Turn an AI-generated report into a claim ledger, retrieve independent evidence, and record a reviewable verdict for each material claim.
Before using an AI report to choose a vendor or plan a project.
A source-backed research audit that separates supported, contradicted, and unverifiable claims.
A plausible citation can still point to the wrong author, date, or conclusion. Opening independent sources makes the result reviewable before it informs a decision.
Given a market report with 12 claims, retrieve current primary sources for the five material claims and return a verdict plus supporting URLs for each one.
Illustrative format only, not a completed run. Replace the placeholders with verified results.
Claim: [exact statement] Verdict: [supported / contradicted / unresolved] Evidence: [primary URL, date, relevant passage] Decision impact: [what changes or remains unknown]
Search results are leads, not proof. Open the cited pages and retain unresolved claims when the underlying source cannot be checked. This is not a professional audit.
Review access and pricing for each App below. Connect an account where required, then return to this guide and give the example task and your inputs to your Loadout-enabled agent. Setup alone does not run the task.
No dedicated Skill is linked to this guide yet. You can follow the steps with your agent using the listed Actions.
These optional examples share at least one Action with this guide. They are not required Skills or verified implementations of the complete use case.
Mapped to the current public Loadout catalog
Demand and workflow source
How to Audit AI Search Citations Across Kimi, Doubao, and DeepSeek