Building with AI?
Catch factual errors before they ship.
An independent, multi-model fact-checking API for the teams that can’t afford to ship a wrong claim.
Multiple models across five stages, grounded in independent sources. Opposing-side debate, three independent reviewers, full citation trail.
/extract1,000 / day
/assess5,000 / mo
/verify500 / mo
/ask5,000 / mo
Enter the 6-digit code
We sent it to .
Claim your free month
A free month of Developer access — $99 value, no card required.
Full Developer access is applied the moment you sign up.
New accounts only · within 7 days of your first visit
Multiple models. Five stages. One auditable verdict.
Send a claim to the API
Lenz frames it as a precise, falsifiable statement — removing ambiguity before research begins.
Lenz searches, then argues both sides
Multiple queries across authoritative sources. Two models argue opposing sides from the evidence — disagreement is preserved, not averaged away.
Structured output — block, publish, or escalate
A score, sourced citations, and a full reasoning trace. Machine-readable so your workflow branches automatically.
Developer tier — API surface
| Endpoint | What it does | Quota |
|---|---|---|
/extract |
Pull verifiable claims from any text | 1,000 / day |
/assess |
Inline guardrail before output reaches users | 5,000 / mo |
/verify |
Full multi-model pipeline — framing to sourced verdict | 500 / mo |
/ask |
Ask a factual question, get a sourced answer | 5,000 / mo |
Where teams are integrating Lenz
AI content before publishing
Newsletters, blog posts, product updates — claims verified before they reach subscribers or go live.
Data enrichment & CRM
LLMs routinely invent funding rounds, headcounts, and tech stacks. Verify enriched records before they enter your CRM.
Regulated claims
Health, finance, and supplement copy where the audit trail is the deliverable — not an optional extra. Highest stakes, full trace.
Agent guardrail
Wire /assess as a ~10s pre-action check before your agent acts on a factual claim. Full verify only when the stakes warrant it.
Customer support AI
Catch a wrong answer before it reaches the customer — verified at the point of response, not after the complaint arrives.
RAG response gating
Groundedness checks verify retrieval faithfulness — not whether the source itself is wrong. Lenz checks the claim against the open web.
of real-world fact-checks return a dissenting verdict across five frontier LLMs. The choice of model is an invisible editorial decision — Lenz returns a verdict grounded in independent sources, not model consensus.
Read the research →Developer access, free for a month.
One month of Developer access. It applies the moment you verify your email.