If Your Brand Doesn't Exist in AI's Eyes, That $80 Billion of SEO Was for Nothing

A few days ago, a friend told me about buying a birthday present for his mom.

She loves chocolate. In the old days, he'd either trudge through several shops or page through search results, opening a dozen browser tabs, comparing prices until his eyes blurred.

This time he took a shortcut. He dumped the list of chocolates he'd given before into an AI assistant, and in seconds it came back with a recommendation sheet: no repeats, in stock, with prices — plus one line: "Rare white cacao with a raspberry-lime crisp — she hasn't tried the Peruvian one yet."

When he told me this, something clenched in my chest.

It wasn't envy at his AI fluency. It was suddenly realizing one thing: from discovery to comparison to recommendation, not a single brand website was ever opened.

Whether your chocolate is good no longer gets decided by your landing page. It gets decided by AI.

This is not a little gift-shopping anecdote. This is a seismic shift in an $80-billion piece of real estate.

Numbers Are More Honest Than People

First, a set of numbers.

From 2024 to now, website traffic referred by AI services has grown sevenfold. You might say: small base, right? Yes, very small. AI platforms currently contribute just 0.15% of global traffic, while traditional search still holds roughly half.

But the speed is terrifying. Adobe's data shows that between July 2024 and February 2025 alone, AI referrals to US retail sites grew 12-fold. By July 2025, the year-over-year growth hit 4,700%.

Four thousand seven hundred percent. My first reaction was that someone had typed an extra zero.

Worse still, among heavy users: more than a third of people have already made a habit of finding things with AI, leaving traditional search on the bench.

For twenty years, SEO plus paid ranking was the default customer-acquisition posture for enterprises. That business is worth $80 billion. Now the ground is shaking.

Some call this "the second act of search." Act one was about rankings. Act two is about — whether AI is willing to cite you, recommend you, and speak for you.

That's GEO: Generative Engine Optimization.

What Is GEO?

Don't be intimidated by the name. Take it apart and only one word changes: the search engine becomes a generative engine.

The old Page Rank was built on links. Whoever gets linked to most is the authority — all "voting" between sites.

GEO is built on language.

The user has changed too. You used to search "single-origin chocolate award-winning brands" and the machine dumped a row of links. Now the user just says: "I want to buy my mom three single-origin chocolate bars. She's open to new things, budget is this much — what do you suggest?"

Question, answer — like chatting with a shop clerk.

For marketers, the difference lands in four places.

The goal changed. In the SEO era you fought for rankings; in the GEO era you fight to be cited in AI answers.

The metrics changed. You used to watch click-through rates, rankings, organic traffic; now you watch citation frequency, mention rates, and your brand's share of voice inside AI answers.

The path to authority changed too. You used to buy backlinks from high-authority domains; now you compete on E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness. The framework first appeared in Google's Search Quality Rater Guidelines, and Google itself says "Trust" is the ballast among the four letters — and in the GEO era, it weighs even more.

And the user journey changed. In the SEO era: click a link, browse the site, buy — all three steps on your turf. In the GEO era, the answer is fully delivered inside the AI; the user may never come to your site at all, and the purchase itself may even happen one layer removed.

Zero-click. Your website goes from "destination" to "raw material."

Whom Does AI Trust?

So how does AI actually pick? Before recommending a product, it's really working through a few true/false questions: does it know you (do you appear in training data and web sources); can it find you (is your data structure clear); does it trust you (do credible third parties vouch for you); do you deserve this question (does your content match user intent); and can the journey continue (did you give a clear next step).

Note the third one. Researchers tested it with the GEO-BENCH benchmark: content carrying citations, sourced quotations, and statistics can gain more than forty percent in AI citation visibility.

Now look at a telling list: the domains ChatGPT cites most in the US are Reddit, Wikipedia, Amazon, Forbes, and Business Insider.

Notice anything? More than half of these aren't places where brands speak for themselves — they're places where others talk about you.

In the world of GEO, what others say about you is worth far more than what you say about yourself.

Researchers call this new authority "semantic authority": it's not about how many backlinks you have, but whether your content has depth, factual accuracy, and genuine experience behind it. Frankly, this is good news for smaller players who genuinely have the goods — on resources and backlinks you can't outspend the giants, but on expertise and sincerity, you can compete.

Every AI Has Its Own Temper

There's another easily missed detail: AI platforms are not a monolith.

Some models feed only on training data with a knowledge cutoff — to be known by them, you must exist in high-quality sources before their training window. Others are wired into live search — Perplexity is search-first by birth and constantly goes online; ChatGPT and Claude require either a fresh-enough topic or manually enabling web access; Gemini's AI Mode rides on top of Google Search and also ventures out often.

This shapes strategy massively. With pure training-data models, you compete on "historical presence"; with live-search models, you compete on "current freshness" — content must be structured, extractable, shaped like a direct answer to a question, and your product data must feed a real-time stream.

Consumers Aren't Waiting for You to Get It

Don't assume this is still a lab phenomenon.

July 2025 data: 38% of US consumers had already shopped with AI, and were projected to pass half by year's end.

Retailers are moving even faster. Amazon has Rufus, Walmart has Sparky, Taobao has Wenwen, and Carrefour launched Hopla, a ChatGPT-based plugin. At Amazon's Prime Day 2025, Rufus answered natural-language questions directly, and AI could even generate shopping guides to run side-by-side comparisons.

The transaction layer is moving too. In April 2025, ChatGPT launched a shoppable-product carousel and opened a merchant data program letting brands pipe first-party information straight into the platform; five months later, in-app instant purchasing went live. Microsoft Copilot's merchant program and Perplexity's Buy with Pro are all doing the same thing.

The loop is closing: from asking a question to completing the purchase, not a single step needs to pass through you.

So — What Can You Do Right Now?

Frankly, the playbook is still in its wild-growth phase, and nobody dares claim a standard answer. But four things can be done today.

First, measure how you look in AI's eyes. Traditional search engines introduce themselves in your analytics; AI referrals don't — you have to identify ChatGPT, Perplexity, Gemini, and Claude as sources inside your analytics tools. Even the crudest method works: take typical questions about your product, ask each AI, and see whether you appear in the answers. GEO monitoring tools like Goodie, Profound, and Daydream exist for exactly this. The real significance of the exercise is forcing the organization to treat AI as a legitimate marketing channel, not some employee's side experiment.

Second, systematically accumulate E-E-A-T. Case studies, customer stories, white papers, research with numbers — AI loves to lift facts and statistics it can quote verbatim. Add structured markup to your authors so machines can read them; display ratings and testimonials so AI knows not just "what you are" but "why you're worth it."

Third, shift effort from self-praise to being cited. Get PR and marketing working together: publish in industry outlets, go on podcasts, participate in Wikipedia and industry communities — get your name repeatedly appearing in authoritative venues. Only when AI sees you mentioned in multiple places will it dare write you into an answer.

Fourth, test, measure, adjust. The mechanics by which AI selects products keep changing, just like SEO algorithms always did — no fix-it-once solution. Benchmark against competitors regularly, check whether the "you" in AI's eyes matches the persona you're trying to build, and iterate against the behavioral data of your AI-referred traffic.

Jason Forbes, who works on AI brand optimization, has five goals worth watching — I think it's a clean summary: being found (visibility), how you're described (narrative control), where you stand versus competitors, how clear the user's next step is, and — what you got back for the money you put in.

Finally

From the Yellow Pages to Google, from desktop to phone, the ways customers find us have changed several times over. Looking back at history, the winner usually isn't the one with the best product — it's the one that understood the shift of the wind earliest, then pushed its resources onto it.

This time, the changeover is more thorough than the birth of the search engine itself.

The good news: genuine substance is worth more in the GEO era — authenticity, expertise, and trustworthiness are things AI can sniff out. The bad news: the time left for "let's wait and see" is running short.

Think back to that Peruvian white cacao from the opening. AI selects for you, speaks for you, closes the deal for you — and if your brand isn't in its answers, then to the next generation of customers, you effectively never existed.

Explaining to your shareholders "why can't anyone find us" at that point — that will be too late.