Overview
Navigate to Analyze > Fact Check in the left navigation menu.
Fact Check mines every AI answer that mentions your brand for factual claims (pricing, features, funding, leadership, and more), then verifies each claim against your own Source of Truth. The result is a Brand Accuracy score that tells you, at a glance, how often AI models describe your brand correctly.
Fact Check is currently in private beta. Contact your account manager to request access.
How It Works
- Every AI answer that mentions your brand is mined for factual Claims.
- Each claim is matched against your Source of Truth.
- When a claim lines up with a known fact, it is auto-verified.
- Unmatched claims wait for a human verdict: Correct, Incorrect, or Outdated.
- Verdicts feed your Brand Accuracy score and its trend over time.
Activation
Fact Check runs only while it is active. Use the Active / Paused toggle in the page header.
- Manager or Owner role is required to change this setting.
- When active, each AI answer mentioning your brand consumes 1 credit (claim extraction plus auto-verification against your Source of Truth). Answers with no brand mention are skipped at no cost.
- Pausing stops all credit spend and claim extraction. Existing claims and verdicts are kept.
Brand Accuracy
The Overview tab leads with your Brand Accuracy score.
| Element | Description |
|---|
| Brand Accuracy | Share of verified claims that are accurate, 0 to 100%. Target is 80%. Color coded with a period-over-period trend. |
| Distribution | A donut split into Accurate, In review, and Inaccurate. |
| Accuracy evolution | A line chart tracking the score across recent weeks. |
| By category | Accuracy broken down by claim category. |
| By provider | Accuracy broken down by AI provider (ChatGPT, Claude, Perplexity, and others). |
| Bucket | Includes |
|---|
| Accurate | Correct and Auto-verified claims |
| In review | Unverified claims awaiting a verdict |
| Inaccurate | Incorrect and Outdated claims |
Claims
The Claims tab lists every extracted claim. Search by text, then filter and sort.
| Filter | Options |
|---|
| Status | Unverified, Correct, Incorrect, Outdated, Auto-verified, Ignored |
| Impact | High, Medium, Low |
| Category | Pricing, Limits & Quotas, Performance, Comparison, Integration, Features, Security, Availability, Company, Leadership, Funding & Valuation, Customers, Market position, Date, Location, Other |
| Source of Truth | New, Match, Drift, Recurring |
Claim statuses
| Status | Meaning |
|---|
| Unverified | Extracted, awaiting a verdict |
| Correct | Verified as accurate |
| Incorrect | Factually wrong |
| Outdated | Was true, no longer current |
| Auto-verified | Matched your Source of Truth automatically |
| Ignored | Excluded from the accuracy score |
Source of Truth
The Source of Truth tab holds your canonical facts. When a newly extracted claim matches an entry, it is verified automatically with no manual work. The more entries you maintain, the more future answers are checked for you.
Each claim is labeled by how it relates to your Source of Truth: New (not yet linked), Match (aligns with an entry), Drift (differs from a linked entry), or Recurring (seen across multiple answers).
Audit log
The Audit log tab is a chronological feed of extraction and verification activity, so you can trace how each verdict was reached.
Export
Use Export CSV in the header to download claims with their status, category, impact, and providers for reporting or external analysis.
Use Cases
| Use Case | Description |
|---|
| Catch hallucinations | Spot incorrect or outdated facts before prospects see them |
| Protect deals | Find and correct misinformation that could cost a sale |
| Compare providers | See which AI models describe your brand most accurately |
| Measure improvement | Track Brand Accuracy as you correct the record over time |
