Free check · five per hour per network

Can ChatGPT read my site?

Paste the homepage. The checker fetches it the way a crawler does: raw HTML first, then the files and bot user-agents that decide whether an assistant can quote the page.

What the score is made of

Readable is worth 30. When a headless browser is bound, the checker compares the words in the raw HTML with the words after JavaScript runs. Full marks need about 80 percent of the rendered text to already be in the HTML, the H1 included, and the opening lines overlapping. Without that browser, the readable score is provisional: a page with 200 words and one H1 in the raw HTML can earn 27 of the 30, and the result says so.

Crawlable is worth 15. The checker reads robots.txt and asks whether GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Applebot-Extended, and CCBot are allowed. It then fetches the URL as GPTBot, ClaudeBot, and PerplexityBot. A 401, 403, 429, 503, or a challenge page fails that bot. If the same URL answered a normal browser and refused the bot, the result says the block is likely. Google-Extended is reported and not scored, because it is a control token, not a crawler.

Discoverable is worth 30. That covers a title of 10 to 65 characters, a meta description, a canonical link, the html lang attribute, exactly one H1, Open Graph title and description, JSON-LD in the raw HTML, a sitemap with URLs and lastmod, and llms.txt. llms.txt is only 5 points. Evidence that assistants actually request it is still thin, so it cannot carry the score.

Fast is worth 25 and comes from the PageSpeed Insights mobile lab score plus Chrome UX Report field data. If no API key is configured, Fast is left out and the score is out of 75. The checker does not invent a speed number, and it does not quietly rescale the other points to 100.

Limits

Private addresses, localhost, link-local and metadata hosts, and any port other than 80 or 443 are refused. Redirects are checked again at each hop, at most three times. Each response is capped at 3 MB and each fetch waits at most 15 seconds. The same URL is cached for 24 hours, and a cached result does not use your hourly quota. Results are stored so the link keeps working. The stored IP is a short hash, not the address itself.

The homepage and up to two internal links found in the raw HTML are included. The homepage carries most of the readable weight.