Best passage
Which stretch of your page a retriever would actually pick, and how convincingly it answers the question you want to own.
AI search sends out a retriever: it reads your page, picks the strongest passage, and decides whether it is worth bringing back as the answer. This free check does the same job with the same kind of models. Drop in a URL and the question you want to be the answer for. If your page fails the sniff test, the evidence card in your inbox shows exactly which sentence to add.
No dashboard to learn. The models do the work, then hand you the sentence to add.
It pulls the real page, strips the chrome, and splits the content into overlapping passages, the same units AI search retrieval works in. No guessing from your sitemap.
A cross-encoder asks the only question that matters: does this passage answer it? SPLADE then weighs every term, so the verdict comes with evidence rather than vibes.
The verdict renders on the page straight away. Confirm your email and the full card lands in your inbox: matched terms, missing concepts, and the exact sentence to add.
When an answer engine picks a citation, it is not judging your whole site. It is judging one passage against one question, in milliseconds, against everyone else's best passage. Most pages lose that fight without knowing it was on. Retriever shows you the fight, the scorecard, and the fix.
Which stretch of your page a retriever would actually pick, and how convincingly it answers the question you want to own.
The terms the question demands that your page never engages, ranked by weight. Usually the fix is one honest sentence.
Optional competitor compare: their page scored against the same question, so you know whether you are losing on relevance or on reach.
The way AI search retrieval does. It fetches your live page, splits it into overlapping passages, and scores each one with the same family of reranker models AI search uses to pick citations: a cross-encoder for whether the passage answers the question, and SPLADE for term-level evidence. Your strongest passage wins and gets the verdict.
It is the relevance of your strongest passage to the question you want to be the answer for. Strong means the passage reads as the answer. Weak means it reads as background, and a retriever sent out for that question has no reason to bring your page back. The card shows the raw logit and how many passages were scored, so you can see the conviction behind the verdict.
Yes. The verdict and top missing concept render on the page. Confirming your email unlocks the full evidence card: the matched-terms table, the SPLADE breakdown and the numbered fix list. Double opt-in, and nothing is stored until you click the confirmation link. Fair-use caps apply: three checks per email per day.
The winning passage itself, a matched-terms table weighing your page against the question term by term, the missing concepts ranked by how much they matter, advice on answer shape, and the exact sentence to add. If you ran the competitor compare, their page is scored head-to-head next to yours.
This free check scores one page against one question. The paid work runs the same evidence across your whole site and every question your buyers actually ask, then fixes the losing passages and measures the lift. That is the actual job.