What a page needs to be quoted. A checklist, with the source for each item.

This is the one hard requirement every vendor documents. Allow these in robots.txt, and make sure the host or CDN does not block them:

• OAI-SearchBot, for ChatGPT search. OpenAI's crawler overview says a blocked site is not shown in ChatGPT search answers, and that a robots.txt change takes about 24 hours to apply.
• Googlebot, for Google Search, AI Overviews and AI Mode. Google lists allowing crawling in robots.txt, and by any CDN or host, among its best practices.
• Claude-SearchBot and Claude-User, for Claude. Anthropic's crawler page says blocking them may reduce a site's visibility in Claude's search answers.

The training crawlers (GPTBot, ClaudeBot) are separate settings, according to the same two pages. Blocking them is a choice about training, not about search answers.

For Google's AI answers, a page must be indexed and eligible to show with a snippet. Google's AI guide adds: "Indexing and serving aren't guaranteed." A noindex or nosnippet tag removes a page from those answers.

We found no OpenAI or Anthropic tool to submit a page, or to see whether a page is in their index.

Google asks site owners to make sure important content is available in textual form. Hours inside a photo of a sign, or a menu that is only a PDF scan, cannot be quoted.

Whether ChatGPT-User and Claude-User run JavaScript before reading a page is not documented. Our inference: text that is already in the HTML when the page loads is the safe choice.

Google's AI guide says content that people find unique and useful will likely influence a site's presence in AI search more than any other suggestion in the guide. It warns against making a page for every variation of a search.

One academic study points the same way. The GEO paper (KDD 2024) tested ways of rewriting pages and reported that some raised visibility in generative search answers by up to 40%. Its tests were not on local business questions, and they predate today's assistants.

For a local business, specific means the facts a customer asks for: what the business does, where, when it is open, what it charges if it says so, and which areas it serves.

For Google it has a documented job. LocalBusiness structured data tells Google a business's hours, address and phone, and makes a page eligible for rich results. Google does not guarantee they show.

For AI answers, Google's AI guide says: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." It also says the markup must match the text people see.

OpenAI and Anthropic do not document whether their systems read schema.org markup. Claims that it raises the odds of a ChatGPT or Claude citation are unverified.

Not on Google. Its AI guide says you do not need AI text files to appear in Search, and that Google Search ignores them.

We found no OpenAI or Anthropic documentation saying ChatGPT or Claude use llms.txt to choose what to quote. The file is harmless. Any claim that it improves recommendations is unverified.

No vendor documents a freshness rule for local answers. There are hints. Claude's search tool passes each result's page age to the model. OpenAI's help page tells people to check when a source was published or updated.

Our inference: a visible, true "last updated" date and current hours help an assistant trust a page. A new date on old facts does not.

Google says complete and accurate information makes a business more likely to show in local results. No vendor documents a matching rule for assistants.

Assistants combine several sources in one answer, so a wrong number on one site can end up in it. In Ryan Doser's test, 7 of 250 recommendations gave a phone number that did not match the business's Google listing.

An OnMap page is built to this checklist: open to these crawlers, plain text, and structured data that matches the text. It cannot make an assistant choose the page. See how it works.