Skip to main content
POST
The inaccurate claims behind your accuracy score — the false statements AI answers made about your category, from the platform’s FactCheck tab. Returns a flat, paginated list, or group_by one dimension to section them. A claim is { cluster_id, claim, occurrence } plus whatever include adds.
  • occurrence: the % of scoped responses that carry the claim.
  • include: any of theme, reasoning, models, evidence, citation_sources.
    • models[] — which models make the claim: { id, name, occurrence }.
    • evidence[] — refuting knowledge-base snippets: { id, kb_path, kb_snippet, source_updated_at }.
    • citation_sources[] — the cited pages driving the claim: { href, hostname, citation_category, domain_category, snippet, citation_share }. citation_share is a percentage, model-balanced, matching the Citations tab.
  • group_by: 0 or 1 of model, region, persona, prompt, topic, tag, theme (no citation or date — those are scores dimensions). Empty → one flat paginated list; a dimension → one section per value: { <dim>: { id, name }, accuracy, accurate, inaccurate, total_claims, claims: […] }, each carrying all its claims.
  • filter: the same constrained tree as scores — a top-level and of single-field leaves over model/topic/region/persona/prompt/tag (only topic negatable).
A few guardrails return 422 (with a message): citation_sources isn’t available with group_by (its share is scope-sensitive) — request it on the flat list; group_by: ["tag"] can’t be combined with a tag filter (tags are multi-valued, so co-tag sections would escape the filter); and sectioning by model over an unusually large scope (>~2000 inaccurate clusters) — narrow the range or filter.
POST /v2/reports/factcheck/claims/stream takes the same request body and returns Server-Sent Events: one summary event (the info block), then one result event per claim. limit/cursor are ignored; it returns everything by default. Pass max_results to cap.The stream is the flat list only — group_by and the per-claim evidence/citation_sources lookups return 422; use the non-stream endpoint for sectioned or enriched claims.
Response (text/event-stream)

Authorizations

​
X-API-Key
string
header
required

Body

application/json

The inaccurate claims — flat + paginated, or group_by one dim to section them (all claims per section).

​
category_id
string
required
​
start_date
string
required

YYYY-MM-DD, ET, inclusive

​
end_date
string
required

YYYY-MM-DD, ET, inclusive

​
group_by
enum<string>[]

Optional single dim to section the claims (e.g. per model). Empty → one flat claim list.

Maximum array length: 1
Available options:
model,
region,
persona,
prompt,
topic,
tag,
theme
​
filter
FilterNode · object | null

Scope which responses count (see Filtering).

​
include
enum<string>[] | null

Optional per-claim detail fields to add to each claim (see options).

Available options:
theme,
reasoning,
models,
evidence,
citation_sources
​
limit
integer | null

Claims (or sections) per page; default 25.

Required range: 0 < x <= 100
​
max_results
integer | null

Stream only: cap entries returned.

Required range: x > 0
​
cursor
string | null

Response

Successful Response

The response is of type Response Query Claims V2 Reports Factcheck Claims Post · object.