AI Hallucination Firewall
The AI Hallucination Firewall sends brand-related prompts to large language models, compares the responses against brand facts you provide, and classifies each claim as accurate, inaccurate, hallucinated, missing, or outdated — producing a single 0-100 Brand Accuracy Score.
What it does
You provide a set of brand facts — what your product does, pricing, founding details, policies, anything you consider ground truth. CrawlPod periodically sends prompts a real customer might ask an AI assistant about your brand, then compares each claim in the response against your facts.
Every claim is classified individually and rolled up into a Brand Accuracy Score from 0 to 100, so accuracy is trackable as a single number over time rather than re-read from raw AI output on every check.
Why it matters
Industry studies have found AI-generated brand descriptions are frequently inaccurate, and a customer asking ChatGPT or a similar assistant about your brand has no easy way to know whether the answer is correct. An inaccurate or hallucinated claim can shape a purchase decision just as much as a true one, and unlike a wrong statement on a website you control, you can't simply edit an AI model's output — the first step is knowing it's happening.
How it works
- Enter your brand facts in the CrawlPod dashboard.
- CrawlPod sends a set of brand-related prompts to supported AI models.
- Each response is parsed into individual claims.
- Every claim is classified against your brand facts.
- Review the Brand Accuracy Score and the specific claims behind it.
Key concepts
| Classification | Meaning |
|---|---|
| Accurate | The AI's claim matches a brand fact you provided. |
| Inaccurate | The AI's claim contradicts a brand fact you provided. |
| Hallucinated | The AI stated something with no basis in the facts you provided or in public information. |
| Missing | The AI omitted a fact you'd expect it to mention for the prompt. |
| Outdated | The AI's claim was once true but no longer matches your current brand facts. |
Limitations & fair usage
The AI Hallucination Firewall is available on the Business plan and above. Checks run on a fair-use schedule to keep response times reasonable for all customers — accuracy is a point-in-time read of how a given AI model responded when checked, not a continuous live feed.
Related features
- Competitor Poisoning Alerts — monitors what competitors publish that could shape how AI models describe your brand.
- CrawlPod Verified Badge — a public, checkable trust signal that complements accuracy monitoring.
- See the Business plan for what else is included at this tier.