Sentiment

Rankshift automatically analyzes the sentiment of AI responses to your prompts. Each response receives a sentiment label — positive, neutral, or negative — so you can instantly see how AI models are speaking about your brand or topic.

Written By Rankshift

How sentiment is scored?

We look at the context around the brand mention and assign a score between -10 and 10. Here is what each range means:

  • 8 to 10 — Exceptional, very strong positive sentiment

  • 4 to 7 — Generally positive sentiment

  • 1 to 3 — Mild or slightly positive sentiment

  • 0 — Neutral, factual, ambiguous, or not mentioned

  • -1 to -3 — Mild or slightly negative sentiment

  • -4 to -7 — Generally negative sentiment

  • -8 to -10 — Exceptional, very strong negative sentiment

Each score is based on the evaluation of the following parameters:

  1. Tone

  2. Emotional language

  3. Context around the term

  4. Implied sentiment

  • positive, negative, or neutral

  • e.g. joy, anger, sadness, praise, criticism

  • positive or negative events, outcomes, or associations

  • even if unstated directly

Sentiment chart on the dashboard

The dashboard includes a sentiment chart giving you an overview of all AI responses and their sentiment labels.

You can filter the chart by:

  • Tags

  • Personas

  • Models

💡Combine multiple filters for more targeted insights — for example, compare sentiment per LLM for a specific persona or tag.

Viewing sentiment per response

In addition to the overall chart, you can also view the sentiment of an individual response. Each response displays a label : positive, neutral, or negative. Based on the full content of the answer.

Here's how:

  1. Go to Monitoring > Visibility > Prompts in the sidebar

  2. Click on a prompt to view its responses

  3. Click on response

  4. Each response displays a sentiment label: positive, neutral, or negative

💡If a response is labeled negative, open it and check the reasoning to understand why the model takes that tone. This helps you understand not just what is negative, but why — so you can respond in a targeted way.