How AI Decides Which Businesses to Recommend
Ask ChatGPT for the best employment solicitor in Manchester and you won’t get ten blue links. You’ll get three or four names, a line on each, and maybe a suggestion of which one to call first.
That’s a shortlist. And it’s been written before the person asking has visited a single website.
If your business isn’t on it, you’re not losing a ranking battle. You’re just not in the conversation. So it’s worth understanding how AI tools actually decide who makes the cut, because it isn’t quite how Google rankings work.
Step one: AI works out what’s really being asked
Before it recommends anyone, an AI tool has to understand the question. And most questions carry more than they say.
“Best employment solicitor in Manchester” isn’t really one question. It’s several. Who handles employment law in Manchester? Do they act for employees or employers? Are they any good? What do clients say about them? Are they affordable?
Google calls this query fan-out. Its own guidance explains that AI Overviews and AI Mode can split a question into subtopics and run a series of related searches across different sources before writing the answer. Other AI tools work in a similar way when they search the web.
That matters for you, because it means AI isn’t only looking for pages that match the exact words someone typed. It’s looking for passages that answer each of those smaller questions. A business that clearly answers “who do you help, where, and with what?” has far more chances to be picked up than one that just says “expert legal advice”.
Where AI looks for its answer
AI tools draw on two kinds of information, and both shape who gets recommended.
What the model already knows. Every AI model is trained on a huge snapshot of the web. If your business was mentioned often and consistently in that snapshot, the model has a sense of who you are before it searches for anything. If it wasn’t, you’re starting from nothing.
What it finds when it searches. Tools like Perplexity, ChatGPT search and Google’s AI Overviews also look things up in real time. They run those smaller searches, pull back a set of pages, and pick out the passages that best answer each part of the question.
The pages they pull back often aren’t business websites at all. When Ahrefs analysed 750 recommendation-style prompts in ChatGPT, including agency searches like “best web design agencies in London”, it found that “best X” list articles made up 43.8% of the pages cited. Directories, review sites and news coverage turn up regularly too.
In other words, when someone asks AI who to hire, the answer is mostly built from what other people have written about the options.
The signals that decide who makes the shortlist
No AI company publishes a recipe for this. But across the research so far, and what we see when we test AI answers for clients, the same signals keep coming up.
1. A clear specialism
AI has to match your business to the question. “Family law solicitors in Liverpool” is easy to match. “A full-service firm offering a wide range of legal solutions” isn’t. The more specific your language about what you do, who for and where, the more questions you can be matched to. Broad positioning feels safer, but to AI it reads as vague.
2. The same story everywhere
If your website says you’re an implant specialist, your Google Business Profile says general dentistry and a directory lists an old address, AI has to decide which version to believe. Sometimes it picks the wrong one. Sometimes it plays safe and leaves you out. Consistent descriptions across every profile make you a much easier recommendation. We covered this in more detail in AI is rewriting your About page for you.
3. Other people talking about you
AI search leans heavily on independent sources. AI tools show a strong bias towards third-party content, like reviews and independent publications, over what brands say about themselves. Press coverage, client case studies, awards, podcast appearances and trade association listings all count as someone else vouching for you.
4. A place on the lists
Because “best X” articles are such a big part of what AI cites for recommendation questions, being on them matters. The Ahrefs research also suggested that appearing higher on several comparison lists makes a recommendation more likely. For most service businesses, that means trade directories, local “top firms” round-ups and platforms like Clutch, Trustpilot or Google reviews, depending on your sector.
5. Reviews that say something specific
A review that says “great service” tells AI very little. One that says “they handled our commercial lease renewal quickly and kept us updated throughout” tells it what you do and how well you do it. Encouraging clients to mention the specific service they used gives AI real material to work with.
6. Pages that answer specific questions
Your own website still counts, especially the parts AI can lift out cleanly. Service pages that state plainly what you offer, who it’s for, what it costs or how long it takes, and FAQ sections with direct answers, are far easier to cite than long pages that circle the point. AI retrieves passages, not whole pages, so each passage needs to make sense on its own.
Why bigger names often win, and how smaller businesses compete
If you’ve tested AI answers in your sector, you’ve probably noticed the same big names turning up again and again. That’s not a coincidence. The University of Toronto researchers flagged a “big brand bias” in AI search, and it makes sense given everything above. Larger firms get written about more, appear on more lists and have more reviews. They simply give AI more to go on.
That doesn’t mean smaller businesses can’t compete. It means they compete differently.
The big names win the broad questions. “Best accountants in the UK” is a hard fight for a ten-person practice. But “accountants for dental practices in the North West” or “Xero specialists for construction firms” is a different story. The more specific the question, the fewer businesses clearly answer it, and the more a well-defined specialist stands out.
The same goes for lists. You don’t need to appear in a national round-up. Local guides, sector directories, trade association listings and niche comparison articles are a more realistic target, and they’re often closer to the questions your clients actually ask.
Getting on the shortlist
AI doesn’t pick businesses at random, and it doesn’t just copy Google’s rankings. It breaks the question down, looks at what’s been written about the options, and recommends the businesses it can describe with the most confidence.
That’s good news if you’re willing to put the work in. A clear specialism, a consistent story, a presence on the right lists and a few people vouching for you will go a long way.
Wondering whether AI recommends you or your competitors? Get in touch for your free AI search gap analysis.