Quick answer: To get recommended by ChatGPT, Gemini and Perplexity, make your business facts identical wherever they appear, publish pages that answer customer questions in their opening lines, and let the AI search crawlers read your site. Then earn reviews and mentions on other trusted sites, and test a fixed set of prompts every month to see what the assistants say about you.
How AI assistants decide which businesses to mention
Assistants combine what the model learned in training, which can be months or years old, with live retrieval: searching the web or an index at the moment of the question and summarising what they find. You have far more influence over the second.
Gemini and Google’s AI features draw on Google’s index, Copilot draws on Bing, and ChatGPT and Perplexity run their own search crawlers. In every case your chances rise when the assistant can read your pages, finds the same facts wherever it looks and sees other sources backing you up.
Step 1: settle your entity facts
An entity is anything a machine can identify without confusion, such as your business. Write down, in one document, the facts you want every assistant to repeat, and check every listing and page against it.
- Exact business name, spelled the same way everywhere
- Full address and the areas you serve
- Phone and WhatsApp numbers, email address and website URL
- Opening hours
- A two-sentence description of what you do and for whom
- Core services, named the way customers search for them
- Links to your official social profiles
Step 2: make your business details consistent everywhere
Assistants cross-check. If your website shows one address, your map listing another and an old directory a number you no longer use, the safest choice for a model is to leave you out or hedge.
The listings that matter most
- Google Business Profile: feeds Google Search and Maps, which Google’s AI features can draw on; our Google Business Profile guide covers set-up.
- Bing Places for Business: feeds Bing’s local results, which Copilot can draw on.
- Apple Business Connect: controls how you appear in Apple Maps, Siri and other Apple apps.
- Industry and local directories: keep the ones customers actually use, and correct or remove outdated entries.
- Social profiles: the same name, description and contact details on every active account.
Step 3: write answer-first pages
Assistants quote passages, not whole pages. An opening about your commitment to quality gives a model nothing to lift; a direct answer does. This is the heart of how we approach AI SEO for service businesses, starting with service, location and about pages.
- Put the main answer in the first one or two sentences under each heading.
- Use headings that mirror real questions, such as “How long does a website redesign take?”
- Define terms in plain language before using them.
- Use tables for comparisons and lists for processes.
- Be specific about where you work, who you serve and what is included.
Step 4: add FAQs based on real questions
Collect questions from sales calls, WhatsApp chats, enquiry forms and reviews; each one is a prompt someone could type into an assistant. Add four to six to each relevant page, with short, direct answers that help even a reader who never contacts you.
Step 5: add structured data
Structured data labels the facts on a page so machines can read them without guessing. Google says no special schema is needed for its AI features, so treat markup as a way to remove ambiguity, not as a shortcut.
- Organization or LocalBusiness with name, address, phone, logo and sameAs links to official profiles
- Service markup on service pages
- FAQPage markup that matches the visible FAQ exactly
Mark up only what visitors can see. Test each change with Google’s Rich Results Test or the Schema Markup Validator.
Step 6: let AI crawlers read your site
An assistant cannot recommend what it cannot read. It is easy to block AI crawlers by accident through a copied robots.txt file, a security plugin or a CDN setting.
The key distinction is between training crawlers and search crawlers. OpenAI and Anthropic control them separately, so you can decline training and still appear in their search features; Google-Extended works differently, as the table shows.
| Crawler token | Run by | What it does | For AI recommendations |
|---|---|---|---|
| GPTBot | OpenAI | Collects content that may be used to train OpenAI’s models | Your choice; set independently of ChatGPT search |
| OAI-SearchBot | OpenAI | Surfaces websites in ChatGPT search | Allow |
| PerplexityBot | Perplexity | Surfaces and links websites in Perplexity results; Perplexity says it is not used for model training | Allow |
| ClaudeBot | Anthropic | Collects content that may be used to train Anthropic’s models; Claude-SearchBot handles search | Your choice; allow Claude-SearchBot |
| Google-Extended | A robots.txt control token, not a separate crawler, covering Gemini model training and grounding in Gemini Apps | Allow if you want Gemini Apps to use your content; it does not affect Google Search inclusion or ranking | |
| Bingbot | Microsoft | Crawls for Bing search, which Copilot draws on | Allow |
For the details, see OpenAI’s crawler overview, Perplexity’s crawler documentation and Google’s list of common crawlers. Anthropic describes its bots in its help centre.
What to put in robots.txt
A crawler with no rule naming it, and no blanket block, is already allowed. To make your intent explicit, you can add a group like this:
User-agent: OAI-SearchBotUser-agent: PerplexityBotUser-agent: Claude-SearchBotUser-agent: BingbotAllow: /
A crawler obeys only the most specific group that names it, so copy any Disallow lines from your general group into this one. Security plugins, firewalls and CDN settings can still block crawlers that robots.txt allows, so check your server logs.
Step 7: earn reviews and third-party mentions
Assistants are more willing to suggest a business that others vouch for. Your own website is only one voice.
- Ask every satisfied customer for a review on Google and on the platforms that matter in your sector.
- Reply to reviews, including critical ones, calmly and specifically.
- Never buy reviews, offer incentives for them or write them yourself.
- Seek genuine mentions from trade associations, event listings and industry publications.
If reviews are thin or old complaints dominate, the remedy is more genuine feedback, not fewer critics. A structured online reputation management plan can help rebalance the picture.
Step 8: consider an llms.txt file
llms.txt is an emerging convention, first proposed in 2024, for a plain Markdown file at the root of a site, such as example.com/llms.txt, that summarises the business and links to its key pages. It is not an official web standard, you should not assume any particular assistant reads it, and Google says no AI text files are needed for its AI features.
It is still a low-effort extra once the earlier steps are done. Keep it short and consistent with your fact sheet; some WordPress SEO plugins can generate one.
Step 9: test whether you get recommended by ChatGPT and others
Run a fixed prompt set on a schedule. It shows where you stand and whether your changes are working.
Build the prompt set
- Category and place: “Who are good [service] providers in [city]?”
- Problem-led: “How can a [type of business] [solve a problem]?”
- Comparison: “[Your business] or [competitor]: which suits [need]?”
- Brand check: “What do you know about [your business name]?”
Run and record the tests
- Use a logged-out window or a temporary chat so earlier conversations do not colour the answer.
- Run each prompt more than once, because answers vary between sessions.
- Log the date, assistant and prompt, whether you appear, whether the facts are right and which sources are cited.
- Repeat monthly and compare with the previous run.
When an assistant gets a fact wrong, trace it to the source, often an outdated listing, an old page or a third-party article, and correct it there. This testing loop is also where our generative engine optimization engagements begin.
Key takeaways
- Assistants suggest businesses they can read, verify and see confirmed elsewhere, so keep one fact sheet and apply it everywhere.
- Answer-first pages, real FAQs and accurate markup make content easy to quote.
- Allow the search crawlers you want to appear in, and treat training crawlers as a separate decision.
- llms.txt is optional; monthly prompt tests show what to fix next.
FAQ
How long does it take to get recommended by ChatGPT?
There is no fixed timeline. ChatGPT search can reflect a change once OpenAI’s search crawler revisits your pages, while what the model learned in training changes only with new model versions. Track your prompt set monthly.
Should I block GPTBot?
It depends on whether you accept your public content being used to train OpenAI’s models. OpenAI controls GPTBot and OAI-SearchBot separately, so you can block GPTBot and still allow OAI-SearchBot, the crawler behind ChatGPT search results.
Is llms.txt required to appear in AI answers?
No. It is an emerging convention rather than a standard, and Google says AI text files are not needed for its AI features. Treat it as a low-effort extra once your listings, pages and crawler access are in order.
Why does an AI assistant give wrong details about my business?
Usually because it found outdated or conflicting information, such as an old directory entry or a page you forgot to update. Correct the source, align it with your fact sheet and test again once the assistant has had time to revisit it.
Do I need a Bing Places listing if most customers use Google?
It is worth having. Copilot draws on Bing for its web answers, and a Bing Places listing is quick to set up once your Google Business Profile is accurate. Use identical details on both.
Find out what AI assistants say about you
Want to know how ChatGPT, Gemini and Perplexity describe your business right now? We can run a prompt audit, review your listings and crawler settings, and hand you a prioritised list of fixes. Start the conversation through our contact form or message us on WhatsApp.
