The free SOURCE check

See how AI assistants like ChatGPT find, trust and recommend your business, and what to fix first. Free, in about a minute.

Ask it the way a buyer would ask ChatGPT.

We email you a link. Open it and press Show my results.

Optional: what do you track today?
What do you track today? (optional)

It fills in the Evaluation letter, which a website can't show. Skip it and that letter says it's unknown.

We use your email only to send the link. We keep a scrambled copy that can't be turned back into your address, so each email gets up to three checks a day. It's kept about two days, then deleted, and we don't add you to any list. Your website and question go to OpenAI and Anthropic to run the check, and your website to Google (for speed), Wikidata and Wikipedia. We keep the result for 35 days so its page keeps working, and anyone who checks the same site with the same question, tracking answer and country in the next 30 days sees the same result. What you tick about tracking only sets the Evaluation line in your result, and is kept with it. Privacy policy

More about this check

What we can measure about your site in about a minute, three answers to a question your buyers ask, from OpenAI's API and Claude, with web search on, and a quick read of all six SOURCE letters.

What it checks

  • robots.txt: whether it lets in the crawlers that feed AI answers, and which training crawlers it blocks.
  • Crawler names: your homepage asked for as a browser and with six crawlers' names, and the answer each got. It's our server asking, not the crawlers themselves, so a firewall that checks where a request comes from may treat the real crawlers differently.
  • Your HTML: how much of your homepage's text is in the HTML before any script runs.
  • Indexing: the noindex and canonical tags on your homepage, up to 8 pages from your menu, and up to 3 of your pages the AI answers cite.
  • Structured data: the JSON-LD that parses, its types, and your business's profile links in it.
  • Real-visitor speed: Core Web Vitals from real Chrome visitors on phones over 28 days, never a lab score.
  • Three AI answers: your question put twice through OpenAI's API with web search on (close to ChatGPT, not the app itself) and once through Claude with web search on: whether they name you, who else the first names and puts first, and which pages they cite. Three answers to one question are a small, related sample, and the result says so.
  • Outside signals: whether Wikidata or Wikipedia know your business (matched by your website, not just your name), whether the other sites' pages the answers cite name you, and the review figures your own pages state, marked as your own claims. What's found counts; what isn't found never counts against you.
  • Your SOURCE letters: a quick read of Structure, Offsite, Uniqueness, Relevance, Credibility and Evaluation from those pages and those answers, by the Scorecard skill's own checks and scoring, with each letter's issues in what it read, and up to three problems: what stops AI reading your site first, then not being named in the answers, then the rest of what it measured, each with its letter and a first step. For your weakest letter, the book's 3-minute self-score and the full audit look further than this check can.

It reports what it measured, and never calls something missing unless the measurement shows it. What it can't tell you: how your reviews, rankings and press coverage look, what Gemini, Perplexity and Google's AI Mode say about you, or what the assistants say across many questions and many askings. The SOURCE Scorecard skill, which comes with Be the SOURCE, checks all of that in your own Claude, and checks each letter in full. Asking Gemini and Perplexity needs your own API keys for them.