How do AI assistants choose which businesses to recommend?

Most AI assistants recommend businesses by searching live when a customer asks, reading what comes back (websites, business listings, reviews, articles) and summarising it with citations. Each platform uses a search index it runs or licenses, and most send their own crawler to your website, so what they can recommend depends on whether that crawler can read your site and what the wider web says about you. None of them publishes a full ranking formula, so be wary of anyone who claims to know it.

Key takeaways

• ChatGPT, Perplexity, Claude, Gemini, Copilot and Google's AI Overviews and AI Mode all document some form of live web search with source links.
• Each platform has its own crawler; blocking it (often by accident) can keep you out of that assistant's answers.
• Search crawlers and AI-training crawlers are separate for OpenAI and Anthropic, so you can block training and still appear in answers.
• Google's and Microsoft's assistants sit on top of their own search engines: Google's AI features use the Google index and Copilot uses Bing.
• Your Google Business Profile, reviews and mentions on other sites feed into many of these answers.

What happens when someone asks an AI assistant for a business?

The assistant usually turns the question into one or more web searches, reads the top results and writes an answer from them.

This process is called retrieval-augmented generation or grounding: the model fetches current information and bases its answer on it, rather than relying only on what it memorised in training. Google describes its AI features as relying "on our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index" (Google Search Central, 2026).

Two documented details matter for businesses:

• The question gets rewritten. Google's query fan-out issues "multiple related searches across subtopics and data sources" (Google Search Central, 2025). ChatGPT "typically rewrites your query into one or more targeted queries" (OpenAI Help Center, 2026). So a customer's question about "a good florist for a wedding in York" (an illustrative example) may become several searches about wedding florists, York, reviews and prices. Pages that clearly answer those sub-questions have more chances to be picked.
• Location is used. ChatGPT says it "may use an approximate location based on your IP address to provide relevant local results" (OpenAI Help Center, 2026). Google's local results are based on "relevance, distance, and popularity" (Google Business Profile Help, 2026).

How does each AI assistant find and cite businesses?

Each platform documents its search feature and crawlers differently; this section sticks to what they have published.

ChatGPT (OpenAI)

• Where answers come from: ChatGPT search "sometimes partners with other search providers", sending them rewritten queries, and ranks results "using multiple factors intended to help users find relevant, reliable information". For local questions it "may share your location with trusted third-party providers that help provide more local information", such as business listings (OpenAI Help Center, 2026). OpenAI's help page does not name these providers.
• Crawlers (OpenAI, 2026):
  • OAI-SearchBot is "for search". Sites that opt out "will not be shown in ChatGPT search answers", though they can still appear as navigational links. Changes take "~24 hours" to apply.
  • ChatGPT-User visits a page when a user's request needs it. It "is not used for crawling the web in an automatic fashion", and because actions are user-initiated, "robots.txt rules may not apply".
  • GPTBot collects content that may be used to train OpenAI's models. Blocking it does not affect search, because "each setting is independent of the others".
• Eligibility: to appear, allow OAI-SearchBot and make sure your host or CDN does not block OpenAI's published IP addresses. "Placement is not guaranteed" (OpenAI Help Center, 2026).

Perplexity

• Where answers come from: Perplexity runs its own web index. Announcing its Search API in September 2025, it described "an index covering hundreds of billions of webpages" and said the API uses "the same global-scale infrastructure that powers Perplexity's public answer engine" (Perplexity, 2025).
• Crawlers (Perplexity, 2026):
  • PerplexityBot surfaces and links websites in Perplexity's search results and "is not used to crawl content for AI foundation models".
  • Perplexity-User fetches pages when a user asks a question. Because a user requested it, it "generally ignores robots.txt rules".
  • Changes to robots.txt may take up to 24 hours to take effect.

Google AI Overviews and AI Mode

• Where answers come from: the Google Search index and Google's core ranking systems (Google Search Central, 2026). To be shown as a supporting link, "a page must be indexed and eligible to be shown in Google Search with a snippet" (Google Search Central, 2025).
• Local data: Google says AI responses can include "information about local businesses", and that Google Business Profiles and Merchant Center "can help your products and services to be visible in both AI responses and other Google Search results" (Google Search Central, 2026).
• Crawler: ordinary Googlebot. AI features are part of Search, so there is no separate AI Overviews crawler. You limit what appears with the usual nosnippet, data-nosnippet, max-snippet or noindex controls (Google Search Central, 2025).
• No special requirements: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary" (Google Search Central, 2025).

Gemini (the Gemini app)

• Where answers come from: Google describes "grounding (providing content from the Google Search index to the model at prompt time...)" in Gemini Apps (Google Search Central, 2026). So Gemini's live answers draw on the same index as Google Search.
• Maps data: Google's developer documentation for Gemini's "Grounding with Google Maps" feature says it uses data on "over 250 million places worldwide", including "places, reviews, photos, addresses, opening hours", and that place answers are "based on Google user reviews and other Maps data" (Google AI for Developers, 2026). That page documents the developer API, not the consumer app; press coverage reports that the Gemini app also shows Google Maps place information, but we have not found a Google help page that details exactly how.
• Crawler / control: Googlebot does the crawling. Google-Extended is a robots.txt token, with no separate user agent, that controls whether Google's crawled content is used for training Gemini models and for grounding in Gemini Apps and Vertex AI. It "does not impact a site's inclusion in Google Search nor is it used as a ranking signal" (Google Search Central, 2026).

Claude (Anthropic)

• Where answers come from: Claude can search the web, and "every response includes citations" (Anthropic, 2026). Anthropic has not published which search index powers this in its help pages. In March 2025 TechCrunch reported that Anthropic appeared to be using Brave Search, after Brave was added to Anthropic's list of subprocessors (TechCrunch, 2025). Treat that as reported, not documented.
• Crawlers (Anthropic, 2026):
  • ClaudeBot collects content that may be used for model training. Blocking it signals your future content should be excluded from training.
  • Claude-SearchBot "navigates the web to improve search result quality". Blocking it "prevents our system from indexing your content for search optimization".
  • Claude-User fetches pages when a user asks Claude a question. Blocking it "prevents our system from retrieving your content in response to a user query".
  • Anthropic says its bots honour robots.txt and support the non-standard Crawl-delay directive.

Microsoft Copilot

• Where answers come from: Bing. Microsoft's Bing Webmaster Guidelines describe how Bing "discovers, crawls, indexes, evaluates, and surfaces content across Bing search experiences, Copilot, and grounding API results" (Microsoft Bing, 2026).
• What Bing says helps: "URLs focused on a primary topic are more likely to be selected for grounding results", and you should "place essential information near the top" of the page. Structured data "may support clearer grounding but does not guarantee visibility" (Microsoft Bing, 2026).
• Crawler: Bingbot (Microsoft Bing, 2026). We found no separate Copilot crawler in Microsoft's documentation.
• Controls: Bing says NOARCHIVE "prevents content from being used in Copilot responses and grounding results", and NOCACHE limits Copilot to the URL, title and snippet (Microsoft Bing, 2026). Check your site doesn't set these by accident.

Summary table: where answers come from and what you control

What do practitioners observe, as opposed to what's documented?

The section above is what the platforms themselves say. The points below are widely reported by SEO and local-marketing practitioners but are not confirmed by the platforms. Treat them as reasonable working assumptions, not rules.

• Third-party mentions carry weight. Practitioners commonly report that assistants name businesses that appear in "best of" lists, local news, trade directories and review sites. Google does confirm its AI features "can show what's being said about products and services across the web, including in blogs, videos, and forum discussions" (Google Search Central, 2026), but nobody has published how much each source counts.
• Consistency helps. Matching name, address and phone details across listings is long-standing local SEO advice. No AI platform documents a consistency check, but contradictory information gives an assistant less reason to be confident in you.
• Answers vary between runs. Asking the same question twice can produce different businesses. Measure over many runs, not one screenshot.
• ChatGPT's providers. Several analyses by SEO tool vendors have tried to identify which search and local-data providers ChatGPT uses. OpenAI's own help page confirms that it uses third-party providers but does not name them, so we don't repeat the vendor claims here.

What should I do with this?

Make sure every assistant's search crawler can read your site, then put your effort where all of them look.

1. Check robots.txt for accidental blocks on OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot or Bingbot. Our AI visibility checklist has a tested example.
2. Check your host or CDN isn't blocking AI crawlers by default. Some firewall and "bot protection" settings do; OpenAI specifically asks that your host or CDN allows its published IP addresses.
3. Verify your site in Google Search Console and Bing Webmaster Tools, so you can see indexing problems in the two indexes that feed Google's AI features, Gemini and Copilot.
4. Complete your Google Business Profile and Bing Places listing, and keep hours and services current. See our Google Business Profile and reviews guide.
5. Make key facts easy to extract: one page per service, essential information near the top, prices and service areas in plain text.
6. Build genuine mentions on reputable local and industry websites.

How SearchCrow helps

SearchCrow asks the major assistants the questions your customers ask, records whether you are named and which sources each one cites, and shows which of those sources you are missing from. It's a quicker way to run the checks in this guide each month; the guide works on its own too.

FAQ

How does ChatGPT decide which businesses to recommend? When ChatGPT searches the web, it rewrites the question into targeted queries, sometimes sends them to third-party search providers, and ranks results using multiple factors. For local questions it can use third-party providers for business listings. To appear in ChatGPT search answers your site must allow OpenAI's OAI-SearchBot crawler. OpenAI says placement is not guaranteed and has not published its ranking factors.

Where do Google AI Overviews and AI Mode get their information? From Google's search index. Google says these features use its core Search ranking systems to retrieve relevant pages and may run several related searches (query fan-out). To be eligible as a supporting link a page must be indexed and eligible to show in Google Search with a snippet. Google Business Profile and Merchant Center data can also feed local and shopping details.

Does Microsoft Copilot use Bing? Yes. Microsoft's Bing Webmaster Guidelines say they describe how Bing discovers, crawls, indexes and surfaces content across Bing search, Copilot and grounding API results. So being crawlable by Bingbot and indexed in Bing is the foundation for appearing in Copilot answers. Bing also documents meta tags that limit how content is used in Copilot responses.

If I block AI training crawlers, will I disappear from AI answers? Not necessarily. OpenAI and Anthropic separate training crawlers (GPTBot, ClaudeBot) from search crawlers (OAI-SearchBot, Claude-SearchBot), and each setting works independently. Google-Extended is different: it also controls grounding in Gemini Apps, so blocking it may reduce Gemini's use of your content. It does not affect Google Search or AI Overviews.

Do AI assistants use reviews when recommending businesses? Several do, in documented ways. Google's Gemini API documentation says place answers grounded in Google Maps are based on Google user reviews and other Maps data, and Google's AI features can show what is being said about businesses across the web. The exact weighting of reviews in any assistant's recommendations has not been published.

Sources

• OpenAI, Overview of OpenAI crawlers
• OpenAI Help Center, ChatGPT search
• Perplexity, Perplexity crawlers
• Perplexity, Introducing the Perplexity Search API (25 September 2025)
• Google Search Central, AI features and your website
• Google Search Central, Optimizing for generative AI features on Google Search
• Google Search Central, Google's common crawlers (Google-Extended)
• Google AI for Developers, Grounding with Google Maps
• Google Business Profile Help, Tips to improve your local ranking on Google
• Anthropic, Does Anthropic crawl data from the web, and how can site owners block the crawler?
• Anthropic, Enabling and using web search
• TechCrunch, Anthropic appears to be using Brave to power web search for its Claude chatbot (21 March 2025)
• Microsoft Bing, Bing Webmaster Guidelines
• Microsoft Bing, Which crawlers does Bing use?