How to Get Your Business Recommended by ChatGPT (2026 UK Guide)

People used to ask a mate, or Google, for a recommendation. A fast-growing share of them now ask ChatGPT instead. BrightLocal’s 2026 survey found that 45% of consumers now use AI tools to find local businesses, up from just 6% a year earlier, with ChatGPT the tool they reach for most often.

Here is the catch. When someone types “best [your service] in [your town]” into ChatGPT, it names a tiny handful of businesses and quietly ignores everyone else. SOCi’s 2026 Local Visibility Index, which analysed more than 350,000 locations, found that ChatGPT surfaces roughly 1.2% of them. The other 98.8% may as well not exist.

That scarcity is the whole point. A ChatGPT recommendation does not arrive as one blue link among ten, it arrives as an answer, already shortlisted and pre-qualified, from a source the person has decided to trust. Being one of the two or three names it returns is worth far more than being result number seven on a page nobody scrolls. Being none of them costs you the enquiry before you even knew it existed.

This guide is the plain-English version of how to be in that 1.2%. No jargon you have to go and look up, no guarantees we cannot keep, just the moves that actually get a business named. It sits inside our wider guide to answer engine optimisation, and we have tested every step of it, including on ourselves (more on that embarrassing result later).

How does ChatGPT actually decide what to recommend?

ChatGPT recommends what it can retrieve, verify and attribute. It does not rank businesses on a fixed list, and you cannot pay to jump the queue. It pulls from a few overlapping sources and names the businesses that show up cleanly across all of them.

There are three inputs worth understanding:

• Training data. The model learned from a huge slice of the web up to a cut-off date. This is where its general impression of your brand lives, but it is frozen and often out of date, so it is not something you can change this quarter.
• Live web retrieval. When a question needs current, local or specific information, ChatGPT Search fetches live pages, reads them, and cites the ones it used. This is the part you can influence quickly, and it is where most of your effort should go. OpenAI runs dedicated crawlers to feed it, which we come back to below.
• Memory and connected sources. ChatGPT increasingly remembers a user’s preferences and can pull from connected apps and accounts. You cannot game this, but it quietly rewards businesses that people already like, save and mention by name.

The thread running through all three is trust. ChatGPT will only put its name to a recommendation it can stand behind, so it leans on businesses whose details are consistent, corroborated by independent third parties, and easy for a machine to parse. That is a genuinely different job from ranking on Google, where the page itself is the prize. Here, your page is only useful if the model can lift a clean, quotable fact from it and back that fact up elsewhere. Our guide to GEO versus SEO versus AEO breaks down how those disciplines overlap, but the short version is this: Google ranks pages, ChatGPT assembles an answer and needs sources it can cite without getting burned.

Picture what happens behind a single query. Someone asks for a “reliable electrician in Loughborough”. ChatGPT runs a live search, pulls back a handful of pages, and reads them against each other. It finds a firm whose website, Google Business Profile and Checkatrade listing all agree on the same name and number, whose reviews are recent and specific, and whose site says plainly that they cover Loughborough. It can verify that business in seconds, so it names it. The firm two streets over, with a stale address and no reviews it can find, never enters the conversation. Same trade, same town, completely different outcome.

Why does ChatGPT name so few local businesses?

Because most businesses are hard to verify, and ChatGPT would rather stay quiet than name one it is not sure about. The 1.2% figure is not the model being fussy for its own sake. It is the model refusing to guess.

Put that number next to the old Google local “3-pack”, which surfaces a business around 36% of the time, and ChatGPT looks roughly thirty times narrower. Gemini and Perplexity sit somewhere in between, at about 11% and 7.4% of locations respectively, so this is not unique to OpenAI, but ChatGPT is the strictest of the lot. When the machine is not confident, it says nothing rather than risk a bad recommendation.

The businesses that do make the cut tend to share a short, unglamorous list of traits:

• Their name, address and phone number match everywhere ChatGPT looks.
• They have real, recent reviews on the platforms people actually use.
• They are mentioned on third-party sites and directories the model already trusts.
• Their pages spell out what they do, and where, in language a machine can lift word for word.
• Their website lets AI crawlers in and hands over structured data.

None of that is exotic, and none of it is expensive. Most businesses simply have not done it yet, because they built their web presence for Google in 2019 and have not touched the plumbing since. That gap is exactly why the opportunity is real right now, and why it will close as more of your competitors catch on.

How do you get ChatGPT to recommend your business?

Fix six things, in this order: your entity data, your reviews, your content, your structured data, your crawler access, and your third-party presence. Do them properly and you move from invisible to citable. This is the plan we run for clients, in priority order, and the order matters. There is no point earning reviews if your address does not match, and no point writing brilliant content if the crawlers cannot reach it.

1. Lock your entity: one business, one set of facts

Pick one exact version of your name, address and phone number, and make it identical everywhere online. ChatGPT cross-checks your details across your website, your Google Business Profile and the wider web. If your address reads “Suite 4” on one site and “Unit 4” on another, that is a reason to doubt you, and doubt means silence.

• Audit every listing you can find: website, Google Business Profile, Bing Places, Companies House, social profiles and directories.
• Standardise the format down to the punctuation. Decide on “Ltd” or “Limited” and use it everywhere.
• Complete your Google Business Profile in full, with the correct primary category, opening hours and service area. It is still the anchor for local AI answers, and a half-finished profile is a red flag. Our local SEO and Google Business Profile guide walks through the detail.

2. Get reviews, and keep getting them

Reviews and third-party social proof are the single heaviest input into AI recommendations, so treat this as core work, not a nice-to-have. One analysis of more than 23,000 AI citations found that around 57% of brand-evaluation citations pointed to reviews, listicles, forums and case studies, not the brand’s own website. In other words, what other people say about you counts for more than what you say about yourself.

• Ask every happy customer for a review. Google first, then the platforms that carry weight in your sector, such as Trustpilot, Checkatrade or Feefo.
• Aim for recency and rhythm. A steady trickle of recent reviews beats a wall of five-stars from three years ago.
• Reply to reviews, including the awkward ones. Your response rate is a signal in its own right, and it shows a real business paying attention.
• Encourage specifics. A review that names the exact job, the town and the outcome gives the model quotable language to work with, which is worth far more than a bare five stars and “great service”.
• The review signals that sway ChatGPT are the same ones that move Google, so this is not wasted effort. We cover that overlap in do Google reviews affect local rankings.

3. Publish answer-shaped content that names your category and your town

Write pages that answer the exact questions people ask, in plain language, and state clearly what you do and where you do it. ChatGPT lifts sentences it can quote, so give it clean, quotable sentences and stop hiding the useful bits inside marketing waffle.

• Lead each page with a direct, one or two sentence answer, then expand underneath. This page does exactly that, and it is not an accident.
• Name your category and location explicitly. “Emergency plumber in Leicester” beats “your local plumbing solutions” every time.
• Cover the boring, specific questions customers actually have: prices, areas served, response times, guarantees, what happens next.
• Keep it current. Fresh, recently updated pages get cited more often than stale ones that have sat untouched for years. This is the core of generative engine optimisation, which we explain from scratch in what is GEO.

4. Add structured data so machines can read you

Add LocalBusiness or Organisation schema, plus FAQ schema, so ChatGPT can parse your details without having to guess at them. Schema is the label on the tin. It tells a machine, in its own language, precisely who you are, what you offer and where you offer it.

• Mark up your name, address, phone number, opening hours, service area and category using LocalBusiness or Organisation schema.
• Add FAQPage schema to pages that have genuine question-and-answer sections, so those answers can be pulled directly.
• Keep the schema honest and matching your visible content. Marking up claims that are not on the page, or that contradict it, does more harm than leaving it off.

5. Let the AI crawlers in

Check that your robots.txt is not quietly blocking the crawlers that feed ChatGPT. This one is invisible, and it catches good businesses out constantly. You can do everything else on this list and still be locked out at the front door.

• OpenAI runs separate crawlers with separate jobs: GPTBot gathers training data, OAI-SearchBot builds the ChatGPT Search index, and ChatGPT-User fetches live pages during a chat.
• Blocking GPTBot does not block ChatGPT Search. They are controlled independently, so a rule you added last year to keep your content out of training may not be doing what you think.
• If you want to appear in ChatGPT, allow OAI-SearchBot and ChatGPT-User at a minimum. OpenAI lists the current user agents in its crawler documentation, so check yours against that.

6. Earn a presence on the sources ChatGPT already trusts

ChatGPT trusts third parties more than it trusts you, so get your name onto the sites it already reads. Around 85% of the sources behind brand answers are sites the business does not own. You cannot control those directly, but you can absolutely earn your way onto them.

• Get listed on the reputable UK directories for your sector, the ones real customers use, not spammy link farms that will do more harm than good.
• Earn genuine coverage: local press, industry publications, guest articles, supplier and partner pages that mention you by name.
• Claim and fully complete your profiles on the review and comparison platforms your customers already check before they buy.

What quietly keeps businesses invisible?

The usual culprit is not one dramatic failure but several small inconsistencies that add up to doubt. ChatGPT does not owe you a recommendation, and the moment your details stop lining up, it moves on to a business whose details do. These are the mistakes we see most often, and every one of them is duller than people expect.

• Mismatched details across the web. An old address on a directory you forgot you were listed on, a phone number that changed two years ago, a trading name that appears three slightly different ways. Each mismatch chips away at confidence until there is not enough left to name you.
• A blocked or accidental robots.txt rule. A developer added a blanket disallow during a rebuild, or a security plugin blocks AI bots by default, and nobody ever noticed. Your content can be genuinely excellent and completely unreadable to ChatGPT at the same time.
• Thin or ageing reviews. A dozen reviews from 2021 tells the model your best days are behind you. Silence on the platforms that matter in your trade tells it even less than that.
• Marketing-speak instead of plain answers. Pages stuffed with “bespoke solutions” and “market-leading service” give the model nothing it can quote. It cannot recommend what it cannot summarise in a sentence.
• No third-party footprint at all. If the only place your business is described is your own website, ChatGPT has nobody to check your story against, so it stays cautious and stays quiet.

The fix for every item here is already on the checklist above. The reason for listing them separately is that most owners assume they are invisible because of something clever and technical, when it is almost always something dull and fixable they have simply never thought to check.

How do you test whether ChatGPT recommends you today?

Ask ChatGPT the questions your customers would ask, and write down who it names. If you are not on the list, you have your answer. You do not need a fancy tool for this. You need ten minutes and a bit of honesty.

The method we use, and give to clients:

1. Write a list of eight to ten prompts a real customer might type. Mix category and location: “best [service] in [town]”, “who should I use for [job] near [area]”, “recommend a good [category] in [town]”.
2. Run each one in ChatGPT with web search switched on. Note every business it names, and in what order.
3. Run the same prompts more than once, and again on a different day. Answers vary between sessions, so look for patterns rather than reading too much into a single reply.
4. Click the sources it cites. Those citations tell you exactly which sites are shaping the answer, and therefore which sites you need to be on.
5. Repeat monthly and keep the notes. That log is your scoreboard, and it will show progress long before the phone starts ringing.

Read the results as a trend, not a verdict. What you want to see over successive months is your name appearing more often, higher up, and off the back of better sources. A single day where you drop out is noise. A steady climb, or a steady absence, is the signal.

When we ran this exact test on our own agency, for our own category, we came back invisible. Nothing. We build this for a living and ChatGPT still would not name us, because we had not done our own entity and third-party groundwork first. We fixed it in the order above and watched ourselves appear. If the test can catch us out, it is well worth running on you.

How long before ChatGPT starts naming you?

Give it two to three months for steady, repeatable recommendations, with the first small signs inside a few weeks. This is reputation-building, not a switch you flip, so anyone promising instant results is guessing or fibbing.

A realistic shape looks like this:

• Weeks 1 to 2: you fix the entity data, open up the crawlers and publish your first answer-shaped pages. Bot activity on your site starts to pick up.
• Weeks 4 to 6: the odd mention appears, usually in ChatGPT Search and Perplexity first, because they lean hardest on live retrieval.
• Months 2 to 3: recommendations become consistent across AI tools as your sources are re-crawled and your details are corroborated in more places.

Two things speed it up: reviews arriving at a steady pace, and third-party mentions landing on trusted sites. Two things slow it down: a thin review profile, and a website that guards its content from crawlers. The same groundwork also feeds the neighbouring channel, Google’s AI answers, which we cover in how to rank in Google AI Overviews, so the work compounds rather than pulling in two directions.

None of this needs a twelve-month contract. We work in scoped monthly engagements, send weekly updates so you can see exactly what is moving, and there is no lock-in. If you want to know where you stand today, and what the fastest path to being named looks like for your specific business, book a free growth review and we will run the test alongside you.

The short version

Be the business ChatGPT can retrieve, verify and attribute, and it will name you. Be vague, inconsistent or invisible to its crawlers, and it will happily name someone else. Fix your entity data, earn reviews and third-party mentions, write pages that answer the question, mark them up so machines can read them, and let the crawlers in. Then test with real customer prompts, wait two to three months, and test again. Most of your competitors have not started yet. That, in a sentence, is the whole opportunity.