How to Get Your Business Recommended by ChatGPT and AI Search

A customer in Leeds asks ChatGPT who can fix a leaking boiler this week. Another asks Google's AI Mode which café nearby does good gluten-free food. Each answer names a handful of businesses. If yours is not among them, you never find out that the question was asked.

No form or fee buys a place in those answers. The platforms' own documentation describes something more ordinary: AI search tools still depend on crawlers, search indexes and business listings, and they choose sources they judge relevant and reliable. So the way to get your business recommended by ChatGPT and AI search is to make sure these systems can reach your website, that your site and listings say plainly what you do, where and for whom, and that genuine reviews back it up. Then you check, month by month, what the answers say. Below: what OpenAI, Google and Microsoft document, what is still guesswork, and a hypothetical way to measure where you stand.

How ChatGPT and AI search find a business

OpenAI's help page on searching the web with ChatGPT explains that ChatGPT search "sometimes partners with other search providers" and "typically rewrites your query into one or more targeted queries" before sending them; the page links to Microsoft's and Shopify's privacy statements. ChatGPT may estimate a general location from the user's IP address, and if memory is switched on it may use saved details, such as a diet or a home city, when it rewrites the query. For local answers it may share location with "trusted third-party providers" of information such as nearby business listings. Answers that use web search may include citations and a Sources panel.

Two things follow. The customer's question rarely reaches a search index as typed; "emergency boiler repair Leeds" is closer to what gets searched. And, on our reading of OpenAI's description, two customers asking the same thing can see different businesses, because location and memory change the query. That matters when you measure.

Google describes a similar process. Its guide to AI features and your website says AI Overviews and AI Mode may use a "query fan-out" technique, "issuing multiple related searches across subtopics and data sources". The two features may use different models, so "the set of responses and links they show will vary", and AI Overviews "often don't trigger" because Google shows them only when its systems judge that they add something to ordinary search results. If you want the customer's side of this, our guide on how to check an AI search answer covers it.

What the platforms say about getting included

Each platform publishes a short note on how a site becomes eligible. None offers a way to buy or request a place in ordinary answers.

Sources for the table: OpenAI's crawler overview and its help page above; Google's AI features guide and its Business Profile page on improving local ranking; the Bing team's announcement of the AI Performance report; Perplexity's crawler documentation; and Anthropic's crawler help page.

Paid placement exists only as advertising: OpenAI's crawler overview describes OAI-AdsBot, which checks "web pages submitted as ads on ChatGPT". This guide is about the unpaid answers.

First check: can AI search crawlers reach you?

This is the one area where a single setting can remove you completely. The large AI companies now separate the crawler that feeds their search answers from the crawler that collects material for training models:

• OpenAI: OAI-SearchBot is for search, and GPTBot collects content that may be used to train its models. OpenAI says each setting "is independent of the others", so you can allow search while opting out of training. Sites that block OAI-SearchBot "will not be shown in ChatGPT search answers, though can still appear as navigational links". Changes take about 24 hours to apply.
• Anthropic: ClaudeBot collects possible training data, Claude-SearchBot indexes for search, and Claude-User fetches pages when a person asks Claude a question.
• Perplexity: PerplexityBot surfaces websites in Perplexity's results and, according to Perplexity, "is not used to crawl content for AI foundation models".
• Google: there is no separate AI search crawler. Google says Googlebot rules in your robots.txt file (the file that tells crawlers what they may visit) are the control for AI features in Search, while a separate token, Google-Extended, covers training and grounding in some of Google's other AI systems.

The decision is simple to state. If you want to be found, do not block the search crawlers. Blocking training crawlers is a separate choice about how your content may be reused.

The common failure is accidental. Website builders, hosting firewalls and bot-protection services sometimes block crawlers they do not recognise. OpenAI and Perplexity both ask site owners to allow their published IP ranges, and Perplexity warns that Cloudflare's Bot Fight Mode can still challenge its bots even when a firewall rule lets them in. Checking this is a short job for whoever manages your site, and it comes first: good content cannot help if the crawler is turned away.

Say plainly what you do, where and for whom

If the AI turns a question into a search such as "heat pump installer north Leeds", it needs pages that answer that search in plain words. Google's AI features guide lists the basics: make important content available "in textual form", make sure your structured data matches the visible text, and keep your Merchant Center and Business Profile information up to date. Its May 2025 post on succeeding in AI search adds that people ask "longer and more specific questions" in these features and recommends "unique, non-commodity content".

For a small business, good content here is unglamorous:

• A clear page for each main service you offer, and the towns or postcodes you really cover.
• Prices or typical price ranges where you can give them, and what affects the price.
• Opening hours, response times and booking options that match your Business Profile.
• Proof you can stand behind: qualifications and trade registrations you hold, guarantees, and real examples of work.
• Straight answers to the questions customers ask on the phone, written the way they ask them.

Google's guidance on helpful, reliable, people-first content warns against "search engine-first content made primarily to gain search engine rankings" and says that, of the qualities it looks for, "trust is most important". A hundred thin AI-written pages about every suburb are the opposite of what any of these platforms describe.

Microsoft's AI Performance announcement is more specific: clear headings, tables and FAQ sections, evidence for claims, and regular updates. That is sensible advice, but it is guidance, not a published test of how much each change matters.

Your listings and reviews outside your website

For local questions, the answer often comes from listings rather than your own site. Google says businesses with "complete and accurate info are more likely to show up in local search results", and that prominence is based partly on "how many websites link to your business and how many reviews you have". Microsoft's announcement says local businesses can register with Bing Places for Business so that address, hours and contact details stay current and "eligible for inclusion in AI-generated responses". OpenAI does not name the listing providers it uses, so the safest course is to keep your name, address, phone number and hours identical everywhere they appear.

How much reviews influence ChatGPT's recommendations is not published. What is clear is how you may collect them. In the UK, the Digital Markets, Competition and Consumers Act 2024 makes fake reviews and reviews that hide an incentive a banned practice, under rules that took effect in April 2025 (see the CMA's fake reviews guidance page). The CMA's guidance (CMA208) gives examples of what is banned, including offering a refund or gift card to a customer who changes a negative review, and "buying reviews which look like they have been written by individual consumers but have in fact been generated by software applications (such as bots)". Emailing customers in general to ask whether they would like to leave a review, without steering what they say, is not prohibited.

Google's Maps content policy goes further for Business Profiles: merchants may not offer incentives for reviews or "selectively solicit positive reviews from customers". The practical rule is to ask every customer, pay nobody, and never write reviews yourself or with AI.

GEO advice: what is evidence-backed and what is guesswork

The work of trying to appear in AI answers is often called generative engine optimisation (GEO). Microsoft uses the term itself, calling its citation report "an early step toward Generative Engine Optimization (GEO) tooling". Much of the GEO advice sold to small businesses goes beyond what any platform has published, so it helps to sort it.

Speculation is not always wrong. A mention on a respected local site can bring customers directly, whatever it does for AI answers. Just know which advice rests on documentation and which on someone's observation of a system that changes without notice.

A worked example (hypothetical)

This example is invented to show the method. The business, the questions and the results are not real data.

Assumptions: a fictional two-van plumbing and heating firm in north Leeds. The owner picks 10 questions customers really ask, for example "emergency boiler repair Leeds", "who installs heat pumps in north Leeds", "plumber open on Saturday near Headingley" and "how much is a boiler service in Leeds". Once a month, the owner asks each question twice in ChatGPT with search, Google AI Mode and Microsoft Copilot, in a private browser window, naming the town rather than relying on location. That is 60 checks a month. At an assumed one minute per check, plus about 30 minutes to note the results, it takes about 90 minutes; at an assumed £30 an hour for the owner's time, that is £45 a month, or £540 a year.

For each answer the owner logs three things: whether the firm is named, whether the facts are right, and which source is cited.

Suppose that in month one the firm is named consistently in 2 of the 10 questions on ChatGPT and 3 of 10 on AI Mode, and three months later in 4 and 5. That is worth noting, but it does not prove the fixes caused it: models, competitors and questions change too. Counting only answers that repeat across both runs filters some of the noise. The more useful output is the problem list. Wrong hours and a missing service page cost customers whatever the AI does, and they are cheap to fix.

How to measure without fooling yourself

AI answers vary, and no business controls its position in them. OpenAI's help page says plainly that "search results and citations can be incomplete, outdated, or incorrect". So treat any single answer as a sample, not a score.

• Keep a fixed question list. Use the same wording each time, record the date, tool and town, and change the list only on purpose.
• Use the platforms' own reports. Google counts traffic from AI Overviews and AI Mode within the "Web" search type of the Search Console Performance report, alongside ordinary results. Bing Webmaster Tools introduced an AI Performance report in public preview on 10 February 2026; it shows how often your pages are cited in Copilot, Bing's AI summaries and select partner services. Microsoft says its grounding query data "represents a sample".
• Watch outcomes, not mentions. Google's May 2025 post suggests looking at sales, sign-ups and "information lookups about your business" rather than clicks alone. Asking new customers how they found you is still a simple, direct measure that no platform report replaces.

Allow time. Google notes that crawling can take "anywhere from several days to several months", depending on how often its systems decide a page needs refreshing, and OpenAI says robots.txt changes take about 24 hours to affect ChatGPT search. Judge changes over a quarter, not a week.

What it costs, and when to get help

Most of the cost is time. A crawler and firewall check is usually a one-off technical job. Service pages, listings and a fair way of asking for reviews take some days of attention to set up, then a little upkeep. The monthly check above is about 90 minutes.

The risks are mostly in shortcuts: paying for "guaranteed" AI rankings that no platform offers, buying or filtering reviews in breach of UK law and Google's policy, or publishing mass-produced pages that Google's guidance treats as search engine-first content. A decision rule that works for most firms: if your website is simple and you manage it yourself, the checks in this guide are within reach. If you have several locations, a website builder or firewall you cannot see into, or listings scattered across directories you no longer control, outside help will usually save time.

If you would like this handled for you, Evoogen can check how ChatGPT, Google and Copilot currently describe your business, fix the access, website and listing problems we find, and set up a simple monthly check you can keep running. We will not promise recommendations, because nobody can. See how Evoogen works with small businesses.