AI Answer Citation Audit Worksheet
Trace an AI answer’s product claims to its linked sources and distinguish accurate support, weak attribution and unsupported statements.
When to use this template
Use this when an AI answer names your app but you need to know whether its factual description, access conditions and supporting links are accurate.
Choose another approach when: You want to label any unfavorable answer as misinformation, force a recommendation through hidden instructions, or assume that every listed source supports every sentence.
Gather these inputs
- The exact public answer and its visible linked citations.
- Current authoritative product facts and public URLs.
- A clear distinction between factual claims and subjective judgments.
Copy and adapt the template
Edit the text below and copy it into your own document. Edits stay in this browser field and are not submitted or saved by this page.
Worked example
Illustrative example, not a Crany customer result. Replace every example detail with facts you can verify.
An illustrative answer says a fictional app automatically sends handoff reminders. Its citation points to a page describing a planned reminder feature. The audit marks support as partial and outdated in context: the source discusses a plan, while the answer describes released behavior. The owner makes the public feature status clearer, including the limitation beside the description. The audit does not claim to know the model’s internal reasoning and does not treat a successful page update as proof that the external answer has refreshed.
How to use it
- Break the answer into factual claims and follow each relevant citation.
- Compare the cited passage with current authoritative facts and preserve the original observation.
- Improve ambiguous owned content where warranted, then record a separate retest.
Check before using it
- The cited page is inspected instead of judging by its title.
- Subjective preference is not mislabeled as a factual error.
- The correction stays within facts the product owner can substantiate.
What to measure
Track corrected owned-source ambiguities and the accuracy of later sampled answers. A source audit can identify a fixable explanation problem; it cannot prove which sources caused the answer or guarantee an external correction.
Research and sources
Reliable sourcing and accurate public content support the audit premise. The claim-level assessment is an original editorial technique, not a way to inspect a model’s private reasoning.
- Google: Creating helpful, reliable, people-first content — checked 2026-10-23. Original value, useful answers, transparent sourcing and a clear audience.
- Google: AI features and your website — checked 2026-10-23. Search eligibility, visible content, AI feature reporting and the absence of special AI markup requirements.
About Crany
Crany is an AI marketing co-founder in development for software founders and vibe coders, designed to strategize, execute approved marketing and improve based on results. This free resource is usable independently of the product. Join the public waitlist.