Skip to main content
Yes. Prefer scores every mention of your brand in AI answers as positive, neutral or negative, and rolls them into a Sentiment mix per engine and against rivals. It measures how the AI describes you, not what customers or reviewers feel, and links each negative theme to its sources. Last updated 3 October 2026 Yes. Prefer reads the tone of every mention of your brand in AI answers and labels it positive, neutral or negative. Prefer then rolls those labels into a mix you can track over runs, split by engine and set against each competitor.

​
What Prefer’s sentiment measures, and what it does not

Prefer’s sentiment measures how AI engines describe your brand in their answers, not how customers feel about you. The Sentiment docs put it plainly: it is not a measurement of user opinion, review scores or social posts. It is how the model summarizes you. That is a narrower read than social listening, and it is the one that matters for AI search. When a buyer asks ChatGPT or Perplexity about your category, the summary they read is the thing being scored. A brand can have strong reviews and still be described with a caveat that an engine picked up from 1 old forum thread.

​
What the Sentiment tile shows

The Sentiment view in Prefer has 6 parts, from the overall mix down to the URLs behind a negative theme.
  1. Mix strip: the share of positive, neutral and negative mentions in scope.
  2. Positive trend: how the positive share moves run over run.
  3. Drivers: theme quotes pulled word for word from engine answers, grouped by topic.
  4. Competitor benchmark: your mix next to each tracked rival’s.
  5. Per-engine mix: the split by engine, to spot models that describe you differently.
  6. Negative-source drilldown: the cited URLs driving negative mentions.
You also see sentiment on each prompt’s detail page across runs, and as a column in the Competitors Compare matrix.

​
How to act on a negative theme

A rising negative mix on 1 engine usually traces back to 1 or 2 cited sources, so check the sources before you rewrite anything. The docs lay out the routine:
  • Open Monitor, then Sentiment, or scroll to the Sentiment tile on the Overview.
  • Click a negative driver to read the full engine response that contains it.
  • Open the negative-source panel to see which cited URLs shape that theme.
  • Create an Action for it in the Action Center: a correction request, an outreach, an updated owned page or a support case.
Verify the fix the same way: after the source changes, watch the per-engine mix and the driver quotes over the next few runs. Prefer labels every change vs the previous run, so you can tell whether a shift held. For background on the metric, read our brand sentiment definition, and see how to correct what ChatGPT says about your brand. The Answer Engine Insights module is where the Sentiment view lives. To see the sentiment themes AI uses about you today, run a free AI visibility audit.

​
Frequently asked questions

No. Prefer’s sentiment is scored only from the language AI engines use in their answers. It does not read social posts, review scores or customer surveys. Treat it as how the model summarizes you, which is what a buyer reads when they ask the engine.
Yes. Prefer splits the Sentiment mix by engine, so you can see if Perplexity describes you more harshly than Gemini. The docs note that a rising negative mix on one engine while others stay flat usually traces back to one or two cited sources.
Yes. Prefer’s Sentiment tile includes a competitor benchmark that places your mix next to every tracked rival’s, and the Competitors console Compare tab shows Sentiment as one of its 4 columns.

​
Also asked as

  • Can Prefer tell me if ChatGPT says negative things about my brand?
  • How does Prefer score sentiment?
  • Does Prefer do AI sentiment analysis?

​
Sources