Hallucination Detection
Identify potentially inaccurate or outdated claims about your brand as they appear in AI-generated outputs. Hallucination Detection compares AI answers against your verified brand facts and highlights incorrect, uncertain, or outdated claims that need correction — available on the Professional plan and above.
Why this matters
Large language models generate answers by pattern-matching across training data and retrieved sources, not by looking up a canonical fact sheet for your brand. That means a model can confidently state a price you no longer charge, a feature you never shipped, or a founding date that's simply wrong — and repeat it consistently across sessions until the underlying sources it drew from are corrected. Left unaddressed, these claims shape how prospects and customers understand your brand before they ever reach your site.
Common categories of hallucinations we catch
- Outdated pricing or plan details: a model citing a discontinued tier, an old price point, or a feature that has since moved between plans.
- Feature claims: capabilities attributed to your product that don't exist, or real capabilities described inaccurately enough to set the wrong expectation.
- Company facts: founding year, headquarters, funding status, or leadership details that drift from what's actually true.
- Competitive mischaracterization: claims that misstate how you compare to a named competitor, in either direction.
What you can interact with
- Provider selection: review potential issues per AI platform.
- Brand scope: narrow findings to a specific brand/product line.
- Refresh: re-check after you publish corrections or clarifications.
What the data represents
- Potential inaccuracies: statements that may be wrong, outdated, or overly confident.
- Uncertain areas: topics where AI answers vary or lack clear support.
- Confidence signals: guidance for triage, not a guarantee of truth/falsehood.
How we get this data
We check how AI platforms describe your brand and highlight claims that appear inconsistent, outdated, or not well supported by common public information. For claims that need more scrutiny than a single-model answer can provide, findings can be escalated to the AI Council Chamber, which compares independent outputs from multiple models and uses anonymous cross-review to identify agreement, disagreement, and genuine hallucination risk before you spend time on a correction.
Recommended response workflow
- Confirm the claim internally (what is correct today?).
- Update your official pages to state the correct information clearly.
- Add an FAQ entry for common misconceptions.
- Re-check later to confirm the issue is decreasing over time.
Where this fits in your workflow
Hallucination findings feed into the same Reports area as your prompts, citations, and competitor data, so a correction cycle can be tracked and scheduled alongside the rest of your AI visibility monitoring instead of living as a one-off spreadsheet task.
