ChatGPT recommends just 1.2% of local business locations, so most firms never surface in its answers at all (SOCi Local Visibility Index, 2026). It picks the names it does using two layers: what its training data already believes about your category, and what live search retrieval, built on a Bing-connected index, can corroborate through reviews, directories and your own pages. UK businesses that fix crawler access, tidy their entity data and earn reviews on the platforms AI actually reads tend to see the answers move within two to three months, though nobody can guarantee the outcome.
You typed the question your customers type. Maybe "best boiler engineer in Bristol", or "which accountant should a contractor in Leeds use". ChatGPT answered without hesitation, named three or four firms, and yours was nowhere. That absence is not random and it is not permanent. This guide walks through how ChatGPT actually arrives at those names, why the list is so short, and the specific, checkable signals that put a UK business on it. Every claim here is something you can verify or test yourself, and none of it means rebuilding your website.
How does ChatGPT choose which businesses to recommend?
Two layers decide what ChatGPT recommends. The first is training data: a snapshot of the public web that tells the model which names even exist in your category. The second is live search retrieval, which pulls today's pages through an index connected to Microsoft Bing. The businesses that get named are the ones that show up consistently in both.
What the model already believes
ChatGPT's underlying model learned from an enormous snapshot of the public web, and that snapshot hardens into what the model "knows" about your trade: which firms exist, what people say about them, how often anyone mentions them at all. A business with years of coverage, reviews and directory listings is baked into that picture. A business with a thin, quiet web presence barely registers as existing. You cannot reach into the training data and edit it, but you decide what the next snapshot finds, and the way to influence it is to start laying down consistent public evidence now, before that snapshot is taken.
What it looks up live
Ask about a specific business and ChatGPT rarely trusts memory alone; it runs a live web search. That retrieval leans on an index connected to Microsoft Bing, and the overlap is not a hunch, it is measured: 87% of ChatGPT search citations matched Bing's top organic results in one study of around 100 queries (Seer Interactive, 2025). The consequence is about as blunt as consequences get. If Bing has not indexed your website, ChatGPT's search layer can barely find you, and it does not care one bit how well you rank on Google.
The sources it cites
Open the citations under any ChatGPT answer about local businesses and the same handful of source types keep reappearing: review platforms, directories, Wikipedia, news sites and forums. Reddit and Wikipedia were the model's two most-cited domains across a three-month study of 100 million AI citations (Semrush, 2025). UK queries follow the same shape but with a local accent: Trustpilot, Google reviews, Yell, Tripadvisor for anything hospitality, trade platforms like Checkatrade, and the local press and community forums on top. Your own website is one voice in that chorus. It counts, but it is seldom enough by itself, because the model is looking for several independent sources telling the same story.
Why ChatGPT recommends so few businesses
ChatGPT is ruthless in a way a results page never is. Across more than 350,000 locations analysed, it recommended just 1.2% of local businesses, while the very same brands turned up in Google's local three-pack 35.9% of the time (SOCi Local Visibility Index, 2026). An answer names a handful of firms. A results page lists dozens. That is a thirty-fold difference in how many businesses get to be visible at all.
The gap comes down to the job each format is doing. A results page hands you a stack of links and leaves the choosing to you. An AI answer does the choosing for you, which is exactly why it commits to two to five names and quietly discards everyone else. Earning one of those slots means giving the model corroboration: several independent sources agreeing on who you are, what you do and whether customers actually rate you.
And the audience waiting on the far side of that filter is not niche. More than half of UK adults, 54%, now use AI tools such as ChatGPT (Ofcom Online Nation, 2025), and ChatGPT clocked 1.8 billion UK visits in the first eight months of 2025, up from 368 million over the same stretch in 2024 (Ofcom, 2025). Across the Atlantic, 45% of consumers now reach for AI to find local businesses, up from 6% a year earlier (BrightLocal Local Consumer Review Survey, 2026). That last figure is American, but UK habits have a long history of tracking the same curve a year or so behind. A channel this selective, expanding this fast, pays back the businesses that take it seriously.
The signals that get a UK business into the answer
Five signals decide whether a UK business turns up in ChatGPT's answers: consistent entity data wherever your business is mentioned, reviews on the platforms AI systems actually read, listings in UK directories such as Trustpilot and Yell, website pages built like direct answers, and a site OpenAI's crawlers can actually reach. Not one of them works in isolation. Stacked together, they build the corroboration the model is waiting for.
1. Consistent entity data
To an AI system your business is an entity: a name tied to an address, a phone number, a set of services and a body of reviews. Every mention of that entity across the web either firms up the model's confidence or chips away at it. So your trading name, address, phone number and service description need to match, exactly, across your website, Google Business Profile, Bing Places, Trustpilot, Yell and your social profiles. The killer is not the missing listing, it is the contradictory one: an old address or an abandoned profile that keeps feeding a second, wrong version of you into the record, so fix those rather than looking past them, or see our service for correcting wrong facts about your business online. LocalBusiness schema markup on your site then hands crawlers the facts outright, instead of leaving them to guess.
2. Reviews on the platforms AI actually reads
Reviews are the strongest third-party evidence the model can go and fetch, and here is the part most people miss: it reads the words, not just the stars. A detailed written review that names your services and your location hands a retrieval system something it can quote verbatim. Put your effort where AI answers already cite in your sector: Google reviews and Trustpilot for most businesses, Tripadvisor for hospitality, Checkatrade or TrustATrader for trades, Doctify for healthcare. Run a dental practice or clinic? Start with our guide for UK dentists. Ask every single customer, reply in public, and never, ever buy them.
3. UK directories and citations
Directories feel like a relic, and they are precisely the kind of structured, factual page a retrieval system reaches for when it wants corroboration. The UK shortlist is refreshingly short: Trustpilot, Yell, FreeIndex, Bing Places and whichever directory leads your trade or profession. Give Bing Places extra care, because ChatGPT's retrieval runs on a Bing-connected index and this is where that index looks first. Fifty listings is not the goal. A handful that are complete, accurate and consistent with one another is, and it is exactly the groundwork our citations and authority work covers.
4. Answer-shaped pages on your website
Retrieval systems lift passages, so the pages that get quoted are the ones that answer the question outright. Every core service wants a page that says what you do, where you do it, how pricing works and how fast you respond, and says it in the opening paragraphs, not buried three screens down. Picture a purely hypothetical emergency plumber in Sheffield whose page leads with response times, call-out fees and areas covered: ChatGPT has something clean to quote. A homepage that opens with "Welcome to our website" gives it precisely nothing to work with. Our full UK guide to AI search optimisation goes deep on page structure.
5. A site OpenAI's crawlers can reach
OpenAI runs two separate crawlers, and the distinction matters: GPTBot gathers content for model training, OAI-SearchBot indexes pages for ChatGPT search, and you can allow or block each one independently in your robots.txt file (OpenAI developer documentation, 2026). Plenty of businesses block these bots without ever meaning to, because some firewalls and CDN security settings shut out AI crawlers by default. Before you spend a penny on anything else, check whether your website is blocking AI crawlers. Block OAI-SearchBot and you vanish from ChatGPT's search citations completely, no matter how flawless everything else is.
A 10-minute test you can run yourself
You can take the measure of your own ChatGPT visibility in about ten minutes, and it costs nothing. Ask ChatGPT the questions your customers actually ask, in their own plain words, with web search switched on. Watch which businesses it names, then open the citations and note which websites it leaned on to get there. That list of sources is your working map of where visibility is really coming from.
- Write down five questions your customers ask, in their words. For a hypothetical Bristol letting agent that might be "best letting agent in Bristol for landlords" or "who manages student properties in Clifton".
- Ask each one in ChatGPT with web search enabled. Use a fresh chat for each question so earlier answers do not colour later ones.
- Record which businesses are named, in what order, and what is said about them.
- Open the citations under each answer and list the domains. This shows which review platforms, directories and pages ChatGPT trusts in your category.
- Ask ChatGPT about your own business by name. Note anything wrong: old addresses, dead phone numbers, services you no longer offer.
- Search Bing for site:yourdomain.co.uk. If few of your pages appear, the retrieval layer cannot cite you.
- Open yourdomain.co.uk/robots.txt and look for GPTBot or OAI-SearchBot next to a Disallow rule.
Run the same questions again next month, and the month after, and keep every result. Answers wander from one session to the next and shift as the indexes refresh, so any single test tells you far less than the trend does. The businesses named, set beside the sources that named them, are your true competitive map, and it very often looks nothing like the Google rankings you have been watching all this time.
What to expect month by month
Set your expectations to the real clock, not an overnight one. Technical fixes, unblocking a crawler or straightening out a listing, tend to be read within weeks. The answers themselves usually take two to three months to move, because the retrieval systems behind them refresh at their own uneven pace. Nobody can promise you a named spot, and the moment someone does, walk away.
Month 1. Fix access and facts. Unblock crawlers, correct every listing, claim Bing Places, submit your site through Bing Webmaster Tools and rewrite your core service pages so they answer questions directly. Crawlers typically re-read changed pages within days to weeks.
Months 2 to 3. The retrieval layer catches up. Your pages start appearing in Bing results, citations under relevant answers may begin to include your site or your refreshed listings, and answers to specific questions can change. This is when steady review growth on the right platforms starts to compound.
Month 4 onwards. Broader queries shift more slowly still, and the training-data layer only moves when a new model ships, which is nobody's to control. So keep testing monthly, keep the reviews coming, and treat every piece of local press or sector coverage as a long-term deposit that pays out later. Progress on this timescale is the norm, not a promise, and how fast it comes depends heavily on how crowded your category is.
Common mistakes that keep businesses invisible
The same handful of mistakes keep businesses invisible: treating ChatGPT visibility as a quick trick, blocking AI crawlers without realising it, pouring effort into one platform while ignoring the sources the model actually cites, and leaving old, contradictory details scattered across the web. Most are fixable in weeks. Every one of them is far cheaper to prevent than to unpick after the fact.
- Blocking crawlers by accident. Firewall and CDN defaults quietly remove some sites from AI answers. Check this before you spend money on anything else.
- Ignoring Bing. Many UK businesses tune everything for Google and never claim Bing Places or check Bing indexation, yet Bing's index feeds ChatGPT's search layer.
- Testing once and drawing conclusions. One question, one phrasing, one day tells you almost nothing. Answers vary, so trends are what count.
- Buying or faking reviews. Fake reviews are banned under UK consumer protection law, and they corrupt the one signal you most need to be genuine.
- Publishing claims without corroboration. Calling yourself the best plumber in Sheffield on your own site does not persuade a system built to cross-check independent sources.
- Deleting old listings instead of correcting them. A dead listing with your old address keeps feeding wrong facts into the record. Update it instead.
- Expecting instant change. The model is not watching your website in real time. Fixes are read in weeks and reflected in months.
Questions people ask
These are the questions UK business owners bring us most often about ChatGPT recommendations. The short of it: you cannot buy your way into the organic answer, the timescale is months not days, ordinary search work still earns its keep, and the review platforms that matter depend on your sector. The longer answers follow.
How long does it take to get recommended by ChatGPT?
Technical fixes such as unblocking crawlers and correcting listings are usually read within weeks, and the answers themselves typically shift over two to three months as the retrieval indexes refresh. Competitive categories take longer still. No agency can guarantee a named place in any AI answer, so treat a firm promise as a red flag rather than a selling point.
Can I pay to get my business on ChatGPT?
No. OpenAI started showing clearly labelled adverts to some free-tier users in 2026, but those sit apart from the answer itself. The businesses named inside a recommendation are chosen by the model from its training data and its live citations. That visibility is earned through evidence, not bought as a placement.
Do I still need normal SEO if I care about ChatGPT?
Yes. ChatGPT's search layer leans on a Bing-connected index, and 87% of its search citations matched Bing's top organic results in one study (Seer Interactive, 2025). Indexable pages, sound technical foundations and genuinely useful content are still the base layer. AI visibility work builds on that base, it does not replace it.
Which review platforms does ChatGPT read for UK businesses?
It depends on your sector, which is the whole reason the 10-minute test earns its place. Google reviews and Trustpilot surface across most categories, Tripadvisor rules hospitality, Checkatrade and TrustATrader carry weight for trades, and Doctify shows up for healthcare. Run the test, read the citations under the real answers, and back the platforms ChatGPT already trusts in your category rather than the ones you assume it should.
Sources
- SOCi, Local Visibility Index 2026
- Seer Interactive, 87% of SearchGPT Citations Match Bing's Top Results (2025)
- Semrush, The Most-Cited Domains in AI: A 3-Month Study (2025)
- Ofcom, Online Nation 2025
- BrightLocal, Local Consumer Review Survey 2026
- OpenAI, Overview of OpenAI Crawlers
Stagg Studios does this for you
Entity data, reviews, directories and answer-shaped pages, kept in order every month on the site you already have. One simple price: £100 a month or £1,000 a year, no lock-in.
Get startedSee pricing