Skip to main content
Sentiment tells you whether an AI mention was framed positively or negatively. Accuracy monitoring is a different question entirely: was the AI’s answer even correct? An AI response can be perfectly polite and still be wrong — describing a feature you don’t have, a pricing model you’ve retired, or a positioning claim that contradicts what you’ve told Surfais is actually true. Accuracy monitoring exists to catch that.

​
How it works

Accuracy monitoring checks every AI mention of your brand against a short list of facts you’ve confirmed about yourself — your positioning claims. When an AI response says something that contradicts one of those confirmed claims, it’s flagged as an alert on this page: the prompt, the platform, the excerpt from the AI’s actual response, and the reasoning for why it was flagged.
Accuracy monitoring only runs once you’ve confirmed at least one positioning claim. Until then, this page shows “No claims to monitor yet” rather than silently checking nothing — go to Setting up positioning claims to add the facts you want AI answers held to.

​
Why it needs confirmed positioning claims first

Surfais doesn’t guess at what’s true about your brand — it checks AI answers against facts you’ve confirmed, not an automated best-guess. This is deliberate: a claim only becomes something Surfais checks against once you’ve explicitly reviewed and confirmed it. A handful of clear, specific claims — what you do, what you don’t do, your actual pricing model, a factual detail competitors sometimes get wrong about you — makes for a much more useful check than a long, vague list. See Setting up positioning claims for how to add and confirm claims.

​
Reading an accuracy alert

Each alert on this page shows the prompt that produced the mismatch, the platform and date, a short excerpt of the AI’s actual response, and a plain-language explanation of what it contradicts. The summary card at the top of the page shows how many open alerts you have and how many confirmed claims they’re being checked against, so you always know the scope of what’s being monitored. If an alert doesn’t hold up — the AI’s answer wasn’t actually wrong, or the mismatch isn’t worth tracking — you can dismiss it with an optional reason. A dismissed alert won’t resurface for that specific mention.
Repeat flags for the same prompt, platform, and country collapse into a single alert showing the most recent occurrence, rather than piling up a card per historical scan. A brand misrepresented in one country but described accurately in another still shows as two separate alerts, since each market is a genuinely different result.

Setting up positioning claims

Add and confirm the facts accuracy monitoring checks against.

Sentiment monitoring

The related but distinct check for tone rather than factual accuracy.

The Mentions feed

Read any mention in full context, accurate or not.

Brand profile

The wider set of details Surfais holds about your brand.