GEO in 2027: how AI assistants decide which brands to name

For a while the advice was easy: rank in the top ten and the AI answer will probably cite you. Ahrefs measured exactly that, twice, and the answer changed within months.

From three quarters to just over a third. Most pages cited in an AI Overview today are not on page one for that search.

This cuts both ways. A smaller site with a genuinely useful, specific page can be cited without outranking the giants. And a page-one ranking no longer protects you from being skipped.

Peec AI analysed 30 million sources cited across ChatGPT, Google AI Mode, Gemini, Perplexity and AI Overviews. The most-cited domains, in order, were Reddit, YouTube, LinkedIn, Wikipedia and Forbes.

Look at that list again. Four of the five are places where people, not brands, do the talking.

It means your own website is only part of your generative footprint. What customers say in forums, what your team publishes on LinkedIn, whether there is a clear video explaining what you do, and whether independent publications mention you all feed the answer.

A note of caution: these rankings move. A Semrush study in mid-2025 had Quora first for AI Overviews. Do not chase one platform. Aim to be described accurately in several.

If you depend on local customers, this is the most important chart in the article. SOCi checked about 350,000 locations of multi-location brands in the US and asked a simple question: how often does each system recommend them?

More than a third appear in Google's map pack. About one in a hundred are recommended by ChatGPT.

Assistants are far pickier than the map. The same study found business details were only about 68% accurate on ChatGPT and Perplexity, which hints at why: when the information about a business is thin or inconsistent, the assistant plays safe and names someone else.

It would be easy to read all this and conclude that the assistant's word is final. It is not, at least not yet.

BrightLocal's 2026 consumer research found 45% of consumers use ChatGPT or similar tools for local business recommendations. But only 18% of US consumers felt ready to contact a business an AI tool recommended without checking further, and about 43% of those who started with AI went to Google afterwards.

So the journey looks like this: the assistant builds the shortlist, then the person verifies it. Reviews, your Google Business Profile and your website are the verification step. GEO gets you onto the list. Classic local SEO and a credible site keep you on it.

Here is what we would put on a one-page plan.

Decide the 20 to 40 questions you want to be the answer to. Real buyer questions, in their words. Run them through the main assistants and write down who gets named today. That is your baseline.

Make your facts identical everywhere. Business name, address, phone, categories, services, prices and hours on your site, your Google Business Profile and every directory that matters in your sector.

Publish things only you can publish. Your own data, your process, your pricing logic, worked examples. Assistants have a thousand generic explainers to choose from and very few first-hand ones.

Earn mentions where assistants read. Answer questions in the communities your buyers use, get reviewed, be quoted by trade publications. This is slower than publishing a blog post and worth more.

Let the crawlers in. Check that your robots.txt is not blocking the AI crawlers you want to be read by, and that key content is not hidden behind scripts.

Re-run the same question list each month and count how often you are named, and for which questions. That citation rate is your GEO scoreboard.

Alongside it, watch branded searches and the "how did you hear about us" field on your enquiry form. In our view, those two will show the effect of AI visibility before referral traffic does, because most people who see your name in an answer do not click through at that moment.

Review quarterly. Answers shift week to week, and a weekly read mostly measures noise.