Forget SEO. The new battlefield is called GEO.

The other day a number popped up in my feed and stopped me cold.

This year, traffic to retailers from chatbots and AI search engines is up 520% year-over-year.

520%. You read that right. Not 5%, not 20% — more than fivefold.

I went straight to Adobe's shopping report to verify, and sure enough, it's their number. Then I spotted another piece of news: last week OpenAI signed a big deal with Walmart, and now you can order Walmart products directly inside the ChatGPT chat window.

Think about what that means.

It means people are starting to skip Google and ask ChatGPT instead.

And "people" is going to mean a lot more of them this year.

Which is why every retailer is suddenly sitting up and taking notice.

A new business worth 850 million RMB

For the past few decades, if a brand wanted to be found, there was really only one thing to do.

SEO. Search Engine Optimization.

Find a way to push your web page to the top of Google's search results — the higher the better, ideally above the fold on page one. This single game has fed an entire cottage industry of consultants, agencies, and blogs.

And now?

Now the problem is that users have changed their habits and gone to ask the AI. The AI doesn't give you ten blue links — it just reads the answer out loud to you. So if a brand still wants to be mentioned, what does it do?

A new term has surfaced. GEO.

Generative Engine Optimization.

In plain English, it's "how do you get the AI to mention you in its answer."

I dug into it, and this little industry that didn't even exist last year is on track to hit a market size of roughly 850 million RMB (about USD 120 million) this year, by one market research firm's reckoning. GEO consulting shops are springing up one after another.

The funny part is, I went and poked around the backgrounds of these GEO consultants. Guess what?

Almost all of them came over from SEO.

Old hands, new battlefield

Here's what's interesting.

GEO may be a brand-new term, but the people doing the work are, in their bones, the same crew who were doing SEO for you a few years ago. The line they love to trot out is:

"At its core, GEO is the next stop after SEO."

Is that true?

Right — and wrong.

The part that's right is that the goal is still the same goal: you have to guess what users will ask, and then get your content into the answer. That hasn't changed.

The part that's wrong is that the answers the AI gives and the links Google serves up overlap shockingly little.

Let me give you a number.

The head of a GEO company called Brandlight told me: a year ago, the sources the AI cited and the pages ranking at the top of Google overlapped about 70%. In other words, the AI was basically just reading Google's top ten back to you.

Today, that overlap has dropped below 20%.

In other words, of the ten answers the AI now serves up, eight are things you would never find on page one of Google.

What does that mean?

It means the old playbook of "squeezing into Google's top ten" is failing, fast.

The AI doesn't like your long-winded articles

So what does the AI actually like?

After talking to several people doing GEO in the trenches, this was the most interesting thing I heard.

The AI doesn't like long-winded essays. It likes short, structured content you can lift and use as-is.

Think about the Google-era logic.

You search "how to make braised pork belly," and the page-one result is probably a two-thousand-word blog post. It opens with a story about the author's grandmother, then a section on how to shop for pork belly, and only then lazily gets to the recipe. Why write it that way?

Because Google's algorithm rewards long content. More words, longer dwell time, the right keyword density — rank goes up.

But the AI isn't buying it.

What the AI actually likes is content like this:

• An FAQ with dozens of questions, each answered in a few dozen words;
• A comparison table — Product A vs. Product B — with battery life, price, and specs laid out at a glance;
• A spec sheet where every line can be lifted out and read directly to the user.

The Brandlight executive gave an example that I think nails it. He said:

"An article bragging about how great your brand is can answer one question, at most. But an FAQ can answer a hundred."

An FAQ is, in effect, a hundred hooks baited for the AI. Whatever it asks, you've got a hook that can catch it.

The more specific the question, the bigger the opportunity

But the truly counterintuitive part of GEO is somewhere else.

What's really counterintuitive is that the questions users ask an AI are far more specific than the ones they ask Google.

Think about the Google era — how did people search? "Best SUV," "cheap phone," "intro to personal finance" — one broad phrase tossed in, and then you sift through the results yourself.

Over on the AI side?

A user asks: "Chevrolet Silverado or Blazer — which has more range?"

Or: "For 35-year-old sensitive skin, which serum won't break me out in summer?"

Or: "I'm installing a 3-ton AC unit on the second floor — what's the maximum outdoor unit noise in decibels I should accept?"

See the pattern?

Users aren't asking "who are you" anymore — they're asking about a specific model, a specific scenario, a specific number.

So here's the problem. That website copy of yours that reads "our company is a dedicated enterprise committed to delivering exceptional products and services to our customers" — can the AI use that to answer any of the questions above?

Not a single one.

So the brands that will win in the AI era aren't the ones writing more. They're the ones writing more specifically.

For every specific question, give it a specific, standalone, lift-and-use answer. That is the core move of GEO.

Nobody is counting ROI yet — and by the time they try, it'll be too late

At this point, you might be asking a very practical question.

Is this worth it?

I put the question to a client in the industry — Brian Franz, Chief Technology Officer at Estée Lauder. Here's what he said, verbatim:

"At this stage, I don't want to look at the ROI of each individual piece of content. The thing hasn't blown up yet — but it feels like it's about to."

That stopped me for a second, and then I kind of got what he meant.

What he's saying is, you don't do GEO right now because it's going to drive sales tomorrow.

You do GEO right now to claim real estate.

Every answer the AI gives pulls from only three to five sources. Once a brand or a model gets "remembered" by the AI, the cost for a latecomer to muscle in goes up exponentially.

It's a bit like Google rankings in 2005. The ones who poured money into SEO back then have been cashing that dividend for two decades. The ones who sit it out now will probably have to spend ten times the effort to get the AI to mention them by 2030.

So Franz says, don't count ROI right now. What you count is not falling behind.

The most surreal part

Okay, the last piece of this story is the one that hits me the hardest.

Guess who is writing all that AI-friendly content the brands are producing — the FAQs, the spec tables, the comparison checklists?

The AI.

Yes.

Early on, everyone assumed "the AI won't train on things written by AI," so they bent over backwards to keep their content from being touched by AI. And the result? The Brandlight executive just shrugged:

"Turns out, that's not how it works. The AI happily eats AI-written content."

So the closed loop now looks like this: AI writes the content → AI trains on that content → AI reads that content out loud in its answers.

Full circle. The AI has stitched brands, content, and traffic together — the whole lot.

In closing

I've been turning this over in my head for a few days.

And I think I've figured out one thing.

In the SEO era, you were courting an algorithm.

In the GEO era, the algorithm you're courting can talk back.

The difference sounds small, but the moves are completely different.

You used to write keywords, stack backlinks, and pad the word count so a machine could parse you.

Now you write specific Q&As, real data, and the exact details a user would actually ask out loud — so that a machine that talks will choose to read your answer back to a human.

Strip it all the way down, and the logic of GEO comes down to a single sentence:

Can you, in one passage, directly answer a specific person's specific question?

If yes, you win.

If no, then as far as the AI's mouth is concerned, your brand doesn't exist.

I don't know how this war is going to play out. But there is one thing I'm pretty sure of.

The brands that are still writing "we are a company dedicated to..." website copy are, in all likelihood, going to be casualties of this war.

Here's hoping your brand isn't one of them.