Question 1

What is Brand FactCheck?

Accepted Answer

It is a per-window accuracy check on what AI answers say about your brand. Citlyze extracts the factual claims in tracked answers, verifies each one against a knowledge base of facts you approved, and turns the results into an Accuracy Score, claim themes, and the sources behind wrong claims.

Question 2

What counts as a factual claim?

Accepted Answer

A single verifiable statement: a price, a founding year, a policy, a product spec. Each claim is atomic, one checkable fact, so a verdict is always about one thing. Subjective statements like tone and sentiment are not scored here; those belong to brand perception tracking.

Question 3

What do accurate, inaccurate, and unverifiable mean?

Accepted Answer

Accurate means the claim matches an approved fact. Inaccurate means it contradicts one, and the verdict carries an error class: hallucination for invented details, outdated for facts that used to be true, incorrect for the rest. Unverifiable means your knowledge base has no covering fact, which is a coverage gap, not an error.

Question 4

How is the Accuracy Score calculated?

Accepted Answer

Accurate claims divided by accurate plus inaccurate claims, per measurement window and per engine. Unverifiable claims are excluded so gaps in your knowledge base never count against you, and low-confidence verdicts wait in a review queue and are not scored until you confirm them.

Question 5

How do I build the knowledge base?

Accepted Answer

Three ways: a guided crawl extracts facts from your own pages, starting from the ones AI answers cite most; document upload runs the same extraction on a PDF or text file; and a manual editor covers facts that live nowhere else. Facts carry categories and effective dates, and a coverage meter shows which categories are still thin.

Question 6

Why do proposed facts need my approval?

Accepted Answer

Because every verdict is only as trustworthy as the fact it was checked against. Extracted statements are proposals until you approve them, nothing is verified against an unapproved fact, and rejected statements are never proposed again. Facts untouched for a long time get flagged for a freshness review.

Question 7

How do I find which sources spread a wrong claim?

Accepted Answer

Inaccurate claims cluster into themes, and each theme lists the sources cited in the answers that repeated it, ranked by citation frequency. That points the correction where it works: the page a wrong price actually comes from, rather than the engine that only repeated it.

Question 8

What can I do about inaccurate claims?

Accepted Answer

Two remediation paths. When the offending source is your own site, the fix flows into recommendations with the claims and answers as evidence. When it is a third-party site, Citlyze builds an outreach pack per source: the wrong claims, where AI repeats them, and the correct facts with your sources, ready to copy or export.

Question 9

Can I get alerted when accuracy drops?

Accepted Answer

Yes. Alert rules fire on a new inaccurate theme or an Accuracy Score drop past your threshold, by email or through Slack and Teams webhooks, and scheduled reports include an accuracy section for the weekly digest.

Question 10

Which plans include Brand FactCheck?

Accepted Answer

Brand FactCheck is included from the Pro plan up. Each plan carries a cap on verified claims per measurement window, and a capacity add-on raises it when your prompt set produces more claims than the cap covers.

What is AI getting wrong about your brand?

A source of truth you approve, fact by fact

Every claim gets a verdict

Find the source, fix the record

Brand FactCheck questions

Keep exploring

Does AI recommend you when buyers ask?