Your Next Client May Have Asked AI First

A while back, I did something that startled me.

I wanted to find a restaurant for hosting a client dinner. My hands moved faster than my brain, and by the time I noticed, I had already typed a whole paragraph into an AI chat box: somewhere quiet, with a private room, around three hundred yuan per person, within two kilometers of the client's hotel...

When I finished, I froze for a second.

How long has it been since I last opened a search engine on purpose?

Go ask the people around you. Trip plans, background research — they ask AI first. That page of blue links a search returns looks more and more like something from a previous era. Right?

Users have changed. So, tell me: where will your traffic come from?

Behind all this is a migration happening right now. Let me walk you through it.

The Layer Underneath Search Got Swapped Out

First, a question: for the past twenty years, what has everyone making content and building brands online been busy doing?

Doing SEO — Search Engine Optimization. Keywords, backlinks, title tags, page speed: every trick in the book to nudge a page a little higher — and higher still — in Google's and Bing's search results.

The logic is simple. A search engine hands you a list; whoever sits near the top gets clicked. It's a race for position.

But now, the list is disappearing.

ChatGPT, Gemini, Claude, Perplexity — these AI products now catch users' questions. Instead of tossing you ten links, they hand you an answer, written out in full. At the end of the answer, they might attach a few cited sources.

Someone put it precisely: SEO is about letting people find you; GEO is about getting AI to cite you.

What is GEO? Generative Engine Optimization. Plainly put, it's the study of one thing: when AI writes its answer, how do you get it to cite your content?

How Does AI Write Its Answers?

To understand GEO, you first have to understand how AI search actually works.

Traditional search engines do "retrieve + rank": crawl web pages, build an index, and when a user searches, the algorithm ranks results by backlinks, keywords, and click behavior.

AI search adds one more step: synthesis. The technical name is RAG — Retrieval-Augmented Generation. The large model first retrieves a batch of material, then works up the answer in its own words, based on that material.

Like what?

Like an open-book exam. AI first pulls a few books off the shelf, opens them, and writes its answer with the books in front of it.

Now think: at this point, what's the one make-or-break question?

How well the books are written is secondary. What matters is: when AI pulls books off the shelf, will it pull yours?

SEO competes on ranking; GEO competes on being written into the answer.

And the way users ask questions has changed completely.

The venture capital firm a16z offered a set of figures in its 2025 analysis: in traditional search, users type an average of just 4 words; in AI conversations, the average prompt is 23 words, and a single session often runs 6 minutes.

Four words looks like: "Vienna vegetarian restaurant."

Twenty-three words looks like: "I'd like to find a vegetarian restaurant with a great atmosphere in Vienna this weekend, somewhere quiet for dinner for two — any highly rated recommendations?"

See the difference? The first is looking something up in a dictionary; the second is asking a friend who knows the territory.

AI will synthesize several sources and sum up a handful of restaurants for you, with reasons attached. If your content happens to be cited, the user has indeed seen you. Even though they may never once have clicked through to your site.

This is "zero-click." The answer is delivered; no link is clicked; no traffic arrives.

The Finding That Stings Most: AI Doesn't Trust What You Say About Yourself

In 2025, several researchers published a paper specifically measuring how AI search actually picks its sources. The conclusion stings a little:

AI engines overwhelmingly favor authoritative third-party content and give the cold shoulder to brands' self-published content.

What does that mean?

Your own site says "we're No. 1 in the industry" — useless; AI won't cite it. An industry publication writes "this company's product currently holds the largest market share" — AI cites that.

Like what?

Like matchmaking. Praise yourself, and the matchmaker just smiles; when your neighbor praises you, when your former boss praises you, the matchmaker takes notes.

AI doesn't look at how you introduce yourself. AI looks at how others introduce you.

This rewrites the logic of the content business. With SEO, the main battlefield was your own turf: tweak your own pages, stack your own keywords. With GEO, half the battlefield is outside: how third-party media mention you, whether industry reports include you, and what identity you hold in AI's knowledge base.

Your brand needs a "household registration" (hukou — China's official registry that proves you exist on the books) in AI's world.

So, What Do You Do?

One thing upfront: don't panic — SEO isn't dead.

Content quality, user intent, page speed — these things matter in any era. Structured data isn't a new invention either; people were doing it in the SEO era, and GEO has simply amplified its importance. SEO will not die at GEO's hands. Search has grown into multiple platforms and multiple forms; AI has become an engine; a new lane has simply opened up.

So what specifically should you do? I've sorted it out for you — roughly five things.

First, get that "registration" fully sorted. Brand, founder, product — use structured data like Schema.org and JSON-LD to fill in every field, box by box. Attach every unique identifier you can: your entity ID on Wikidata, the author's ORCID (a standard author identifier). AI reads your site the way a clerk reads a file. The clearer the file, the more likely it is to be trusted.

Second, structure your content so machines can take it apart. Definitions, data, conclusions, frequently asked questions — write them in blocks, add subheadings, use tables where tables belong, and put a summary at the top of the article. Don't think of this as humans bending to suit machines. Good structure is a pleasure for humans to read, too.

Third, build up "what others say about you." Mentions in third-party media, inclusion in industry reports, reviews from real users — this hard-earned word of mouth is worth more than a hundred pieces you write praising yourself. It is the most valuable asset of the GEO era.

Fourth, walk on two legs. Keep doing SEO to serve the users still on search; layer GEO on top to serve the users who have already moved into AI. Before you publish a piece of content, ask yourself one more question: can AI retrieve this passage? Will it cite it?

Fifth, switch to a new dashboard. You used to watch rankings, click-through rates, time on page. Going forward, you'll also need to watch: how many times you've been cited by AI, and within AI's answers, how big your share of the mentions is. Tools like AI Search Grader already exist for this kind of monitoring. If you don't change your metrics, you won't even realize you've lost the ground.

One more detail, while we're at it. Researchers found in live tests that AI engines favor sources with a presence across multiple languages and continuously updated content. So don't publish and then walk away. For brands going overseas, going multilingual just gained one more reason.

But Every Coin Has Another Side

GEO is not magic. There are a few realities I have to warn you about.

AI's retrieval and citation mechanism is largely a black box. Nobody can tell you why it cites A and not B. What you can do is make your own side solid.

It favors big brands and established media. For small brands and new players, squeezing into the answers is hard. That's unfair, but that's the reality today.

Measurement is a hassle too. Search rankings give you a position to check; whether you appear in AI's answers — there today, gone tomorrow — is far more painstaking to track.

And something more fundamental: users have stopped clicking. Is your business model still built on "clicks"? When your content gets cited by AI, who owns the copyright, how attribution gets counted — the industry itself is still fighting over these questions.

One Last Thing

Back to that evening at the beginning.

I still ask AI first when looking for a restaurant. Chances are, so do you.

This isn't users betraying you. This is users moving house — into the building called AI. However beautifully your storefront is decorated, if the people in the building can't see it, they can't see it.

So stop scrambling for position in the old building.

From now on, your content is written for two readers: one human, one machine. The human decides whether to buy; the machine decides whether you get a chance to be seen.

Understand GEO a little earlier, and get called on by AI one day sooner.

May you be called on soon.