AEO: How to Get AI Assistants to Recommend Your Business

You know SEO: you build your site so Google ranks it high on the results page. AEO — Answer Engine Optimization, sometimes called GEO — is the same idea moved one step further. The goal is for your business to appear inside the finished answer a customer gets from ChatGPT, Perplexity, or from Google above the list of links.

The goal hasn't changed: new customers. What changed is where the customer makes the decision.

How Is AEO Different From SEO?

Put simply: SEO competes for a spot on a list. AEO competes to be the substance of the answer. In SEO, the click is what counts. In AEO, what counts is whether your name gets said at all.

• SEO: the customer sees 10 links and chooses. AEO: the customer sees 3 names, or one.
• SEO: position matters. AEO: presence matters — there is no page two or three outside the answer.
• SEO: you optimise for keywords. AEO: you optimise for questions asked in full sentences.
• SEO: you check results in a ranking. AEO: you check results by asking the question and seeing who shows up.

One important thing: AEO doesn't replace SEO. It sits on top of it. A site Google can't read is equally invisible to AI assistants — the same crawlers, the same plumbing, the same foundation.

How Do AI Assistants Pick the Businesses They Name?

Nobody publishes the exact algorithm, and anyone claiming otherwise is selling you a story. The mechanism itself is no secret though: a model answers from what it has read — the material it was trained on, plus the web it searches at the moment the question is asked. Six things follow from that, all of which you can actually influence.

1. A Site That Can Actually Be Read

AI crawlers (OpenAI's GPTBot, PerplexityBot, Google-Extended) have to be able to reach your site and pull text from it. If robots.txt blocks them, if content only loads after a click, or if your key facts are baked into images — your site doesn't exist for them. Speed matters too: a page that takes 8 seconds to load often gets abandoned halfway.

2. Clear Facts About the Business

A model has to be able to say in one sentence who you are: what you do, for whom, where, since when, during what hours, on what number. Those facts need to be written out plainly, in the page's actual text — not in a downloadable PDF, not inside a photo of a business card, not only in six-point footer type.

3. Structured Data (schema.org)

The same facts, additionally written into your page's code in a format built for machines. LocalBusiness describes the company, Service describes each individual service, FAQPage covers questions and answers. Your customer never sees it, but Google and language models read it as settled fact rather than something to be inferred from marketing prose.

4. Consistent Details Across the Web

Your name, address and phone number should be identical everywhere: on your site, in your Google profile, in trade directories, on Facebook, on any invoices available online. Every discrepancy makes a model unsure it's dealing with the same business. And when it's unsure, it usually names somebody else.

5. Content That Can Be Quoted

Customers ask in full sentences: "how much does a timing belt change cost", "how long does resin flooring take to cure", "do you build custom furniture for small flats". A page that uses those questions as headings, with a short, specific answer under each, hands a model a ready-made passage to quote. A page with a single "about us" paragraph hands it nothing.

6. Reviews and Mentions Beyond Your Own Site

A model is far more willing to name a business somebody else has said something good about. Google reviews, entries in local directories, mentions in trade publications, projects written up by partners — that's the raw material behind the phrase "recommended because…". Without it, all you have are your own claims.

What Nobody Can Guarantee You

You cannot buy a place in a ChatGPT answer. There are no ads there, no subscription for the top slot, and nobody to call about it. Answers differ between users, between days and between model versions — the same prompt asked twice can produce two different lists of businesses.

If someone promises you "first place in ChatGPT", they are selling something they don't control. What you do control is a different question: whether your business is even a candidate to be named.

Checklist: What You Can Do Yourself

You can tick off a good chunk of this list in one evening, without a developer and without spending anything:

• Fill your Google Business Profile in completely and add the full list of services.
• Check your name, address and phone number everywhere your business is listed, and make them identical.
• State plainly on your site: what you do, for whom, in which city and area, with what lead time, in what price range.
• Replace generic headings with the questions customers actually ask, and add a short, specific answer under each.
• Ask customers from the last three months for a Google review.
• Ask ChatGPT and Perplexity the questions a customer would ask, and save the answers as your baseline.

Where You'll Usually Need Help

• Structured data (schema.org) for the business, the services and the FAQ sections — implemented correctly and validated.
• The technical access layer: robots.txt, sitemap, an llms.txt file, load time, content that renders without JavaScript.
• Content architecture: service pages and local pages built around real questions rather than slogans.
• Regular measurement — the same questions every month, so you can see whether anything moved.

One more thing, about timing: this doesn't work in a week. Changes to your site have to be crawled, and models refresh what they know at their own pace. Realistically we're talking weeks, not days — which is exactly why a written-down starting point is worth having.