The short version: ChatGPT does not rank businesses the way Google ranks websites. It pattern-matches your question against whatever it was trained on plus whatever it can pull from a live search, then it repeats whichever names show up most often, most consistently, and most recently across that mix. If your business barely exists in that mix, it will not get mentioned, no matter how good you are at what you do.
What’s happening when you ask ChatGPT for a recommendation
When someone types “what’s the best accountant for freelancers in Manchester” into ChatGPT, one of two things happens behind the scenes. If it’s answering purely from its trained knowledge, it’s drawing on patterns baked in during training, which for most people right now means data that’s a year or two old, frozen in time, full of whatever was written about loudly and often. If it’s using ChatGPT search or browsing (which it does by default for anything time-sensitive or local), it’s running a live web search, pulling back a handful of pages, and summarising what it finds.
This second mode works a lot like the way other AI tools decide what to cite, and I’ve written before about how AI engines decide which website to cite first, because the mechanics overlap heavily with how ChatGPT picks businesses to name. It’s grabbing pages that are clear, well-structured, and repeated across multiple sources, not necessarily the pages that are most accurate.
The sources it’s leaning on
From testing this with client accounts and my own, the same handful of source types come up again and again:
- Roundup and “best of” articles from third-party publishers (think NerdWallet, TechRadar, PCMag style listicles for the relevant industry)
- Review aggregators like G2, Capterra, Trustpilot, and Google Business Profile reviews
- Reddit threads, because Reddit gets weighted heavily in a lot of AI search results right now
- Wikipedia and other reference-style pages for anything with a long history
- The business’s own website, but only if the content answers the question directly rather than just describing the brand
Notice what’s missing from that list: your opinion of your own business, your awards page, and your “about us” story. None of that carries much weight unless it’s echoed somewhere else the model can find.
A real test I ran, with the numbers
Back in October I ran a small experiment with a client who runs a boutique PR agency in the Midlands. We wrote out 20 realistic prompts a potential customer might type, things like “best PR agency for a tech startup in the UK” and “small PR agency alternative to a big London firm.” We ran each one through ChatGPT with search on.
Her agency, which by every measure I could see does better client work than most of its named competitors, showed up in 3 of the 20 answers. A competitor with a weaker portfolio but a Trustpilot page with 140 reviews, three separate mentions in industry roundup articles, and an active Reddit presence from happy clients showed up in 14 of the 20. Same quality of work, wildly different visibility. That gap wasn’t about who’s better. It was about who left more digital fingerprints in the places ChatGPT goes looking.
That’s the uncomfortable part nobody likes saying out loud: ChatGPT isn’t judging your business, it’s counting mentions. A mediocre business with 200 scattered reviews and five press mentions will out-recommend an excellent business with none, every single time, because the model has nothing else to go on. It has no way to taste your coffee or sit in your accountant’s office. It only has text.
What you can control
You cannot buy your way into a ChatGPT answer the way you could once buy backlinks in 2011. There’s no ad slot, no paid placement, no account rep to call. What you can do is make yourself easier for the model to find and repeat:
- Get reviewed, and often, on the two or three platforms most relevant to your industry, not scattered across twelve
- Get named in third-party roundup content, which usually means pitching journalists and niche bloggers directly, the same relationship-building that used to get you a backlink now gets you an AI mention
- Write pages on your own site that answer the exact questions your buyers ask, in plain factual language, not brand-voice marketing copy dressed up as an answer
- Keep your business name, location, and category consistent everywhere, because inconsistent details make a model less confident about grouping mentions of you together
- Add clear structured information to your site, and if you’re technical enough or have someone who is, look at whether an llms.txt file is worth setting up for your site so AI crawlers can find your key pages faster
None of this is a trick. It’s the same groundwork that’s always mattered for word of mouth, just now the loudest word of mouth on the internet is a language model reading everyone else’s word of mouth back to a user in real time.
Where copywriting and clarity move the needle
One thing I keep seeing on client sites is content written to impress a human reader with tone, rather than answer a question a machine can lift and repeat. There’s a real difference between the two, and it’s worth understanding what copywriting means within marketing before you rewrite a page purely to please an algorithm. A page that says “we’re passionate about delivering excellence” gives ChatGPT nothing to quote. A page that says “we handle payroll for businesses with 5 to 50 staff, starting at 150 pounds a month, based in Leeds” gives it a fact it can reuse.
The bit most people get wrong: recency and volume beat prestige
I’ve had clients assume that a Forbes mention from three years ago will carry weight forever. It won’t, or at least not as much as they hope. Models weight recent, repeated mentions higher than a single big one from years back, especially once ChatGPT search is pulling live results rather than relying on frozen training data. A local plumber with 40 fresh Google reviews from this year will often out-mention a national brand whose last big press wave was in 2019. Freshness and repetition matter more than the size of the logo that once wrote about you.
This is also why automating your review requests, your press outreach, and your content publishing matters more now than it did five years ago. If you want a fuller picture of how that kind of automation works day to day, I’ve broken down what AI marketing automation is and how it works in more detail, and it’s worth reading before you try to build this by hand across a dozen platforms.
Testing your own visibility, step by step
You don’t need special tools to check where you stand. Here’s exactly what I do with clients:
- Write out 15 to 20 realistic prompts a customer might type, using the language a customer would use, not your own industry jargon
- Run each one through ChatGPT with search enabled, and note whether you appear, in what position, and alongside which competitors
- Repeat the same test in Perplexity and Google’s AI Overviews, because the source patterns overlap but aren’t identical
- Log the competitors who keep showing up and go look at what they have that you don’t, usually reviews, press mentions, or Reddit presence
- Fix the biggest gap first, then rerun the test in 60 days, not 60 minutes, because these results shift slowly
If you want a head start on crafting the actual test prompts, I’ve put together a guide to writing sharper ChatGPT prompts for business tasks that’s useful for this exact exercise, not just for content generation.
Where a good AI marketing tool helps
Once you know where the gaps are, some of the fixing can be sped up with the right software, whether that’s review-request automation, listing management, or content tools that help you publish consistently instead of in bursts. I’ve reviewed a shortlist of AI powered marketing tools that are worth using for a business, and most of them earn their place by doing the boring repetitive part of this, the reviews, the listings, the consistency, rather than promising to game the model directly.
If this all feels like more than you have time to run yourself, this is exactly the kind of gap analysis I do hands-on with clients through my AI consultant work for small businesses, usually starting with the same prompt-test I described above so we know precisely where the gaps are before we touch a single page.
More on ChatGPT prompts: the complete ChatGPT prompts guide groups all of these by problem.
Frequently asked questions
Can I pay ChatGPT to recommend my business?
No, there’s no paid placement or ad slot inside a normal ChatGPT answer. The only way to influence it is indirectly, by increasing genuine mentions of your business across reviews, press, and your own website content until the model has enough consistent signal to repeat your name.
Does ChatGPT recommend businesses based on quality?
Not directly. It has no way to test your product or sit through your service, so it relies on how often and how recently your business is mentioned across the web. A business with more visible mentions will often be named ahead of a better business with fewer of them.
Why does a competitor with worse reviews still get recommended over me?
Usually it’s volume and recency rather than quality. A competitor with 100 recent Trustpilot reviews and a mention in a roundup article will out-rank a business with five glowing reviews from three years ago, because the model is counting fresh, repeated signal, not scoring satisfaction.
How often does ChatGPT update who it recommends?
When it’s using live search, the answer can change from one week to the next depending on new reviews, articles, or forum posts. When it’s answering from trained knowledge alone, the picture can stay frozen for months at a time, which is why testing your own visibility every 60 days or so gives a more honest read than checking daily.
