How to Get Your Business Recommended by ChatGPT and Other AI Assistants

“Getting into ChatGPT” sounds like a new discipline, and a lot of new services are sold on that basis. Most of it is not new. When an assistant recommends a plumber or a café, it is usually because it searched the web for that question and summarised what it found. A business it cannot find, or cannot read, is not in the answer.

So this guide starts with the part every provider documents: which crawler fetches your pages for which assistant, and what happens if you block it. Then it covers what Google says about its own AI features, what helps an assistant describe your business accurately, and which popular tactics have no documented effect at all.

Every major assistant that answers with live web results uses a crawler with a published name, and most operators run separate crawlers for searching and for training models. That split matters, because it lets a business stay out of training data while remaining findable in answers.

Google's AI Overviews and AI Mode are the exception: they are built on Google Search, so they use the pages Googlebot has already indexed rather than a separate AI crawler.

Open yoursite.com/robots.txt in a browser. It is a short text file listing which crawlers may read which parts of your site, and it is the single most common reason a perfectly good business is missing from AI answers. A line that disallows a search crawler, or a blanket rule that disallows every crawler not named, removes you from that assistant's results.

OpenAI is unusually clear about the consequence: sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. It is just as clear that this is separate from GPTBot, the crawler that collects training data. Blocking GPTBot keeps your pages out of future model training; it does not remove you from ChatGPT search. Anthropic documents the same split between ClaudeBot, for training, and Claude-SearchBot, for search, and Perplexity says PerplexityBot is not used to collect content for training AI models.

Website builders, hosts and security services sometimes offer a single switch to block AI crawlers. Before turning one on, find out which crawlers it blocks. A switch that blocks search crawlers along with training crawlers trades a privacy preference for invisibility in answers, and most local businesses would not make that trade knowingly. OpenAI says a change to robots.txt can take around 24 hours to reach ChatGPT's search results.

Google's documentation for site owners says there are no additional requirements to appear in AI Overviews or AI Mode, and no special optimisation needed. A page must be indexed and eligible to be shown in Google Search with a snippet. It says explicitly that you do not need new machine-readable files, AI text files or special structured data.

That makes the useful work for Google's AI features the same work as for Google Search: pages that can be crawled and indexed, text that answers real questions, and, for a local business, a complete and accurate Business Profile, since that is what Google's local results draw on. If you want to keep something out of AI Overviews, Google points to the same snippet controls Search has always had, such as nosnippet.

An assistant asked for a recommendation searches, reads several pages and summarises them. No provider documents how it weighs those pages, but the practical consequence of how they work is clear enough: if your website says you open at 8, your Google profile says 9 and an old directory listing says you have closed, there is no single fact for an assistant to repeat, and a careful one may leave you out rather than repeat a wrong one.

So the highest-value work here is unglamorous. Use the same business name, address, phone number, hours and list of services on your website, your Google Business Profile, the equivalent listings Apple and Microsoft run for their own maps, and any directory or association page that lists you. Fix or remove old listings that are wrong.

Reviews belong here too. Google says the number of reviews and your ratings are part of how well known, or prominent, a business is in its local results, and those results are among the pages assistants read. Genuine reviews, collected by asking every customer the same way, help in both places.

An assistant can only repeat what is written down. A home page that says “quality service since 1998” over a photograph gives it nothing; a page that says what you do, for whom, where, and how to book gives it a sentence it can use. Write the services page a customer would ask an assistant about: each service you offer, the area you cover, your hours, how to contact or book, and anything that genuinely sets you apart.

Put that information in text on the page, not only inside images, PDFs or a menu that loads after the page does. A crawler that reads the page once, without clicking anything, should come away with the facts.

Answer the questions customers actually ask, in their words, with the answer first. “Do you work on weekends?” followed by “Yes, Saturdays 9 to 1” is exactly the shape an assistant can lift into a reply. This is the same advice that has always made pages useful in search, which is the point: it helps in both.

A new industry has grown up around “generative engine optimisation”, and much of it rests on claims no provider has made. Be sceptical of anything that promises a recommendation, because none of these companies publishes how it chooses one.

Ask the assistants yourself, the way a customer would: “Who is a good plumber in Leeds?” or “Where can I get a gluten-free birthday cake near Hove?” Turn on web search where the assistant offers it, since that is the mode that reads your site, and note which sources it cites. If your competitors appear and you do not, the cited pages show you where the assistant is getting its information.

Then check your analytics. Most analytics tools show the website a visitor came from, and visits from chatgpt.com, perplexity.ai and similar assistants are a direct measure of whether you are being cited and clicked. A trickle of those visits is usually the first sign the work above is landing.