Direct answer
Compare sentiment only within the same prompt, provider, model, market, and time scope. Pair each aggregate with its eligible mention count and read the answers. Detected AI sentiment is not customer sentiment or factual accuracy.Data required
- Completed answers with SELF and competitor mentions
- Correct entity aliases and competitor relationships
- Detected sentiment values and eligible mention counts
- Fixed provider, model, country, language, topic, tag, funnel, and period filters
- Full answer text for contextual review
Workflow
- In Analyze > Sentiment, set a fixed analysis scope and inspect the SELF distribution.
- In Cockpit > Compare, select the same scope and compare competitor sentiment.
- Record each value with its eligible mention count.
- Separate topics, prompts, or providers that contribute most to the difference.
- Read positive, negative, mixed, and neutral cases in Monitor > Response Analysis.
- Verify whether the language describes the brand, a product, an incident, or a quoted source.
Interpretation
Sentiment summarizes detected language around eligible mentions. A difference can reflect prompt mix, small samples, or one repeated narrative. It does not show what users believe. Pair sentiment with Mention Rate or Coverage. A favorable sentiment value from a small set of SELF mentions can coexist with broad brand absence. Do not rank brands without showing eligible counts.Possible next actions
- Test whether a negative pattern is concentrated in one factual claim or topic.
- Correct inaccurate public information with a clear authoritative source.
- Investigate provider-specific wording before changing positioning.
- Track a matched answer cohort after a corrective update.
Limitations
- Automated sentiment can miss nuance, conditional language, irony, and mixed evaluations.
- Competitors may have different eligible mention volumes.
- Answer variability can move small samples sharply.
- Sentiment does not measure recommendation, citation, reputation, or revenue.