When Users Stop Clicking, Can Your Brand Still Show Up in AI Answers?

A while back, a friend who does B2B marketing complained to me.

He told me he'd discovered something really frustrating. In the past, when a customer searched "how to choose an industrial electricity plan," his company ranked in the top three on Google's first page, steadily pulling in leads. But now, customers aren't clicking on him anymore. It's not that his ranking dropped — it's that customers aren't clicking on any link at all.

Why?

Because the customer just asked ChatGPT. ChatGPT gave them an answer, they read it, and left.

He didn't lose to a competitor. He lost to an "opponent" that doesn't even exist. An AI-generated answer.

This stuck with me for days. Search engines are being replaced by something new: generative engines. And around this new thing, an entirely new discipline is being born — GEO, Generative Engine Optimization.

Today, I want to explain this to you clearly.

First, Let's Clear Something Up: What Exactly Is a "Generative Engine"?

Have you used ChatGPT, Google Gemini, Perplexity, or Claude?

You type in a question, and it gives you an answer. Sometimes it even includes source links. That answer is "generated" by the AI itself — not copied verbatim from any webpage.

AI systems that can generate their own answers are called generative engines.

ChatGPT is made by OpenAI. Gemini is made by Google. Claude is made by Anthropic. Perplexity is built by an independent company. LLaMA is Meta's. You've probably seen these names before.

Let me give you a few real examples. Someone asked, "Who can give me a stable industrial electricity plan?" — ChatGPT listed several energy suppliers and analyzed the pros and cons of each. Someone asked, "Which brand makes good high-end windows?" — Gemini directly recommended a few brands with key review points. Someone asked, "What's the best SEO company?" — Perplexity gave an answer with source links, organizing the core strengths of several companies into a comparison table.

Think about it. In the past, to find these kinds of answers, a user had to open Google, click through ten webpages, and compare everything one by one. Now? AI does it all for them.

They share one thing in common: underneath, they're all large language models, trained by reading billions of articles on the internet. When you ask them something, they use statistical relationships to "guess" the most reasonable answer.

Notice I put quotes around that word. Guess.

Because these models don't actually "understand" what you're saying. They're just calculating: what word is most likely to come after this one. So sometimes they're scary accurate, and sometimes they're spectacularly wrong.

Once you understand this, everything else makes sense.

Three Types of Engines, Three Ways to Play

After looking into it, I found that generative engines fall roughly into three categories. Each has a different "temperament," and the way to get your brand in front of them differs accordingly.

Type one: Pure training-based. For example, Claude (when not using search) and Meta's LLaMA.

These models rely only on what they learned during training. When you ask them something, they pull from "memory." You can't influence them in the short term. All you can do is long-term digital PR, lay down content, and slowly build presence.

Type two: Search-based. For example, Google AI Overviews, Google AI Mode, and Perplexity.

These engines pull web content in real time to generate their answers. To get cited, you need traditional SEO. Create great content, rank well, and get your page selected as a source.

Type three: Hybrid. For example, Google Gemini and ChatGPT Search.

These engines rely half on training knowledge and half on real-time search. Foundational knowledge comes from the model; fresh information comes from the web. You need to nail both.

You see, even though the goal is the same — "get your brand to show up in AI answers" — the strategy is completely different depending on the engine. That's why GEO isn't a one-size-fits-all game.

So What Exactly Is GEO?

I've been talking about engines for a while. Now let's talk about GEO.

GEO is about getting your brand, products, and content to be cited, mentioned, and even directly integrated into AI answers more frequently.

There are several ways this can happen: your brand or product appears directly in an answer; you get linked or cited as a source; your content gets silently pulled in when the AI does background research; or you're consistently indexed by high-quality platforms and eventually absorbed into the model's training data.

How do you make it happen? Here are the core moves.

Publish content on trustworthy platforms — content that has context, structure, and original information. Note the three keywords: context, structure, originality. AI loves information that provides complete background. It doesn't love dry, bare-bones product pitches.

Build presence on platforms that actively feed into AI training. Wikipedia, Reddit, top-tier media. These platforms are the wellspring of AI knowledge. If you have content there, the models will drink from it sooner or later.

Strengthen your authority through digital PR, data-driven thought leadership, and industry commentary. Simply put, you need the AI to see you as an "authoritative source" in your field — not just another random webpage.

Optimize at the technical and semantic level, making your content machine-friendly and easy to extract. Whether your content can be easily read and accurately extracted by AI is a technical craft in itself.

You might say, isn't that pretty much the same as SEO?

It's not.

The goal of SEO is to rank at the top of search results. The user still has to click through to see you. GEO goes a step further. You don't wait for the user to click. Your brand itself becomes part of the AI's answer.

A user asks ChatGPT, "Which brand is good?" The AI says, "Brand X performed well in reviews." The user never clicked a single link, but your brand is already in their head.

That's the most fundamental difference between GEO and SEO: SEO competes for rankings. GEO competes for "AI recommendation."

Someone once said something that really captures it: the future of AI search isn't about competing for rankings — it's about competing for recommendations. If you don't give AI a reason to recommend you, you'll be left behind.

A Key Mechanism: RAG, the Reason AI Can "Look Things Up"

I mentioned earlier that large models sometimes make things up. That's because they rely only on what they memorized during training — their knowledge has a cutoff date and can't update in real time.

So how do search-based and hybrid engines ensure reliable answers? Through one thing: RAG, Retrieval-Augmented Generation.

In plain terms: the AI first goes online, searches around, pulls in relevant content, then combines it with its own training knowledge to construct an answer.

This makes answers far more reliable. First, it can access the latest, verified facts. Second, users can see the sources — transparency goes up. Third, there's no need to frequently retrain the model, which lowers costs and makes output more stable.

This mechanism is hugely important for GEO.

Why? Because RAG means the AI will proactively crawl web content. If your content gets selected, it has the chance to become part of the answer. That's your entry point for "optimization."

Different engines use different search indexes behind the scenes. ChatGPT uses Bing's index. Gemini uses Google's index. Perplexity has its own. This means if you perform well on Bing, it helps with ChatGPT's search-based answers. If you perform well on Google, it helps with Gemini.

What the Claude Leak Taught Us

In May 2025, Claude's system prompt was leaked. This was incredibly valuable.

Why was it valuable? Because for the first time, it gave us a clear look at when an AI model decides to "look things up" and when it doesn't.

The answer: by default, Claude doesn't search the web. It answers directly using its own training knowledge. It only triggers a search — and therefore the possibility of citing external sources — when a question involves time-sensitive information, requires multi-perspective analysis, or falls outside its training scope.

What does this mean?

If the content you write is stable, encyclopedia-style knowledge — like "what is brand positioning" — the model can answer that itself. It won't bother searching. It won't cite you. You get zero exposure.

Exposure only happens where the model "genuinely needs" external content. Things like real-time price comparisons, the latest market data, original research findings, frequently updated tool pages. These are things the model can't answer on its own — it has to search. That's when your content has a chance to surface.

You need to make your content specific enough, fresh enough, or irreplaceable enough. Interactive tools, real-time market comparisons, frequently updated price tables, original research reports, first-hand experience reviews. These are the kinds of content AI can't avoid. Only then will your content be searched, mentioned, and linked.

So the first step in GEO isn't thinking "how do I write more beautifully." It's thinking "what content do I have that AI can't answer without searching for?"

Has Traffic Actually Changed?

Okay, now for the most practical question: has generative AI actually killed search traffic?

A lot of people think ChatGPT will replace Google. But the real picture is more interesting than you'd imagine.

Gartner once predicted that traditional search engine query volume would drop 25% by 2026. Gartner analyst Alan Antin said generative AI is becoming a substitute answer engine — questions people used to ask Google, they now ask AI directly.

That prediction was only half right.

Search volume hasn't dropped. According to Semrush's data, people are actually searching more on Google. ChatGPT hasn't replaced Google — it has generated additional search demand.

But click-through volume is indeed dropping sharply.

Ahrefs' research has shown this clearly. After AI Overviews appeared, users clicked on fewer links. In May 2025, Google AI Mode launched across the US. By early October of the same year, it became available in Europe too. It's a conversational search experience where users can complete the entire purchase decision journey inside the AI interface — no need to jump to any website.

Why?

Because in the past, when a user searched "best laptop," they had to open ten webpages and compare for themselves. Now they just ask AI: "What laptops under 1,500 euros can edit video, play games, and have long battery life?" AI organizes it all in one shot.

Users stop clicking.

A McKinsey study confirms this trend: purchasing decisions increasingly happen before the user ever reaches your website. Attribution is getting harder. The user might go directly to a store. They might search your brand name. They might have been recommended by AI — but you can't trace it at all.

Honestly: clicks and traffic were never good KPIs to begin with. Over 97% of traffic on most websites never converts. AI search has simply exposed a truth that was always there, just papered over.

Here's our take:

Generative AI is changing how people search, not whether they search. Total search volume is growing, but behavior has shifted. Faster, more conversational, and often zero-click. The main battleground for brand exposure is migrating from websites to AI answer interfaces.

So Should You Actually Do GEO?

My answer is simple: Yes.

This isn't the time to hesitate. Look at these facts:

Google AI Mode has launched in over 200 countries. ChatGPT gets more than 3.8 billion visits per month (data from Similarweb). Generative engines are growing faster than you think.

But how exactly should you invest? My advice depends on your situation:

If your SEO is already in good shape, allocate an additional 20% to 25% of your existing SEO budget to GEO. You're already standing on a solid foundation — GEO is adding another story on top.

If your SEO is still catching up, do SEO first. GEO is built on top of SEO. If the foundation isn't solid, building on top is just throwing money away.

Seer Interactive did an analysis and found a correlation of about 0.65 between a brand's first-page Google ranking and how frequently it gets mentioned by large language models. (Of course, correlation doesn't equal causation.)

SEO, content marketing, digital PR — these things aren't outdated. They're the foundational infrastructure for winning AI exposure.

Before You Start, Think Through a Few Things

If you've decided to start doing GEO, don't just dive in and start changing content. First, take the time to think through a few things.

First, in your industry, who is AI actually citing?

Go into ChatGPT, Perplexity, and Gemini. Search the most common customer questions in your industry. Look at which brands get mentioned in the answers and which sources get cited. Map out the patterns. Which brands come up again and again? What did they do to get cited? These insights are your starting point.

Second, why does the content that gets cited actually get cited?

Everyone writes articles. So why do some get picked by AI as sources while others get ignored? Study the content that does get cited. Look at its structure, format, information density, and timeliness. Generally speaking, highly structured, specific, and continuously updated content is more likely to get crawled and cited by AI.

Third, what can you do to get cited more?

Push from several directions: on the content level, nail your answer format and relevance, and build a solid page structure. On the structural level, improve your site architecture and make content zones clearer. On the external level, do digital PR, spread across knowledge platforms and aggregators, and expand your digital footprint.

There's a trap to watch out for. Don't think doing some basic SEO and tweaking titles is enough. The logic of AI search is different from traditional search. You have to give AI a reason to "recommend you." Is your content specific enough? Fresh enough? Irreplaceable enough? If AI can answer the question on its own, why would it come to you?

A Few Final Words

My friend later came back with some feedback.

He said he followed this line of thinking and re-audited his content assets. He cut out the generic, encyclopedia-style articles and focused his energy on content AI couldn't answer. Real-time price comparisons, exclusive research data, first-hand experience reports.

A few months later, he noticed ChatGPT and Perplexity had started citing his content. No surge in clicks, but his brand was starting to appear in AI answers. Customers would call and say, "I searched on ChatGPT and you got recommended."

He said he suddenly understood something: in the age of AI, the definition of exposure has changed. It's not exposure only when someone clicks your link. AI mentioning your name — that's exposure.

Search engines have changed. The intent to search hasn't. Users are still looking for answers — those answers just show up in a different place now.

The old playbook was about trying to be everything to everyone, the one-size-fits-all answer. That era is over. Now it's about delivering the right information, to the right person, at the right time. AI isn't a tool that makes you more generic. It's an opportunity to make you more precise.

You only need to ensure one thing. When the AI generates that answer — your name is in it.