AI Fact Check API: Accuracy of AI Answers | Genezio

The fact check reports of the Genezio API measure how often answer engines give the correct facts about your brand. A topic of the type fact_checker asks the engines about facts of your brand, such as a price, a feature or a policy. Each scenario has the expected answer, thus each answer passes or fails.

Report	Route
Over time	`GET /customer/v1/brands/{brand_id}/reports/fact-check`
By topic	`GET /customer/v1/brands/{brand_id}/reports/fact-check/topics`
By engine	`GET /customer/v1/brands/{brand_id}/reports/fact-check/models`

A row of the fact check report

{
  "date": "2026-09-28",
  "passed": 41,
  "failed": 9,
  "total": 50,
  "pass_rate": 0.82,
  "confidence_interval": { "lower": 0.72, "upper": 0.89 },
  "change": 0.11,
  "significant": true
}

Field	Meaning
`pass_rate`	From 0 to 1: `passed` divided by `total`. `null` when there is no answer.
`confidence_interval`	The Wilson interval of `pass_rate`, at 90%.
`change`	Over time only: the rate minus the rate of the previous period.
`significant`	Over time only: `true` when the intervals of the two periods do not overlap.

Find the facts to correct

The report by topic gives the scenarios of each topic with the lowest rate first, and the report by engine gives the engine with the lowest rate first. Thus, the first row tells you which fact, or which engine, to correct first.

Filters of the fact check reports

start_date, end_date, models and topic_ids, as in the other reports. The report over time also takes granularity: day, week or month.