On this page

AI Visibility Sentiment

Beta

This feature is in Beta. This feature may continue to evolve. This documentation may not yet reflect the latest updates.

Sentiment shows you how AI models describe your product across real user prompts. As part of AI Visibility, it measures the tone of AI-generated responses. You can track not just whether your product appears in AI answers, but how it's perceived.

How sentiment works

Amplitude computes sentiment at the response level, then aggregates it across all prompts. Amplitude classifies each AI-generated response into one of four categories:

  • Positive: The response frames your product favorably.
  • Negative: The response frames your product unfavorably.
  • Mixed: The response contains both positive and negative framing.
  • Neutral or failed: The response has no discernible sentiment, or the sentiment analysis failed.

Calculations use only responses with valid sentiment. Amplitude excludes neutral and failed responses to keep percentages consistent and comparable across time and competitors. Mixed responses contribute to both the positive and negative totals, so results still reflect partially negative responses.

Amplitude extracts themes independently from sentiment classification, and they can appear across multiple sentiment categories.

How Amplitude calculates the sentiment score

Amplitude bases sentiment percentages on valid responses only. This keeps totals consistent and comparable.

Example:

CategoryCountPercentage
Total responses100
Positive4040 / 80 = 50%
Negative1010 / 80 = 12.5%
Neutral30excluded
Failed20excluded

Valid responses: 80 (total minus neutral and failed).

The overall sentiment score is the percentage of positive responses. Amplitude uses this score to compare perception across your product and competitors.

The UI displays only positive and negative as top-level metrics. It doesn't show mixed sentiment as its own category.

Explore sentiment

  1. Open AI Visibility.
  2. Select Sentiment.
  3. Review the overall sentiment score and trends.
  4. Analyze the top themes associated with each sentiment.
  5. Drill into individual responses for specific examples.

Act on sentiment

Use sentiment insights to improve how AI models describe your product:

  • Reinforce positive themes: Create dedicated content and messaging around themes that are already working.
  • Address negative themes: Publish targeted pages or clarifications for recurring concerns. See AI Visibility recommendations for specific actions Amplitude suggests.
  • Align positioning: Adjust product marketing to match how AI interprets your product.
  • Track over time: Monitor sentiment shifts as you make content and messaging changes.

Limitations

Sentiment reflects how AI models describe your product, not direct user feedback. Keep the following in mind:

  • Amplitude excludes neutral and failed responses from calculations to avoid distorting percentages.
  • Mixed responses contribute to both positive and negative sentiment, so they still influence overall results.
  • Themes may appear across multiple sentiment categories depending on context.
  • Amplitude doesn't automatically generate negative prompts (for example, "worst tools" or "alternatives to avoid"). Add them manually to your prompt set if you want to analyze negative framing without introducing bias.

To learn how to send AI Visibility data, go to Send AI Visibility data to Amplitude.

Was this helpful?