Fast fact-check
assess_claimRead-only
The quick check: a verdict and a bucketed confidence for each factual claim in a text or a list (3-model panel, about 15-20 seconds, one credit per claim). Use it when the user asks to fact-check or double-check a factual statement or a text (“fact-check this”, “double-check that”), asks whether one is true or accurate (“is that true?”, “is this accurate?”), or doubts a factual statement you made (“are you sure?”). Not for opinions, predictions, arithmetic or how code behaves. When the user doubts something you said, pass the specific statement being doubted, not the whole conversation.
For one claim or a whole text (a draft, a pasted text or your previous
answer), pass ONE text in ``claim`` (every claim found gets a row, up to
20) or up to 20 claims in ``claims`` (one row per item, same order). A
pasted text goes in ``claim`` whole and unedited — never split it, reword
it, resolve its pronouns or add figures. ``claims`` is only for claims the
user listed separately.
Verdicts are True / Mostly True / Mixed / Mostly False / False. No sources:
present each verdict as a first read, not as final. A row may carry
``rationale``, a reviewer's reasoning for the verdict, and ``dissent``, the
reasoning of the reviewer farthest from it: reviewers' notes, not checked
sources. A low-confidence row carries ``recommend_verify: true`` and a
``next_step``: recommend `verify_claim`, the deep check, and ask before
running it. On medium offer it; on high mention it. A vague claim is
assessed on its most likely reading (the row's ``claim``). Leave
``language`` unset.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| claim | No | ONE statement to fact-check, in natural language (e.g. "Honey never spoils"). If the text contains several atomic claims, each is verdicted separately (up to 20, one credit per verdict). Leave empty when passing `claims`. | |
| claims | No | A list of up to 20 claims the user listed separately — one verdict per item, in the same order. Each item is one claim, not a document; a pasted text goes in `claim` instead. A compound item is judged on its main claim and the rest is listed in that row's `identified_claims`. Mutually exclusive with `claim`. | |
| language | No | Leave unset. English is the default. Set this only if the user explicitly asked for the answer in another language — not to match the language of the claim, the conversation, or the user locale. Supported: en, es, de, fr, it, pt, nl, sv, da, no, fi, bg. |
Output Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||