparison for you, and ends with a recommendation. A few citation links trail at the bottom.

You got your answer. But you didn't click a single link.

That's what's called "generative search." The user no longer has to dig through web pages to find information — AI does the finding for them, the reading for them, the summarizing for them.

What does that mean?

It means that if your content isn't selected by AI, isn't cited by it, then in the user's eyes, you don't exist.

So what exactly is GEO?

GEO — short for Generative Engine Optimization.

In plain terms, it's one sentence: get your content cited by AI search engines.

With traditional SEO, you obsess over "how do I get onto the first page of Google." With GEO? You're thinking "how do I become part of the AI's answer."

Those are two completely different goals.

SEO chases position. GEO chases mention frequency.

Think about it: ask ChatGPT the same question five times and you might get five different answers. It's not Google's "number one is number one" fixed ranking. Every single time, it reorganizes the information from scratch.

So in the world of AI search, there's no such thing as "ranking first." What there is, is this: across how many different ways of phrasing a question does AI mention your brand?

That frequency is your AI share-of-voice (mention frequency). The higher it is, the more chances you have of being seen by users.

How do AI search engines actually work?

This step is critical. You have to understand how AI finds information before you can make it find you.

Let me break it down for you — here's what happens behind the scenes when a user asks AI a question.

Step one: decompose the question.

AI doesn't throw your whole sentence at a search engine as-is. It first splits your question into several smaller questions and searches them separately.

For example. Someone asks AI: "What's the best VPN for watching Netflix in Europe that can handle streaming?"

AI won't just search that whole long string. It splits it into "best VPN 2026," "VPN for watching Netflix," "European VPN servers" — searches each one, then stitches the results together.

Step two: fetch the content.

AI pulls relevant passages from the web and from its own knowledge base. Most AI uses a technology called RAG (Retrieval-Augmented Generation)（a method that extracts specific passages from web pages and feeds them to a large language model as raw material）.

Step three: synthesize the answer.

AI weaves information from multiple sources into a single coherent passage. It's not copy-paste — it's reorganization.

Step four: add citations.

Citation links to the sources are appended to the answer.

Notice something? Here's the key point.

Your content doesn't need to answer the user's long question. Your content just needs to get found by one of the smaller questions AI decomposes out.

That's the biggest strategic difference between GEO and SEO. Doing SEO, you think about what keywords the user will search. Doing GEO, you think about what smaller questions AI will split the user's question into — and then you make sure your content hits those smaller questions.

What's the relationship between GEO and SEO?

A lot of people's first reaction when they hear GEO is: is SEO dead?

No. Nowhere close.

Let me tell you a few things that haven't changed.

First, E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness)（the four-part quality framework Google uses to judge content） is still king. Experience, expertise, authority, trustworthiness — Google values this stuff, and AI values it too. AI only wants to cite you when what you've written has real substance.

Second, the technical fundamentals still matter. Fast page loads, mobile-friendly design, crawler accessibility — none of those old rules have been retired.

Third, content quality is still hard currency. Those keyword-stuffed filler articles? AI looks down on them just the same. What AI wants to cite is content that's genuinely useful and has real depth.

Fourth, backlinks still work. AI uses real-time web search to gather material. Pages with strong backlinks get picked up by AI more easily. Plus, lots of backlinks means you show up frequently in a dataset called Common Crawl（the public web crawl that most large language models are trained on）. The more the training data has "seen" you, the more AI "recognizes your face."

But a few things have genuinely changed.

Change one: how users phrase questions has changed. The average Google search is 4 words. The average AI search is 23 words. Users describe their situation in full detail, the way you would talking to another person.

Change two: how success gets measured has changed. SEO looks at rankings, click-through rate, traffic. GEO looks at citation count, brand mentions, share-of-voice.

Change three: a high Google ranking doesn't mean AI will cite you. Research has found that the overlap between links on Google's first page and the sources AI actually cites has already dropped from 70% to below 20%. AI has its own preferences.

In a sentence: SEO is the foundation of GEO, but the foundation alone isn't enough — you still have to build the GEO floor on top of it.

Why you have to start in 2026

You might feel that AI search is still niche, no rush.

Let me give you some numbers.

ChatGPT has over 800 million weekly active users. Google's AI Overviews show up in billions of searches every month. Perplexity handles millions of queries a day. Apple has baked AI search directly into Safari.

These aren't "future trends." This is happening right now.

And the behavior of AI search users is completely different.

On Google, users bounce in seconds. In AI search, users stay an average of 6 minutes per session. Six minutes. The user is genuinely talking with you — digging in, following up.

Here's the more critical part. Traffic coming from AI search is smaller in volume than Google, but the conversion rate is higher. Why? Because the user has already gotten a recommendation from AI. When they click through, they arrive with purchase intent. Vercel says 10% of their new signups now come from ChatGPT recommendations.

AI search traffic is the kind that's small in volume but high in quality.

If you don't do GEO now, your competitors are already doing it. By the time they've filled up their AI share-of-voice, it'll be too late for you to catch up.

How? Six concrete actions

Alright, principles covered. Now the practice. Here are six actions — just do them.

Action 1: First, make sure AI can read your content

This one sounds like a no-brainer. But it's the most common problem we've seen.

Go check your robots.txt file. Lots of sites, without realizing it, have blocked AI crawlers. Cloudflare recently changed its default settings to block AI bots outright. If you use Cloudflare, your AI traffic has probably already been cut off automatically — and you don't even know it.

Go into your server logs and search for the "ChatGPT-User" identifier to see whether AI crawlers have actually visited.

There's another trap: JavaScript rendering. AI crawlers aren't human — they don't click, don't scroll, don't switch tabs. If your pricing page uses interactive sliders or tab switching to display different plans, then that content is invisible to AI.

If you want AI to cite you, first let AI be able to read you.

Action 2: Structure your content so AI can extract it

AI plucks information from your page. The easier you make it to extract, the higher your odds of getting cited.

How?

Use clear heading hierarchy. H1, H2, H3 — one topic per level.

Use lots of lists and numbered items. Someone analyzed ten thousand real queries and found that pages with structured lists, citations, and data saw 30% to 40% higher visibility inside AI answers.

Put the conclusion first in every paragraph. Don't bury the answer after a long windup. What AI is hunting for is information blocks it can lift directly.

Keep paragraphs to two or three sentences. AI struggles to slice up big, dense blocks of text.

Action 3: Aim at the sub-questions AI will actually search

Remember "decomposing the question" from earlier? AI splits the user's long question into smaller questions to search.

You have to think through: inside the user's long question, what smaller questions will AI break out? And then make sure your content answers those smaller questions.

For example. The user asks: "For a small e-commerce store under ten thousand subscribers, which email marketing platform should I pick?"

AI might split it into: "best email marketing platforms 2026," "e-commerce features of email marketing," "email marketing pricing for small businesses."

You have to make sure your page has content that hits each of those three directions separately.

Action 4: Put authority signals out so AI trusts you

When AI decides who to cite, it evaluates the credibility of the source. You have to give it signals.

Include expert quotes — with name, title, and company. AI reads that as a strong authority signal.

When you use data, label the source. "According to Semrush clickstream data（the record of which links users click across the web）" carries far more weight than an assertion with no origin.

Showcase first-hand experience. Your own real cases, observations, and data. AI weighs this more and more heavily.

Make clear who the author is and what their credentials are.

Action 5: Stay fresh

This one is extremely important.

AI has a strong preference for freshness. We see a very steep curve in the data: once content is more than 3 months old, AI's citations of it fall off a cliff.

3 months is the cutoff.

Every quarter, at minimum, pull out your most important content and refresh it. Update the data, swap in new cases, add new developments.

Action 6: Build authority off-site

This is the easiest one to overlook.

AI doesn't only learn about you from your own website. It learns about you from the entire internet. Getting mentioned on third-party sites, getting discussed in forums, getting recommended on Reddit — all of it shapes how AI describes you.

Even brand mentions with no backlink are weighted by AI.

The fastest trick: find the pages that AI already cites frequently, and figure out how to get your brand onto those pages. Go leave thoughtful comments in relevant Reddit threads, email blog authors and ask to be included. Someone used this method and went from completely invisible to AI to getting their first mention in under an hour.

The fastest GEO shortcut isn't optimizing your own site — it's wedging yourself into the pages AI already trusts.

What's different about each AI search engine?

The core principles are universal. But each platform does have its own temperament.

ChatGPT holds the largest market share — roughly 70% of AI search. It mixes real-time web search with its training data. It likes content that's comprehensive, clearly sourced, and rich in professional signals. And it's increasingly able to deliver real recommendation traffic.

Google AI Overviews blends traditional search ranking signals with AI synthesis. Rank well on Google and your AI Overviews visibility probably won't be bad either. Structured data markup earns bonus points.

Perplexity cares most about citations — it searches the live web, favors fresh content, and is the most transparent about sources. SaaS products convert surprisingly well on Perplexity.

Gemini is the fastest-growing platform. It's deeply tied to Google's search infrastructure. Do Google SEO well and your Gemini visibility will naturally be solid too.

Claude leans toward synthesizing information rather than citing directly. It likes content that's clearly structured and logically strong. Apple has already announced integrating Claude into Safari — which means its influence is about to climb sharply.

Don't step in these traps

The GEO face-plants I've seen fall into just a few categories.

On the content side: keyword stuffing — AI hates this just as much as Google does; content too thin, no depth — AI won't cite it; information outdated — past 3 months without updates, citations crater; data without a source — credibility takes a hit; big dense blocks of paragraphs — AI can't slice them up.

On the technical side: robots.txt blocking AI crawlers without you knowing; important content loaded dynamically via JavaScript — invisible to AI; content hidden behind logins, paywalls, or collapsed menus; pages loading too slowly for crawlers to grab.

On the strategy side: treating GEO and SEO as two separate things when they're actually one; mass-publishing AI-spammed articles — a negative for both SEO and GEO; only optimizing your own site and ignoring off-site; still using the old Google Analytics setup, which can't see the full picture of AI search.

How do you measure whether it's working?

This step is the easiest to overlook, but also the most critical.

Traditional SEO metrics — rankings, clicks — basically fail in the face of AI search. Because most AI search is zero-click. Users get the answer inside AI and leave — they never come to your site.

You have to switch to a different set of metrics.

Share-of-voice — the most important. It measures: across a large set of different ways of phrasing questions, how frequently does your brand appear in AI's answers? Think of it as your "mention rate." The higher this percentage, the more AI exposure you get.

Competitive ranking. Just looking at yourself isn't enough — you have to compare against competitors. Where do you rank in your industry for AI visibility?

Citation tracking. Which specific pages are getting cited by AI? How many times?

Brand description accuracy. How does AI describe your brand? Is it accurate? Is it flattering?

AI referral traffic. Check server logs for "ChatGPT-User" to see whether AI crawlers have shown up.

If you don't want to pay for tools, there's a simple, low-tech approach.

Pick 10 to 20 questions relevant to your business — especially the ones users ask when making a purchase decision. Take those questions to ChatGPT, Perplexity, and Gemini. See whether your brand shows up, how it gets described, and what gets cited. Do it once a month and track the changes.

Clunky, but it works.

What happens next?

The GEO playbook is still early. The manual is being written as we run. But a few directions are already fairly clear.

AI search will keep growing. Apple has put it inside Safari, Google is expanding AI Overviews, ChatGPT is connecting to e-commerce. The number of users will only go up.

GEO and SEO will increasingly look like one thing. Not one replacing the other — merging. SEO feeds GEO, and GEO in turn helps SEO validate content direction.

Freshness weighting will keep rising. That 3-month cliff will become a rule everyone knows.

Multimodal will rise. AI is getting better and better at understanding images, video, and audio. Brands that invest in multiple content formats now will have the advantage later.

Personalization will change the game. The same question asked by different people may get different answers. That means you have to cover the same topic from multiple angles to capture visibility across different audiences.

One last thing

Back to my friend from the beginning.

What did he do next? He spent two weekends going through his website's robots.txt and unblocking AI crawlers. He restructured his core product pages — using lists where lists belonged, front-loading answers where answers belonged. Then he went to Reddit and a few industry forums and seriously answered a dozen-plus relevant questions.

A month later, he told me ChatGPT had mentioned his product for the first time.

Just two weekends' worth of work.

The barrier with GEO isn't how hard it is — it's when you start. The earlier you act, the more you're building on empty ground with no one competing. Once everyone else wakes up, it gets crowded.

I don't know what AI search will ultimately evolve into. But there's one thing I'm fairly certain of: the way users look for answers has already changed. And where the users are, your content has to be.

That's my judgment — not necessarily all correct. But this is something worth starting to think about right now.