Brand Narrative vs sentiment score: which one explains what AI says about you?
A head-to-head of a single sentiment score, a manual claim log and Promptwatch's Brand Narrative and Fact Check, with a verdict on when each is enough.
Quick verdict
A sentiment score tells you that something is off. It cannot tell you what. If an AI engine keeps describing your product as "too expensive for small teams," a score of 62 out of 100 hides that sentence completely. Brand Narrative, which Promptwatch shipped on 30 October 2026 (changelog), works at the level of the individual claim, and that is the level where you can actually do something.
Use a plain sentiment score when you only need a trend line for a board slide. Use claim-level analysis when someone has to fix a page, a pricing explanation or a review profile. For most teams that care about what buyers read, the second is the one that changes work.
The problem with one number
Imagine two brands with the same average sentiment. In the first, AI answers are mildly positive about everything. In the second, they are glowing about features and quietly damaging about price and support. The score is identical. The marketing response is not.
Sentiment as a single figure also mixes engines, prompts and topics. A bad week in one engine can be averaged away by a good week in another. And it says nothing about whether the model is wrong. A negative claim that is true is a product problem. A negative claim that is false is a content problem, often with a fixable source.
What each approach actually gives you
A sentiment score
One number per response or per period, usually with a trend. Promptwatch also reports it: sentiment is scored per response from 1 to 100, with a weekly AI summary. That remains useful as an early warning. It answers "is the tone moving?" and nothing more.
A manual claim log
A spreadsheet where someone reads answers and writes down every claim about your brand: price, features, support, reputation. Add columns for the engine, the prompt and a count of how often it appears. This works without any tool, and for a 20-prompt panel it is a fine weekly habit. It breaks down when you have hundreds of responses a week, when several people tag claims differently, or when you want the same treatment for competitors.
Brand Narrative
According to the changelog entry, Brand Narrative pulls the individual claims AI makes about you and your competitors out of every tracked response, with examples like "affordable for small teams" or "its mobile app lacks features." Repeated claims are grouped, so you see how often each one appears and which models say it. Each claim is filed under a theme such as Pricing, Features, Support or Reputation and scored positive, neutral, negative or mixed. Per theme, that rolls up into a score from 0 to 100 you can compare against competitors. It also shows sentiment per competitor, which exposes the cases where AI speaks about a rival more warmly than about you.
It is on every plan.
Fact Check
The second half of that release is where the diagnosis happens. You mark a claim Accurate or Inaccurate, and Fact Check reads the pages crawled from your own site to find the ones that state the claim or its opposite. Each check lists the pages and quotes it found, plus a recommended next step. This is the part that answers "is our own website the reason?" Fact Check is on Business, Agency Scale and Enterprise. Each check uses agent credits, and the estimate is shown before it runs.
Side-by-side
| Sentiment score | Manual claim log | Brand Narrative | Fact Check | |
|---|---|---|---|---|
| Shows which claim is the problem | No | Yes | Yes | Yes |
| Groups repeated claims | No | By hand | Yes | Uses a chosen claim |
| Theme scores against competitors | No | By hand | Yes, 0 to 100 per theme | No |
| Finds pages on your site that state the claim | No | By hand | No | Yes, with quotes |
| Effort as volume grows | Low | High | Low | Low, credit cost |
| Promptwatch plan | Not plan-gated in the fact sheet | Any | Every plan | Business, Agency Scale, Enterprise |
Source for the Promptwatch columns: Promptwatch changelog, read 2 November 2026.
What other tools do here
We looked at our own catalog rather than guess. Qwairy's entry notes a Fact Check feature that is still in beta. The AthenaHQ, Scrunch AI and Ahrefs Brand Radar entries do not record claim-level extraction, so we cannot say whether they offer it. Treat that as unknown, not as absent, and ask the vendor for a demo on a real prompt set before you decide. For a wider look at the category, see our buyer scorecard.
Who should pick which
- A sentiment score alone if you report tone to executives once a quarter and nobody is expected to act on it.
- The manual log if you track a small prompt set, work alone, and want to learn what claims look like before paying for anything.
- Brand Narrative if your questions sound like "why do we lose on price in AI answers?" or "which competitor gets the better reputation score?" It is the cheapest way in, because it is on every plan.
- Brand Narrative plus Fact Check if you suspect your own site contradicts itself, for example an old pricing page next to a new one. The Business plan is $579 a month, which is the honest cost of that second step.
A workable routine
Start with the claims, not the score. Each week, open the theme with the lowest score, read the most repeated negative claim and decide whether it is true. If it is true, route it to product or pricing. If it is false, mark it Inaccurate, run Fact Check if you have access, and fix the page it points to. If you do not have access, search your own site for the claim and its opposite, and check your oldest pages first.
Then fold it into your wider fact-checking process. Our guide to how to fact-check what AI says about your brand covers the source-tracing steps, and sentiment monitoring tools are worth reading if tone trends still matter to you.
One caution. Claims are extracted from tracked responses, so the output is only as good as your prompt set. If your prompts never ask about pricing, a Pricing theme will be thin, and a thin theme looks like good news. Add buyer-style questions on price, support and alternatives before you read anything into the scores.
Bottom line
The score is a smoke alarm and the narrative is the floor plan. If you only get one, get the one that tells you which room is on fire. Promptwatch is the one we would evaluate first here, because claim-level analysis arrives on every plan and the site check sits one tier up, so you can start with the claims and add the site check once they justify it.