Can the assistants read
your website at all?

Four things decide whether you can be cited.

Not four things that decide whether you are cited — that is a different and much longer conversation. These are the preconditions. Fail them and the question of what your page says never comes up.

• 01

  30 points

  Crawler access

  Whether robots.txt lets the answer-time agents through — OAI-SearchBot, ChatGPT-User, Claude-User, PerplexityBot, bingbot. These fetch a page while an assistant is writing its reply. Block one and you are out of that assistant's answers regardless of how good the page is.

• 02

  25 points

  Machine-readable structure

  Whether the page declares what it is: JSON-LD that parses, an Organization node with a name and sameAs links, one h1, headings that break the page into retrievable sections, and semantic landmarks separating the article from the navigation.

• 03

  25 points

  Answer readiness

  Whether there is anything on the page to quote. Text present in the server response rather than assembled by JavaScript, headings phrased as questions, FAQ markup, and paragraphs short enough to survive being lifted whole.

• 04

  20 points

  Authority signals

  Whether an assistant has grounds to attribute you: a title and description worth reusing, a canonical URL, outbound citations, named authors, and the about, pricing and case-study pages that answer questions about the company itself.

What does the checker actually do?

It fetches your homepage, robots.txt, sitemap and llms.txt exactly as a crawler would, then runs twenty-three checks against what came back. Nothing is guessed and nothing is generated — every point in the score traces to something in the response, and every finding names the value it read.

Does a good score mean ChatGPT will recommend me?

No, and any tool claiming otherwise is selling something. This measures whether the assistants can reach, parse and attribute your site — the preconditions. Whether you then get recommended depends on what the page says, who else says it, and what your competitors have published. A site that fails these checks cannot be recommended; a site that passes them can be.

Why does blocking GPTBot not cost the same as blocking OAI-SearchBot?

They do different jobs. GPTBot collects training data, so blocking it keeps your content out of a future model — a reasonable choice with a slow, diffuse cost. OAI-SearchBot and ChatGPT-User fetch live pages so ChatGPT can cite them in an answer it is composing right now. Blocking those removes you from answers immediately. Many sites blocked both with one pasted snippet, having only meant the first.

You only look at the homepage. Isn't that limited?

It is, and it's the honest limit of a free instant check. The homepage is where site-wide signals live — schema, crawler policy, entity declarations, the links to your other pages — so it's the highest-signal single page to read. Depth beyond that is what a full audit is for.

Do you store the site I check?

The report is held in memory for fifteen minutes so the fix list can be unlocked without re-running the check, then dropped. If you unlock the detail, we keep the email address and the domain it was run against.