Where do AI assistants learn about a business?
The companies behind the assistants read the web with programs called crawlers, and each one describes what its crawlers do in its own documentation. OpenAI defines three: OAI-SearchBot finds pages to show in ChatGPT's search answers; GPTBot reads pages for model training; ChatGPT-User opens a page when a user asks for it (OpenAI: overview of crawlers).
Anthropic makes the same distinction: ClaudeBot works for training, Claude-SearchBot for search results, and Claude-User opens a page when a user's question calls for it (Anthropic: web crawlers). Perplexity's crawler, PerplexityBot, reads pages to show them in search results and link to them; Perplexity states that it is not used for model training (Perplexity: crawlers).
Google works differently. The AI overviews in search results are fed by Google's search index. Google states that no extra requirement, special file or special markup is needed to appear in them; the requirement is that the page is indexed and can be shown in search results with a short snippet. That is an eligibility requirement, not a guarantee that you will appear (Google Search Central: AI features and your website).
An assistant's knowledge of your business is largely limited to what it can read about you on the web.
Why doesn't ChatGPT know my business?
There may not be a single reason, and assistants may not explain why they chose a source in an answer. Still, there are a few concrete obstacles you can check on your side.
Access. OpenAI states that a website closed to OAI-SearchBot will not be shown in ChatGPT's search answers (OpenAI: overview of crawlers). Your website's hosting or a security setting may have blocked these crawlers without you noticing.
A page to read. If your business name, what you do, your address, opening hours and contact details do not appear as text on a web page, the crawler has nothing to read. Details that exist only inside a photo or a poster may not count as text.
Time. A website that is new or has just changed takes time to be crawled. OpenAI notes that a change in access settings can take about a day to be reflected on the search side.
Even when all of this is in place, your business may not appear in an answer. The assistant decides which business is mentioned for which question; the providers' documentation lists no setting that guarantees a mention for a particular question.
Is my website open to AI assistants, and how can I tell?
Type your domain name into your browser's address bar, followed by /robots.txt. The short text that opens tells you which crawlers may read what on your website. The providers state in their documentation that their search and training crawlers follow this file; a crawler that opens a single page at a user's request may behave differently.
If the file says “Disallow: /” under the line “User-agent: *”, your website is closed to every crawler that has no separate rule of its own. If the same line sits under OAI-SearchBot, Claude-SearchBot or PerplexityBot, only that assistant's search is closed.
It is also possible to stay open to search and closed to training. OpenAI states that you can allow OAI-SearchBot and block GPTBot; Anthropic also keeps its training crawler and its search crawler separate. At Google, the name Google-Extended controls whether your content is used for training Gemini and in answers in Gemini apps; it does not affect your presence in Google Search (Google: common crawlers). This is a choice, and the decision is yours.
Google Search Console and Bing Webmaster Tools show whether your pages have been indexed. Bing's tool also lists which of your pages Copilot and Bing's AI summaries cite as a source (Bing Webmaster Blog: AI performance report).
What should a website have so that assistants can read it?
According to Google, no special method is needed: the work for AI features is the same as the work for a good search result (Google Search Central: guide to content for AI). In practice, look at four things.
Basic details as text. Put your business name, what you do, your address, phone number and opening hours on the page as plain text. Microsoft warns that text hidden inside tabs or expandable sections may not be read by AI systems (Microsoft Advertising: content for AI answers).
A direct answer to the question. Put the question your customer asks in a heading and the answer right below it: “Are you open on Sundays?”, “Do you offer delivery?”. Among Microsoft's recommendations in the same article are clear headings, tables and question-and-answer sections.
Your own information. Google's guide recommends first-hand information over text that repeats what everyone already knows, and says that producing a separate page for every search phrase breaks its rules when the aim is to steer search results or AI answers. That is why your own products, your own price list and your own working hours are valuable.
The same details everywhere. The name, address and hours on your website should match what your map listing and social accounts say. Details that differ from place to place can mislead readers and assistants alike.
If you have heard of llms.txt: it is a proposal for giving AI tools a short summary of a website (llmstxt.org); it is not a standard. Google states that no new file is needed for AI features; do not expect it to change anything by itself.
How does an out-of-date website show up in an assistant's answer?
An assistant may pass on what it has read. If your website still shows old opening hours or a product you no longer sell, that information can end up in the answer. A customer may arrive at a closed door or ask for a product that no longer exists; the assistant may not know the information is out of date.
Keeping things current is therefore two steps: first correct the information on your website, then let search engines know about the change. Bing recommends reporting changed pages through IndexNow, which can help the current version reach search and AI answers sooner (Bing Webmaster Blog).
Information that goes out of date includes opening hours, prices, products that sell out or arrive, and announcements. If changing them is a chore, the information on your website may go stale.
Is a separate page or a special service needed for assistants?
According to Google's guide, it is not: the guide recommends good content written for people, not separate content written for AI, and it adds that trying to collect artificial mentions around the web is not as helpful as it might seem (Google Search Central: guide to content for AI).
Be cautious about offers that promise you a place in assistant answers: even Google does not guarantee one. What can be done is the preparation described above.
Who does this preparation?
You can ask whoever built your website three questions: which crawlers does the robots.txt file leave open, are the basic details on the page as plain text, and are changes reported to search engines? The answers show where to start.
With IntegrAI, your website is prepared to be discoverable on Google and in AI assistants such as ChatGPT, Gemini, Claude and Perplexity — on your own domain. There is no separate fee for discoverability on Google and in AI assistants. You update your website from WhatsApp: you write, the change is prepared, the preview comes to you, you approve. This preparation does not mean an assistant will mention your business; it covers only preparing your website to be discoverable.