Score
Answerability Score
Retrieval does not read your page. It reads a fragment of it, out of order, with no surrounding context. This scores whether your paragraphs survive that, and names the specific ones that do not.
- Scores the fragmentsNot the whole page
- Names the weak paragraphsBy position
- Paste, never fetchNothing leaves the page
Check it
Paste it here
Result
- Paragraphs —
- Start with an unresolved “it” or “this” —
- Over 80 words —
- Heading-like lines —
- Average paragraph length —
- Weakest paragraph —
A paragraph is answerable if a stranger could read it alone and learn something true. That is exactly the condition retrieval puts it in.
What the result means
A paragraph starting with “It” is a broken fragment
Retrieved alone, “It costs one credit” names nothing. The subject lived in the previous paragraph, which was not retrieved. Repeat the noun; the redundancy is the point.
Long paragraphs get truncated mid-thought
Past roughly eighty words a paragraph will be split, and the split rarely lands where the meaning does. Shorter blocks survive chunking intact.
Headings are retrieval anchors, not decoration
They carry disproportionate weight in matching a question to a fragment. A page with no headings is a single undifferentiated block to a retriever.
Write the answer in the first sentence
Assistants quote the opening of a fragment far more often than its conclusion. If the answer arrives in sentence four, it usually does not arrive at all.
Questions
Above 80 means most paragraphs would survive being retrieved alone. Below 50 usually means long paragraphs, missing headings, or heavy use of pronouns across breaks.
No. It measures something different: whether a fragment of your page still makes sense once separated from the rest of it.
Because retrieval breaks the page apart. Repetition that looks redundant to a human reading top to bottom is what keeps a fragment meaningful on its own.
Want this run against your whole site?
Tell us the page your assistant lives on and the questions you care about. A person runs them and sends back the raw answers.