What is Truth Check?
Truth Check fact-checks the claims AI engines make about your brand against your Context, scores their accuracy and drafts fixes for wrong ones.
Truth Check fact-checks what AI engines say about your brand. It pulls the factual claims engines make about you from their answers, compares each one with your Context, and shows which are right, which are wrong and how to fix the wrong ones. You'll find it in AI visibility, on the Truth Check page.
Your Context is the source of truth, so build it first: records, team answers, brand profile, documents and synced pages. With nothing in Context, every claim comes back as Can't verify.
Each claim gets one of three verdicts. Accurate means it matches your Context. Wrong means your Context contradicts it, and Truth Check shows what your sources actually say. Can't verify means your Context neither confirms nor contradicts it. Your accuracy score is accurate claims divided by accurate plus wrong, so claims your Context can't confirm don't count.
How to set up and use Truth Check
- Open AI visibility and choose Truth Check.
- On Setup, choose the prompts to fact-check. Pick prompts with one right answer, such as pricing, features, availability or policies. Starter can check up to 10 prompts, Growth up to 15 and Enterprise up to 50.
- Once a week, each chosen prompt is checked on each engine using its newest answer. Click Check now on Overview to check anything that's due straight away.
- On Overview, read your Accuracy score against the previous period, Weekly accuracy, Top wrong claims, and the breakdowns By theme and By engine. Switch the period between 7, 14, 28 and 90 days.
- Open Claims to filter by verdict, theme or status, or to search claims.
- Click a wrong claim to see what the AI says, what your sources say and where it comes from.
- Under Fix it, follow the recommended route: Fix your page, Publish the right answer, or Ask the publisher to correct it. Click Draft it, then copy the draft or choose Open in editor.
- Mark the claim as fixing while you work on it, or dismiss it.
Similar wrong claims are grouped, so one fact repeated across several engines is one thing to fix. Facts Truth Check can't verify become questions in Context, under Identified gaps. Answer once and future checks use it. If a claim marked wrong is true, choose Actually correct and tick Add to Context as a confirmed fact, so every AI feature treats it as confirmed. You can also mark a claim Not relevant or flag a Wrong verdict.
- Checking answers and drafting fixes use AI tokens from your plan.
- Truth Check can use only a capped share of your plan's monthly AI tokens. Once it reaches that share, checks pause until next month, though fix drafts you ask for still run.
- The Answer X-ray in Evidence shows a fact check for each answer.
Tip: start with prompts about pricing and features. Wrong facts there cost you the most, and they're usually the easiest to correct at the source.
Still stuck? We typically reply within 1 business day.
Contact support