Every AI answer cites a source. Almost none of them verify what that source is worth. Lens closes that gap, at the document and claim level, with evidence attached to every flag.
49.9% of new English-language articles are AI-generated. Retrieval is drinking from that river. The gap is where hallucination hides.
Get early access to the source-scoring API. Design partner program: partner pricing, direct roadmap input, and one integration path that fits your framework.
URLs and spend, one file, no tags, no integration. Get back spend by measured quality band, thin-content exposure, and the named worst pages with the exact reasons.
The same engine and the same evidence, wired into the workflow you already run. Hosted, private cloud, or fully on your infrastructure.
The delivery shape changes. The engine underneath does not.
Not domain reputation. Not model self-assessment. Every score attaches to the exact passage it read from and the plain-English reason it fired. The score is challengeable at the sentence level, not the black-box level.
No third-party AI sits between your call and the score. The scoring path is Lens infrastructure end to end, and enterprise deployment does not change that. No hidden dependency, no vendor behavior change downstream of yours.
Same document, same result. Every score reproduces against the recorded release. That property is what makes the output defensible under finance, brand, client, or regulator review.
A user's answer, a media buy, or a regulated workflow.
Inside your product, on a file you already have, or inside your stack.
An API response, a scored PDF audit, or a deployment plan.
Your engineers, your media team, or your compliance function.
Start with the one closest to your problem. Lane confusion is not a problem — we route between them if the fit changes after the intro call.