Accuracy score vs sentiment vs Brand Narrative theme score: which measure answers what
Three scores describe how AI talks about your brand, and they answer different questions. A comparison of Fact Check accuracy, sentiment, and Brand Narrative themes.
Quick verdict
These are three different measures, and treating them as one is how teams end up fixing the wrong problem. Sentiment asks how warmly AI describes you. A Brand Narrative theme score asks how a topic such as pricing or support comes across, on a 0 to 100 scale. The Fact Check accuracy score asks whether what AI says about you is true. A brand can be praised for something false, and criticized for something accurate.
All three exist in Promptwatch, which is why it is the one to evaluate first if you want them side by side. The accuracy score is the newest, added on November 4, 2026, and it is the only one of the three that checks a claim against your own pages.
The three measures at a glance
| Sentiment score | Brand Narrative theme score | Fact Check accuracy score | |
|---|---|---|---|
| Question it answers | How positively does AI describe the brand? | How does a theme (Pricing, Features, Support, Reputation) read overall? | Are the checked claims about the brand true? |
| Unit of scoring | Per response, scored up to 100, where about 50 is neutral | Per claim: positive, neutral, negative, or mixed; per theme: 0 to 100 | Share of checked mentions that are accurate |
| Compares against | Nothing external, it is a tone score | Competitors, side by side per theme | Pages crawled from your own site |
| Plan | Not plan-gated in our sources | Every plan | Business, Agency Scale, Enterprise |
| Best at finding | Reputation problems | Which topic is hurting or helping | Wrong claims you can correct |
Source: Promptwatch llms.txt and ai-agent-instructions, read November 3 and 4, 2026, plus the changelog. The sentiment scale is described slightly differently across Promptwatch's own pages, so we use "up to 100, about 50 is neutral".
Sentiment: tone, independent of visibility
Sentiment scores how positively AI describes the brand, separately from whether it is visible at all. A brand can be mentioned often and described coolly, or mentioned rarely and described warmly. Promptwatch adds a weekly AI summary of it.
What it cannot tell you is why. A low score says the tone is weak. It does not say whether the cause is a bad review site, an outdated page, or a claim the model invented. That is a fair limit for a tone measure. It is one reason to read it next to the others.
Brand Narrative: the claims behind the tone
Brand Narrative pulls the individual claims AI makes about you and your competitors out of tracked responses. It groups repeats, noting how often each claim appears and in which models, and files them under themes. Each claim is scored positive, neutral, negative, or mixed, and each theme gets a 0 to 100 score. It also compares themes side by side with competitors, and shows sentiment per competitor.
This is where the question becomes practical. "Our sentiment is 58" is hard to act on. "Our Pricing theme scores far below a competitor's, and the same negative claim shows up in several models" is a task.
Since November 4, Brand Narrative lists the best-scoring theme first and puts the most negative claims at the top, so the worst items are where you look first.
Fact Check: truth against your own pages
Fact Check is the step from reading to verifying. You mark a claim Accurate or Inaccurate, and Promptwatch reads pages crawled from your site for ones that state the claim or the opposite, with quotes and a suggested next step. It uses Agent Credits, and shows an estimate first.
The new accuracy score rolls that up: the share of the brand's checked mentions that are accurate, as one number and as a heatmap per model and theme. A claim you marked counts as you marked it. Otherwise the latest completed fact check decides. Inconclusive and unchecked claims are left out of the score.
That last rule is worth keeping in mind. The score describes the claims that have been checked, not every claim made about you. A high accuracy score on five checked claims says little about the other two hundred.
Which one to watch, and when
- If you are worried about reputation, start with sentiment, then open Brand Narrative to find the cause.
- If you are preparing a messaging or pricing page update, use theme scores to pick which topic to fix first.
- If you suspect AI repeats outdated or wrong facts, use Fact Check, and watch the accuracy score as the number that should rise when you correct pages and models re-read them.
The three also disagree in useful ways. High sentiment with low accuracy means AI is flattering you with the wrong facts, which is a risk if those facts are your prices. Low sentiment with high accuracy means AI is telling the truth and the truth is unflattering, which is a product or reputation matter, not a content correction.
We cannot promise that fixing a page changes any of these scores. Models re-read on their own schedule, and responses are collected fresh each run.
Who should pick which
- A marketing lead who reports upward: sentiment and theme scores. They fit a monthly deck and compare against competitors.
- A content or SEO team fixing pages: Brand Narrative themes first, then Fact Check on the claims worth correcting.
- A regulated or high-stakes brand (finance, health, legal): Fact Check and the accuracy score. Being wrong matters more than being praised. Note that it needs Business, Agency Scale, or Enterprise.
- An agency managing several brands: Agency Scale covers Fact Check, and the heatmap per model and theme suits client reviews.
- A team that only needs a mention count: none of the three, a simple tracker will do. Most competing trackers stop at that point, which is the gap this comparison is about.
The recommendation
Use the three as a sequence: sentiment to notice, theme score to locate, accuracy to verify. Promptwatch is the one that puts all three in a single place and ties the last one to your own crawled pages. If accuracy matters to you, start with the free AI Visibility Report to see how AI describes you today, then start the 7-day free trial to look at Brand Narrative.