No one can guarantee that an AI assistant will cite you, but some things are within reach and documented by the companies themselves: pages that can be indexed, with clear answers and sources; no accidental blocking of the right crawlers; and a description of your company that is the same everywhere. No special files or dedicated markup are needed: for Google, the official documentation says so. Anything beyond that is a promise no source confirms.
Do you need structured data or special optimization for Google’s AI answers?
No. For AI Overviews and AI Mode, Google states that there are no additional requirements or special optimizations: the Search requirements apply, meaning an indexed page that complies with the policies. You don’t need new machine-readable files or dedicated schema.org markup. Google’s guide to optimizing for AI search adds that you don’t have to break content into small pieces or write in a particular way for AI. Even if a page meets every requirement, the guide notes that Google doesn’t guarantee it will crawl, index, or serve it.
Structured data is still useful for Search: it helps Google understand the content of a page, but it has to describe what users actually see. Google’s policies prohibit marking up content that is not visible or is misleading.
Does llms.txt matter?
Not for Google Search: the same guide says you don’t need to create AI text files because Search doesn’t use them, and that creating them for other services will neither harm nor help your visibility in Google. For the other assistants, here is where things stand: in the official documentation we reviewed (OpenAI, Anthropic, Perplexity, Microsoft), llms.txt is not listed as a file their systems read. So we have no confirmation that it is used, and we don’t present it as if it were. If you publish one, treat it as a low-cost extra, not a lever.
How do robots.txt and AI crawlers work?
The companies distinguish between several crawlers with different purposes. Before blocking anything, it helps to know what each one does:
- OpenAI. OAI-SearchBot powers ChatGPT search: sites that opt out won’t appear in ChatGPT search answers, at most as a plain link. GPTBot collects content to train models and is controlled separately. ChatGPT-User acts on a user’s request and, OpenAI warns, robots.txt rules may not apply.
- Anthropic. ClaudeBot, Claude-User, and Claude-SearchBot have distinct roles (collection for models, user requests, search quality) and respect robots.txt.
- Perplexity. PerplexityBot is used to surface and link sites in results, not for model training; Perplexity-User acts on a user’s request and generally ignores robots.txt.
- Google. Google-Extended is a robots.txt token for using content to train Gemini; it does not affect inclusion or ranking in Search.
Opting out of training and staying citable are therefore separate choices. Check your CDN and firewall too: a block there has the same effect as one in robots.txt.
Why does consistent organization data matter?
Google says Organization data on your home page can help it understand and disambiguate your organization in search results. Recommended properties include name, URL, logo, address, contact details, and sameAs, which schema.org defines as the URL of a reference page that unambiguously indicates the item’s identity: for example, the company’s official profiles.
The practical work is simple: the same name, description, and contact details on your site, listings, and profiles. A Google knowledge panel, on the other hand, can’t be obtained on request: it is generated automatically from information gathered across the web, and the subject can ask for changes without any guarantee.
What kind of content gets cited?
No one can say in advance, but the official guidance points the same way. Google asks for content written for people, with clear authorship and first-hand experience, and its guide talks about original, useful content with a point of view of its own. Microsoft, introducing the AI Performance report in Bing Webmaster Tools, recommends clear headings, tables, FAQ sections, examples, data, cited sources, and regular updates. That report shows how often pages are cited in Bing and Copilot AI answers.
In practice: a direct answer up front, details right after, linked sources, an honest update date. Google also warns that chasing inauthentic mentions around the web isn’t as helpful as it might seem.
What can’t be promised
No ranking, no citation, no timeline. OpenAI itself doesn’t guarantee inclusion in ChatGPT search. What you can do is remove obstacles, write pages that deserve to be used as a source, and track what happens with verifiable tools such as Search Console and the Bing report.
If you want to know how your site is read today, that is part of our SEO, GEO, and AEO service.