Skip to main content

​
Overview

Sentiment Score measures the overall tone of AI responses when your brand (or products) are mentioned, on a 0–100 scale. This metric tells you how AI models describe you: whether you show up as a trusted, recommended option, a neutral mention, or associated with concerns and negative language. In practice, most sentiment scores fall between 65 and 85, with higher scores indicating more positive language and associations.

​
How Sentiment is Calculated

​
Brand Sentiment (Brand Prompts)

For brand-level analytics, sentiment is calculated from the brand_mentions table in the analytics schema:
  • Each row in brand_mentions has a sentiment_score field (0–100) for that specific mention.
  • The Brand Sentiment Score is: 
    Sentiment Score = AVG(sentiment_score) across all brand mentions

​
Calculation details

  • Input: Individual sentiment_score values from brand_mentions.sentiment_score
  • Aggregation: Average (mean) of all sentiment scores for the brand in the selected period
  • Range:
  • 0–40: Negative sentiment (critical language, concerns, negative associations)
  • 40–70: Neutral sentiment (factual, low emotional tone)
  • 70–100: Positive sentiment (trusted, recommended, strong brand associations)
  • Filters Applied: Project, date range, AI models, topics/prompts, and optionally brand domain
In SQL terms (as used in our analytical views), this is: 
avg_sentiment = AVG(bm.sentiment_score)

​
Product Sentiment (Product Prompts)

For product-level analytics, sentiment is calculated from the product_mentions table:
  • Each row in product_mentions has a sentiment_score (0–100) for that product mention.
  • The Product Sentiment Score for a given product prompt is: 
    Product Prompt Sentiment = AVG(sentiment_score) across all product mentions for the prompt

​
Calculation details

  • Input: product_mentions.sentiment_score values filtered by prompt_id
  • Aggregation: Average (mean) of all sentiment scores for products mentioned when that prompt was executed
  • Range & interpretation: Same 0–100 scale and bands as brand sentiment
  • Filters Applied: Project, date range, AI models, product prompt IDs, and optionally specific product URLs
In SQL terms (as used in getProductPromptsAnalytics): 
AVG(pm.sentiment_score) as avg_sentiment

​
Display formats

  • Brand dashboard:
  • Sentiment appears in the brand metrics table alongside Visibility and Position.
  • Product prompts dashboard:
  • Sentiment is shown per product prompt, so you can see which prompts drive more positive or negative product language.
Null sentiment values (e.g. when tone cannot be reliably determined) are excluded from the averages.

​
Why Sentiment Matters

​
Brand Reputation in AI Answers

Sentiment directly reflects how your brand is framed in AI-generated content:
  • Positive sentiment:
  • Builds trust and credibility.
  • Often includes language like “trusted”, “reliable”, “recommended”, “leading”, “easy to use”.
  • Makes it more likely that users will consider and choose your brand.
  • Neutral sentiment:
  • Keeps your brand present in the conversation without strong emotional framing.
  • Typical for directory-style mentions or factual descriptions.
  • Negative sentiment:
  • Surfaces concerns, limitations, or criticism.
  • Can damage perception if left unaddressed, especially when it appears in authoritative sources or comparison content.

​
Competitive Perception

Because Orvi AI calculates sentiment per brand and per product:
  • You can compare how positively AI talks about you vs competitors.
  • You can identify competitors that:
  • Have higher visibility and better sentiment (strong threat).
  • Have good visibility but weak sentiment (opportunity for your brand to win trust).

​
Strategic Insights

Sentiment analysis helps you:
  • Detect perception issues early:
  • Sudden drops in sentiment can signal negative reviews, critical articles, or shifting narratives.
  • Understand positioning by topic:
  • Some topics may score highly (e.g. innovation) while others lag (e.g. support, pricing).
  • Track the impact of your initiatives:
  • Monitor whether PR, product improvements, and content updates shift sentiment over time.

​
Understanding Your Sentiment Data

​
In the Overview and Prompts Dashboards

On the Overview (brand) dashboard:
  • The Brand table shows:
  • Visibility (how often you’re mentioned),
  • Sentiment (how positively you’re described),
  • Position (how early you appear).
  • Use filters (models, date range, topics/prompts, brands) to isolate:
  • Specific AI models where sentiment differs,
  • Specific topics where you’re framed more positively or negatively.
On the Product Prompts dashboard:
  • Sentiment is calculated per prompt, across all product mentions generated by that prompt.
  • This lets you see which prompts:
  • Produce positive recommendation language,
  • Or surface more doubts and critiques about products.

​
Interpreting Ranges

Use these ranges as a practical guide:
  • 80–100 (Strongly positive) Your brand or products are consistently framed as trusted, leading, or highly recommended.
  • 65–79 (Generally positive / mixed) Overall good sentiment with some neutral language or mild caveats.
  • 40–64 (Neutral) Mostly factual descriptions; little emotional framing, limited advocacy.
  • 0–39 (Negative) Frequent mentions of drawbacks, concerns, or negative comparisons.
Look at sentiment together with visibility:
  • High visibility + high sentiment = ideal position.
  • High visibility + low sentiment = urgent reputation risk.
  • Low visibility + high sentiment = good perception when mentioned, but not often seen.
Sentiment works best in combination with other core GEO metrics: Together, these metrics answer:
  • Are we present? (Visibility)
  • How are we framed? (Sentiment)
  • Where do we rank vs others? (Position)

​
Improving Your Sentiment

To improve sentiment over time:
  1. Audit negative or neutral mentions
  • Use Recent Chats to read full responses where your brand appears with low sentiment.
  • Note recurring concerns (price, support, reliability, complexity, etc.).
  1. Align product and content with desired positioning
  • Address real product gaps where possible.
  • Update messaging, documentation, and landing pages to clearly answer common objections.
  1. Influence key sources
  • Identify high-impact sources from your Sources / Prompt Sources views where sentiment is weak.
  • Improve or replace those assets (new reviews, updated comparison pages, better how-to content).
  1. Monitor changes over time
  • Track sentiment trends after major launches, pricing changes, or PR campaigns.
  • Watch for divergence between models (e.g. one model lagging behind others in updated perception).
Sentiment is not just a vanity metric — it’s a live, model-level reflection of how AI explains your brand. Combining sentiment with visibility and sources gives you a precise map of where to improve both reputation and recommendations in AI search.
Last modified on January 25, 2026