AI training and AI answers

Out of AI training

While Exclude from AI training is on, Carya publishes the page with two signals:

• Rules in its robots.txt that turn away known AI-training crawlers, such as GPTBot, ClaudeBot, Google-Extended and CCBot.
• A noai, noimageai robots tag and a TDM reservation tag in the page itself.

Search engines aren’t affected, so your page still shows up in search. Turning away Google-Extended also keeps the page out of Gemini’s answers, but not out of Google Search.

A training crawler that comes anyway is turned away: Carya answers it with an error instead of your page, and lets it read only the robots.txt that says why. Your page’s Insights count the ones that came in an AI crawlers card, on every plan, from a day after the setting went live. Where the crawler’s company publishes its addresses (OpenAI, Anthropic, Common Crawl and Meta), Carya checks each visit against them, and a visit from anywhere else is shown as only saying it’s that crawler.

Show up in AI answers

With Show up in AI answers on, ChatGPT search, Claude and Perplexity can read your page and cite it when someone asks about what you do. Reading a page to answer a question isn’t training on it: this setting and the one above are separate.

Switch it off and your page’s robots.txt turns away OAI-SearchBot, Claude-SearchBot, PerplexityBot and meta-webindexer.

What your robots.txt says

Every page’s robots.txt states all three choices in one line, for example Content-Signal: search=yes, ai-input=yes, ai-train=no. Search says whether the page may appear in search results, ai-input follows AI answers, and ai-train follows AI training.

Change a setting

Open the page in the Editor, select nothing so the inspector shows the page, and switch Show up in AI answers or Exclude from AI training. Publish again for the change to reach your live page.

On a page built from Pieces, Exclude from AI training is in Page settings. Show up in AI answers isn’t there, and stays on.