About FactCheck

Overview

FactCheck is Profound's accuracy layer. It measures how AI engines describe your brand and pinpoints which sources are responsible when they get it wrong. It extracts the specific claims AI models make about your brand, checks each one against your own source of truth, and shows you which citation URLs fuel any inaccurate information. The result is a clear, actionable picture of where AI-generated misinformation comes from, so your content team knows exactly what to fix.

When AI gets your brand wrong

AI search engines like ChatGPT, Perplexity, and Gemini don't just surface links, they make direct claims about your brand. They describe your product features, pricing, policies, and service terms as facts. When those claims are wrong, your customers are misled before they ever reach your website. And in many cases, inaccuracies don't come from AI fabricating information. They come from a brand's own published content being out of date.

The problem is that monitoring this at scale is nearly impossible without specialized tooling. Brands may know anecdotally that AI sometimes gets things wrong, but they have no systematic way to measure how often it happens, which prompts are affected, or which sources are responsible. Without that information, there's no clear path to fixing it. Content teams can't update a page they don't know is the problem, and brand teams can't report on accuracy without data to back it up.

For brands where facts matter most, such as those with regulated products, complex pricing, or frequently updated policies, this gap is especially costly.

How FactCheck works

FactCheck brings together three things that were previously separate: claim extraction, source-of-truth verification, and citation attribution.

Profound pulls the claims AI engines make about your brand across all your FactCheck-tagged prompts, compares each claim against your Knowledge Base, and classifies it as accurate, inaccurate, or not relevant. Only factual, actionable, and atomic claims are verified: each claim must assert a single thing that your source of truth can confirm or contradict. Subjective statements are routed to Sentiment instead.

For every inaccurate claim, FactCheck traces the finding back to the specific citation URLs in the AI response that appear to drive it. The citation URL is what makes the data actionable: instead of knowing only that AI is saying something wrong, you know which page or document in your Knowledge Base is the likely source of the problem.

Once you've identified the inaccuracies, run a Profound Agent directly from the FactCheck view to draft the outreach, update your Knowledge Base, create content, or otherwise help answer engines discover the correct information about your brand.

Key features

Claim extraction and verification

FactCheck automatically extracts the specific claims AI engines make about your brand in response to your FactCheck-tagged prompts. Claims are filtered to those that are factual, actionable, and atomic, each asserting one verifiable thing. Subjective claims are routed to Sentiment instead.

Each claim is then checked against your Knowledge Base and classified as accurate, inaccurate, or not relevant. This gives you a structured, consistent view of where AI gets things right and where it doesn't, with no manual review of individual responses. 

Claim clustering by theme

Rather than presenting thousands of individual claims, FactCheck groups similar claims into themes. This makes the data much more digestible and helps you understand which topic areas have the most accuracy problems. Instead of sifting through raw claim lists, you see at a glance that, for example, claims about your cancellation policy are a bigger problem than claims about your pricing.

Citation-level accuracy breakdown

For every inaccurate claim, FactCheck shows you which citation URLs in the AI response are associated with that inaccuracy. This is the feature that turns FactCheck data into a direct action item for content teams. You know which page to update, not just that something is wrong.

Accuracy overview and metrics

The Accuracy Overview section shows what changed from last week to the current week (the time period is customizable): Accuracy Score and Share, top inaccurate claims, movement across themes, and a comparison across reporting periods. Click into any inaccurate claim to see why it's wrong, the ground truth from your Knowledge Base, and every location in the Knowledge Base where the correct information is defined.

User feedback mechanism

If a claim is marked inaccurate but is actually correct, report it directly in the FactCheck UI, and the claim is removed from your view. If a claim isn't something you want to track, mark it as not relevant. Profound uses this feedback to keep improving claim verification.

Run Agents directly from claim views

Take action directly from any inaccurate claim view: run an Agent to draft outreach to the citation source, update your own content, or create new content to address the inaccuracy.