In short
A website is ready for AI agents when assistants like ChatGPT, Claude or Perplexity can read and understand it, if the owner wants them to: a robots.txt that tells training bots apart from search bots, an llms.txt file and structured data. The free WebIntacta scan checks this with a score kept separate from the security grade. Blocking training (GPTBot, for example) does not remove you from ChatGPT search answers.
Reviewed by Jose Soriano ·
ChatGPT, Claude and Perplexity already visit websites to answer their users, and some agents are starting to buy. Your site can let them in, block them or say nothing at all. The scanner tells you where you stand, with a score that is separate from the security grade.
Data protection. Controller: Jose Soriano Castro. Purpose: to send you the diagnostic PDF, email you for 30 days if a security flaw affecting your site is published (with a follow-up scan on day 30) and reply in writing to your question; offers only if you tick that box. Legal basis: your consent. Recipients: nobody, unless required by law. You can unsubscribe at any time with the link in every email or by writing to me. Rights: access, rectification, erasure, objection, restriction and portability. Privacy policy.
STEP 01
It is part of the free scan: you see the score next to the security grade.
STEP 02
Showing up in assistants' answers while not handing your content over for training is a perfectly valid choice. What matters is deciding.
STEP 03
If you want, I set it up: robots.txt, llms.txt and structured data, without touching your security.
Each company now separates the bot that collects content for training from the one that searches to answer questions. Blocking the first does not take you out of the answers; blocking the second does. The names, according to each company's documentation:
GPTBot collects content for training; OAI-SearchBot decides whether you appear in ChatGPT search; ChatGPT-User is the visit ChatGPT makes when a user asks it to, and OpenAI notes that robots.txt may not apply to those visits.ClaudeBot is the training bot; Claude-SearchBot the search bot; Claude-User the one that visits at the request of a Claude user.Google-Extended is not a bot but a token in robots.txt that says whether your content can be used to train Gemini. It does not affect Google Search or your rankings.PerplexityBot indexes for its answers and Perplexity-User visits at a user's request.A common setup: allow the search and user bots, and decide on the training bots separately. What llms.txt is and how to write one is in the llms.txt guide.
I practice what I preach: this site publishes llms.txt, answers in Markdown when asked with Accept: text/markdown, declares Content Signals in robots.txt and describes the business, the services and the prices in schema.org.
The AI agent readiness check is included in the free scan and in the security report.
Security grade and the problems visible from the internet, in a few minutes and with no sign-up.
Free
No sign-up, no commitment
An inside analysis with the read-only connector and a PDF with an action plan.
€129
excluding VAT$129 outside the EU
The same applies to websites and online stores outside the EU: the check is free. All prices
GPTBot is the bot that collects content to train OpenAI's models. Blocking it does not remove you from ChatGPT's answers, which use OAI-SearchBot and ChatGPT-User. Many sites block training and allow search.
No. Google-Extended only controls whether your content is used to train Gemini; Google states that it does not affect inclusion in Google Search or rankings.
A text file at the root of the website that sums up, for language models, what the site is and which pages matter. It is a recent proposal (llmstxt.org) and simple to add. More in the llms.txt guide.
Make sure the search bots (OAI-SearchBot, Claude-SearchBot, PerplexityBot) are not blocked in robots.txt, publish an llms.txt, and describe your business and products with schema.org structured data. The free scan tells you which of these you already have.
No. AI agent readiness has its own score and never adds to or subtracts from the security grade.
Especially for online stores: shopping agents need price, availability and product identifiers (SKU, GTIN) in the structured data.
Free external scan. No sign-up and nothing to install.
Data protection. Controller: Jose Soriano Castro. Purpose: to send you the diagnostic PDF, email you for 30 days if a security flaw affecting your site is published (with a follow-up scan on day 30) and reply in writing to your question; offers only if you tick that box. Legal basis: your consent. Recipients: nobody, unless required by law. You can unsubscribe at any time with the link in every email or by writing to me. Rights: access, rectification, erasure, objection, restriction and portability. Privacy policy.