Do AI Search Providers Publish Optimization Guidelines for Websites?

What does each provider publish?

Which official Google documentation explains AI search eligibility?

Google's documentation is the most complete, and it lives on Search Central:

• Guide to optimising for generative AI features (developers.google.com/search/docs/fundamentals/ai-optimization-guide, published May 15, 2026). It says optimising for generative AI search "is still SEO". Its technical basics are that pages must be indexed and eligible to show a snippet, crawlable including through your CDN, and readable after JavaScript rendering. It lists myths to ignore, including llms.txt, special markup, "chunking" and inauthentic mentions, and says non-commodity content likely matters more than anything else.
• AI features and your website. The same eligibility rule: an indexed page that can show a snippet can appear in AI Overviews and AI Mode, with no extra requirements. It explains that nosnippet, data-nosnippet, max-snippet and noindex limit what AI features can show.
• Spam policies. Spam now explicitly includes "attempting to manipulate generative AI responses in Google Search". Scaled content abuse covers mass-producing pages for every search variation.
• Search generative AI control. In Search Console since August 31, 2026, it lets a site include or exclude itself from AI Overviews, AI Mode and generative features in Discover. It does not affect model training.
• Google-Extended. A robots.txt token controlling use for Gemini training and grounding in Gemini Apps. It does not affect Search inclusion or ranking.
• Generative AI performance report. Impressions from AI Overviews and AI Mode by page, country and device, without queries or answer text.
• Guidance on third-party SEO tools and advice (June 2026), which warns that no tool or service can guarantee performance, including in AI experiences.

A critical look at the guide is in the review of Google's GEO guide.

How to use OpenAI documentation to understand AI search visibility

OpenAI publishes no guide to being cited or recommended by ChatGPT, but its documentation answers several practical questions:

• Crawler access (developers.openai.com/api/docs/bots). OAI-SearchBot powers ChatGPT search. Sites that block it are not shown in ChatGPT search answers, though they may still appear as navigational links. GPTBot collects training data, and ChatGPT-User fetches pages on a user's behalf. Blocking GPTBot does not remove you from search answers; blocking OAI-SearchBot does.
• Referral tracking. OpenAI's publisher FAQ says ChatGPT adds utm_source=chatgpt.com to referral links. That is the basis for measuring ChatGPT traffic in analytics, though not every link carries it.
• Agent access. The same FAQ said accessibility helped the agent in OpenAI's Atlas browser, which used ARIA labels and roles to understand pages. That agent began in OpenAI's Atlas browser, which was retired in August 2026; ChatGPT's browsing now runs in its desktop app and on the web.
• Shopping. OpenAI's help centre says product results are chosen independently and are not ads. Its commerce documentation sets out the product feed format, including required fields such as availability and seller name.
• Plugins. The plugin guidelines set rules for plugins in ChatGPT's directory, including a ban on metadata that steers the model toward one plugin.

Read together, these tell you how to be reachable, measurable and eligible. They say nothing about how ChatGPT weighs one page against another. Be wary of anyone citing "OpenAI's official ranking factors"; no such document exists.

Where to find official AI crawler documentation from Google, OpenAI, Anthropic and Perplexity

• Google: Search Central's crawler overview lists Googlebot and special-purpose tokens, including Google-Extended. AI Overviews and AI Mode use Googlebot, so blocking Googlebot also costs you your snippets and nearly all Search visibility.
• OpenAI: developers.openai.com/api/docs/bots covers OAI-SearchBot (ChatGPT search), GPTBot (training), ChatGPT-User (fetches a user asks for) and OAI-AdsBot, which only checks landing pages submitted as ChatGPT ads.
• Anthropic: its help centre article on crawlers (support.claude.com, article 8896518) describes ClaudeBot for training, Claude-SearchBot for search quality, and Claude-User for fetches a user asks for. Anthropic says blocking the last two may reduce visibility in Claude's answers.
• Perplexity: its developer documentation on bots describes PerplexityBot, which surfaces and links sites and is not used for training, and Perplexity-User, which acts for users and generally ignores robots.txt. It publishes IP ranges for firewall allowlists.
• Apple: its Applebot support article (support.apple.com/en-us/119829) explains that Applebot powers Siri, Spotlight and Safari search, and that Applebot-Extended controls training use.
• Microsoft: Copilot draws on Bing's index, so Bingbot access and Bing Webmaster Tools are what matter.

How the crawlers treat robots.txt is covered in how AI crawlers handle robots.txt, and testing access in how to test whether GPTBot, PerplexityBot and ClaudeBot can crawl your site.

What does Microsoft publish?

Microsoft publishes more practical writing advice than any other provider. Bing's Webmaster Guidelines were rewritten in February 2026 to cover Copilot answers, including a section on prompt injection and AI manipulation that can lead to demotion or delisting. They also explain NOARCHIVE, which keeps content out of Copilot answers, and NOCACHE, which limits use to titles, snippets and URLs. Microsoft Advertising's October 2025 guide to content for AI search answers recommends descriptive headings, question-and-answer pairs, lists and tables, and warns against answers hidden in tabs. Its January 2026 GEO guide for retailers covers feeds, crawled data and off-site data. Bing Webmaster Tools' AI Performance report shows citations in Microsoft's AI experiences. More is in the review of Microsoft's GEO guide and what Bing Webmaster Tools shows for AI.

What do the providers not publish?

• How sources are chosen. No provider explains how it weighs relevance, authority, freshness or reviews when citing or recommending.
• How often indexes refresh, beyond rough figures such as OpenAI's note that robots.txt changes take about a day to apply to search.
• Brand recommendation logic. No provider explains how it decides which brands to recommend for a general question. The closest is OpenAI's shopping guidance: when several merchants sell the same product, ChatGPT considers availability, price, quality and whether the merchant is the maker or primary seller, and its merchant FAQ says richer feed data gives better placement.
• Per-query data. Google and Microsoft offer site reports to any site owner, and Perplexity gives citation analytics only to publishers in its Publisher Program; none of them shows full answers.

That gap is why independent studies, and your own tests, fill in the rest. Treat vendor claims about secret ranking factors with care, and look for evidence from controlled tests. Penalties are covered in whether website practices can trigger penalties in AI search results.

How do you keep up with changes?

Official guidance changes often. Follow Google's Search Central blog and documentation changelog, OpenAI's ChatGPT release notes, Bing's Webmaster blog, and the crawler pages above. Re-read the crawler pages whenever you change robots.txt or firewall rules, since user agents and IP ranges change.

We have not seen any provider publish ranking guidance beyond what is listed here. This page is based on reading each provider's current documentation, and some pages may have moved since.

What might this look like in practice?

Consider an illustrative accounting firm, Lindqvist & Moore, whose web agency blocked "all AI bots" after reading a news story. Its marketing lead checked each provider's crawler documentation and found the rule had blocked OAI-SearchBot, Claude-SearchBot and PerplexityBot, which power search answers, along with the training crawlers. The firm kept GPTBot and ClaudeBot blocked, allowed the search crawlers, left Googlebot alone, and set up Search Console's Generative AI performance report and Bing's AI Performance report. It tracked utm_source=chatgpt.com visits in analytics. Within a month, the firm's pages began appearing again in ChatGPT and Perplexity answers.

What do people ask about official AI search guidelines?

Does OpenAI publish SEO guidelines for ChatGPT?

No. It documents crawlers, referral tagging, product feeds and plugin rules, but not how ChatGPT ranks sources.

What does Google require for AI Overviews?

An indexed page that is eligible to show a snippet. Google says no special markup or files are needed.

Where is Anthropic's crawler documentation?

In Claude's help centre, in the article describing ClaudeBot, Claude-SearchBot and Claude-User.

Does blocking GPTBot remove me from ChatGPT search?

No. GPTBot is for training. ChatGPT search uses OAI-SearchBot.