How businesses get recommended by AI assistants

Ask ChatGPT for “a good plumber in Leeds” and it names a handful of businesses. How does it choose them? Not at random, and not by who paid the most.

AI assistants recommend businesses they can find, verify and quote. They pull from the open web, business directories and review sites, then favour names that show up consistently with clear, structured, corroborated information. You don’t bribe your way in. You earn it by being easy to find, easy to read and backed up by other sources.

The mechanics differ from old-school Google ranking, but the goal overlaps: be the most trustworthy, clearest answer to a question someone’s actually asking. Here’s how that plays out, and what you can actually do about it.

Where AI assistants get their answers

AI assistants build recommendations from a blend of sources: their training data, live web search, and licensed feeds from directories and review platforms. ChatGPT and Perplexity often search the web in real time, then summarise and cite what they find. Gemini leans on Google’s index. The common thread is that they all need to find you first.

That means the businesses getting named are usually the ones with a real website, an up-to-date Google Business Profile, reviews on the platforms people trust, and consistent details everywhere. If your name, address and phone number differ across listings, an AI has no clean fact to repeat.

Why some businesses get named and others don’t

Three things separate the businesses AI mentions from the ones it skips: presence, consistency and clarity. A business with no website rarely gets recommended, because there’s nothing for the assistant to point to. A business whose details conflict across listings gets passed over, because the AI can’t tell which version is true.

And a business that describes itself in vague marketing fluff loses to one that states plainly what it does, where, and for whom. “We deliver bespoke solutions” tells an AI nothing. “Emergency boiler repair in Bristol, available 24/7” gives it an exact match for a real query.

This is the heart of GEO, generative engine optimisation. What GEO is and why it matters covers the concept in full.

A worked example: two electricians, one town

Picture two electricians in the same city, both decent at the job. Ask an AI assistant to “recommend an electrician near me” and only one gets named. Why?

The first has a one-page website that loads fast, a heading that reads “Emergency and domestic electrician in Sheffield, NICEIC registered, same-day callouts,” 47 Google reviews averaging 4.8, and identical contact details on Google, Facebook and three trade directories. The second has a Facebook page, no website, eight reviews, and a phone number that’s written two different ways across listings.

The assistant has a clean, corroborated story to tell about the first electrician. It can say what they do, where, and that real customers rate them. About the second, it has almost nothing it can verify, so it stays quiet. Neither is a better sparky. One is just far easier to recommend. That gap, not skill, is what GEO closes.

What you can actually do about it

You can’t control an AI’s training data. You can control what it finds when it searches the web today. Front-load your most important facts: ~44% of AI citations come from the first 30% of a page, so answer the obvious question in your first paragraph.

The mistakes that keep businesses invisible to AI

Most businesses that never get named by an assistant are making the same handful of avoidable mistakes. None are about being a bad business. They’re about being a hard-to-read one.

• No real website. A Facebook page or an Instagram bio isn’t enough. Assistants want a stable page they can point to, and social profiles are awkward for them to cite.
• Conflicting details. Your phone number, business name or opening hours differ between Google, your site and directories. The AI can’t tell which is true, so it trusts none of them.
• Marketing fluff instead of facts. “Bespoke, holistic solutions for the modern home” tells an assistant nothing. “Bathroom fitting and tiling in Cardiff” gives it an exact match.
• Dead reviews. A handful of old reviews reads as a business that’s gone quiet. Recency matters as much as the star count.
• Hidden basics. Burying what you do, where, and how to reach you below the fold or behind a slideshow. If a person has to hunt for it, an AI will give up faster.

Fix those five and you’ve done more for your AI visibility than any clever trick. The pattern’s clear: assistants reward the businesses that make themselves easy to understand.

Be honest about the timeline

Nobody can guarantee an AI will recommend you, and you should be wary of anyone who promises it. AI models update on their own schedules, and live web results shift constantly. What you can do is stack the odds: be findable, consistent and quotable, and give the tools every reason to pick you.

Like SEO, this compounds. New content and listings typically take 3-6 months to gain real traction. There’s no overnight switch.

Where we fit in

Keeping all this current, the website, the listings, the reviews, the structured content, is ongoing work, not a one-off. That’s the part most owners can’t keep up with alongside the day job.

Our team handles the whole stack: website, Google Business Profile, reviews system, and content written to be quotable by both search engines and AI. You can see what’s covered on the pricing page, and how to get found on Google and ChatGPT goes deeper on the practical side.

The gut check: AI recommends businesses that are genuinely easy to verify and clear about what they do. Sort that out, and you’re already most of the way there.