You're Still Grinding at SEO — Users Have Already Stopped Clicking Links

A recent piece of data made me uneasy.

One platform tallied two million conversations with large language models (LLMs) and uncovered something counterintuitive: when users chat with AI, they barely click links at all.

How bad is it? Plenty of highly exposed links had a click-through rate of 0%.

You read that right. Zero.

What does that mean? It means the SEO playbook we've all run for the past decade has already started to fail in this new world.

A New Word: GEO

When I talk to friends in marketing, everyone is tossing around a new term — GEO, short for Generative Engine Optimization.

What's a generative engine?

It's that chat box you use to ask ChatGPT, Perplexity, Doubao, or Gemini a question. It doesn't hand you ten blue links to pick from — it just hands you an answer.

What the user gets is an answer, not a link.

Which is awkward. The article you slaved over, the page you built, the keywords you stuffed — to an LLM, they may simply get boiled down into a couple of sentences inside the answer.

And those sentences might not even mention your name.

An Overlooked Truth: Whoever Gets Cited, Wins

One research group scanned 25,000 URLs to see whose content the large models actually cite when they answer questions.

They found a few interesting patterns:

First, LLMs love to cite list-style content. Those "Top 10 best XX" or "5 YY worth buying" articles with a clean structure. Why? Because the answer is already laid out — the AI just grabs it.

Second, original research and tool-type content gets cited far more often than generic explainers. Write "What is marketing automation?" and nobody cares. Publish an industry report with real data, and the AI scrambles to cite you.

Third, there's a huge gap between Google and ChatGPT when it comes to citing retail brands — Google only 4%, ChatGPT as high as 36%. Ask the same question, and the two give completely different answers. Which means you can't optimize for just one platform.

But Here's the Trap

A strange phenomenon has surfaced in the industry lately.

Some brands figured: if whoever the AI cites wins, can't I just "buy" the citation?

And so a market has emerged for paying to get AI to recommend you. One SEO heavyweight wrote an article saying outright that this practice is blurring the line between "legitimate optimization" and "ranking manipulation."

In plain English, this is the new version of black-hat SEO.

Remember how Google used to treat sites that bought backlinks? It deindexed the whole site. The AI platforms haven't acted yet — but sooner or later, they will.

Why? Because AI's lifeblood is users' trust in its answers. The moment users realize "oh, the AI's recommendations are paid placements too," trust collapses. The AI companies care about this more than anyone.

So don't take shortcuts.

A Counterintuitive Conclusion: Brand Matters More Than Content

A team that researches AI search has been tracking how various brands perform when AI mentions them. They reached a conclusion I'll pass straight to you:

Most GEO tactics don't actually move the needle. What really determines whether a brand shows up in an AI answer is brand positioning, category mindshare, and how third parties talk about you.

In plain English:

AI doesn't care how well your page is optimized — it cares about "how other people describe you."

If the whole web says you're "the best project management tool," then when a user asks an AI "any good project management tools to recommend?", you'll most likely show up.

If nobody says that about you, no amount of optimization will help.

Which brings us back to an old question: building a brand is something you can't sidestep, no matter how the technology changes.

You used to say "I need to do SEO." Now you say "I need to do GEO." But the underlying logic hasn't changed — you first have to become a brand that other people actually want to mention. Only then can AI mention you.

So Should You Still Do SEO?

Yes. But the role has changed.

Old SEO was "let users find me." New SEO is "let AI understand me."

How?

First, content should be "answer-first." Don't bury the lede — put the core conclusion in the first paragraph. When AI crawls your content, it only looks at the first few hundred words.

Second, keep your structured data tidy. FAQs, tables, lists, comparisons — these are the easiest things for AI to digest.

Third, do more original research. Surveys nobody else has done, exclusive data, first-hand case studies — these are the reasons AI has to cite you.

Fourth, mind multiple platforms. Google, Bing, ChatGPT, Perplexity — they draw on completely different content sources. One case study found that some top brands had no presence on Bing, and then completely vanished from ChatGPT's recommendations. Because ChatGPT leans heavily on Bing's content sources.

You say, "I'll do all of them." But if you skip Bing, you've given up on the ChatGPT lane.

One Last Line

Marketing used to be about "getting users to search for you."

Marketing going forward will be about "getting AI to mention you."

Users don't click links anymore — AI clicks for them.

You think you're losing clicks? No — you're losing being seen altogether.

This is something worth rethinking from scratch, tonight.