Interpreting Metrics

Overview

OneGlance tracks how AI providers mention your brand using multiple metrics. This guide explains what each metric means, how it’s calculated, and how to use it for strategic decisions.

All metrics are derived from actual AI provider responses to your prompts. OneGlance does not simulate or estimate—every score is grounded in real LLM output.

Core Metrics Explained

GEO Score (Generative Engine Optimization)

The GEO Score (0-100) measures your brand’s overall AI visibility. It’s a weighted average of four components:

// From packages/services/src/analysis/analysisPrompt.ts:262
overall = round(
  (visibility_value × 0.25) +
  (rank_value × 0.25) +
  (sentiment_value × 0.25) +
  (recommendation_value × 0.25)
)

Visibility (25%)

How prominently your brand appears in responses.Factors:

Coverage: How much text discusses your brand
Placement: Where your brand first appears
Structural prominence: Headings, lists, emphasis
Frequency: Number of mentions
Contextual framing: Role in the response

Rank (25%)

Your position in recommendation lists.Mapping:

#1 → 100 points
#2 → 80 points
#3 → 65 points
#4 → 50 points
#5 → 40 points
#6+ → 30 points
Mentioned but unranked → 15 points
Absent → 0 points

Sentiment (25%)

Tone of mentions (positive/neutral/negative).Scale:

81-100: Enthusiastic superlatives
60-80: Favorable with caveats
41-59: Neutral/factual
21-40: Significant drawbacks
0-20: Actively discouraged

Recommendation (25%)

Strength of endorsement.Types:

Top pick: 100 points
Strong alternative: 80 points
Conditional: 60 points
Mentioned only: 30 points
Discouraged: 10 points
Not mentioned: 0 points

Interpreting GEO Score Ranges

80-100: Dominant AI Visibility

What it means: Your brand is a go-to recommendation for this prompt topic.Typical characteristics:

Listed in top 3 absolute positions
Positive sentiment with superlatives (“best”, “excellent”, “standout”)
Recommended without major caveats
Multiple mentions throughout response

Example response excerpt:

“For sales pipeline management, HubSpot stands out as the top choice for mid-market teams. Its intuitive interface, robust integrations, and excellent customer support make it the clear leader in this space. HubSpot consistently ranks #1 for ease of use…”

Strategic actions:

Maintain current positioning (don’t relax)
Identify what content/sources AI providers cite
Expand to adjacent prompt categories
Monitor competitors for shifts

60-79: Strong AI Presence

What it means: Your brand is reliably mentioned and favorably positioned.Typical characteristics:

Appears in top 3-5 positions
Positive sentiment, often with conditions (“great for X use case”)
Recommended as a strong alternative
Discussed alongside category leaders

Example response excerpt:

“While Salesforce dominates enterprise CRM, Pipedrive is an excellent choice for small to mid-sized sales teams. It offers streamlined pipeline management at a fraction of the cost, though it lacks some advanced features.”

Strategic actions:

Push for #1 ranking: Address caveats mentioned by AI
Increase content volume on differentiators
Target prompts where you rank #4-5 to break into top 3

40-59: Moderate Visibility

What it means: Your brand appears but isn’t prominently recommended.Typical characteristics:

Ranked #5-8 or mentioned without ranking
Neutral sentiment (factual descriptions)
Listed among “other options” or “also consider”
Limited coverage (1-2 sentences)

Example response excerpt:

“Other CRM tools worth considering include Freshsales, Copper, and Close. Each has its strengths, but they’re generally less popular than HubSpot or Salesforce.”

Strategic actions:

Diagnose why you’re not higher: lack of content? competitor dominance? outdated info?
Publish authoritative content targeting this prompt topic
Get featured in high-authority review sites (G2, Capterra) that AI providers cite
Test prompt variations to see if wording affects ranking

20-39: Low Visibility

What it means: Your brand rarely appears or is positioned negatively.Typical characteristics:

Mentioned only in passing or as a negative example
Appears in “long tail” (rank #10+)
Contrastive mentions (“unlike [YourBrand], Competitor X…”)
Low sentiment or factual errors

Example response excerpt:

“Many teams have moved away from legacy tools like SugarCRM in favor of more modern alternatives such as HubSpot and Pipedrive.”

Strategic actions:

Investigate if AI has outdated information about your brand
Check for factual errors in AI responses (file corrections if possible)
Increase PR and content marketing to update AI training data
Consider rebranding or repositioning if perception is entrenched

0-19: Absent or Harmful

What it means: Your brand is missing from AI recommendations or actively criticized.Typical characteristics:

Not mentioned at all in responses
Appears only with warnings or discouragements
AI lacks awareness of your brand in this category
Competitors dominate all rankings

Example: Prompt “What are the best CRM tools?” returns no mention of your brand.Strategic actions:

Confirm your brand name and domain are correct in workspace settings
Audit if AI providers even know your brand exists (test with direct brand queries)
Launch aggressive content marketing and link-building campaigns
Get listed on major review platforms
Consider partnerships or PR to raise awareness

Detailed Metrics Breakdown

Presence Metrics

Found on the Dashboard and Prompts pages.

Presence Rate

Definition: Percentage of AI responses where your brand was mentioned. Calculation:

presenceRate = (responses_with_brand_mention / total_responses) × 100

Example: 10 prompts × 5 providers = 50 responses. Brand mentioned in 33 responses.

Presence Rate = (33 / 50) × 100 = 66%

Interpretation:

80-100%: Excellent category awareness
60-79%: Strong presence, some gaps
40-59%: Moderate presence, significant room for growth
20-39%: Low awareness, major visibility issues
0-19%: Critical awareness gap

If your presence rate is below 50%, start with brand awareness prompts rather than competitive prompts. Example: “What is [YourBrand]?” instead of “Compare [Competitor A] vs [Competitor B]”.

Mention Count

Definition: Number of times your brand is referenced in a single response. Counting rules (from packages/services/src/analysis/analysisPrompt.ts:48): Counts as a mention:

Exact brand name: “HubSpot”
Domain reference: “hubspot.com”
Sub-products: “HubSpot CRM”, “HubSpot Marketing Hub”
Well-known abbreviations: “SFDC” for Salesforce

Does NOT count:

Brand name only in URLs/citations (not prose)
Partial string matches (“Hub” ≠ “HubSpot”)
Brand mentioned in echoed user question (only counts in AI’s own answer)

Strategic use:

High mention count (4+): Brand is central to the response → good visibility
Low mention count (1-2): Brand is peripheral → increase content depth on this topic

Visibility Score (0-100)

A five-dimension calculation of how prominently your brand appears.

Coverage (25%)
Placement (25%)
Structural Prominence (20%)
Frequency (15%)
Contextual Framing (15%)

Question: How much text discusses your brand?Scoring:

0-5: Name-dropped in a word or fragment
6-15: One brief sentence
16-30: Short paragraph (2-3 sentences)
31-50: Multiple paragraphs
51-75: One of the primary subjects
76-100: Dominates the response

From packages/services/src/analysis/analysisPrompt.ts:190:

A. Coverage (25% weight) — How much space does the brand occupy?
Measure the proportion of the response's substantive content dedicated to {Brand}.

Question: Does your brand occupy high-attention positions?Scoring:

80-100: In a heading, title, or explicit “top pick” slot
60-79: Numbered/bulleted list, top 3 positions
40-59: Numbered/bulleted list, position 4+
20-39: Inline paragraph prose (no emphasis)
1-19: Parenthetical or footnote
0: Absent

Visual example:

## Best CRM Tools

1. **HubSpot** ← High structural prominence (heading + #1)
2. Salesforce
3. Pipedrive

Also consider Zoho, Freshsales, and Copper. ← Low structural prominence

Question: In what role does your brand appear?Scoring:

90-100: Direct answer to user’s question
70-89: Actively recommended or highlighted
50-69: Compared alongside peers
30-49: Mentioned as context or example
10-29: Contrastive/negative reference
1-9: Only in a question (not substantive content)
0: Absent

Example: Direct answer (90-100):

User: "What's the best CRM for startups?"
AI: "HubSpot is the best CRM for startups due to..."

Example: Contrastive reference (10-29):

User: "What makes Pipedrive better than competitors?"
AI: "Unlike HubSpot, which can be complex, Pipedrive focuses on simplicity..."

Final visibility calculation:

visibility = round(
  (coverage × 0.25) +
  (placement × 0.25) +
  (structural_prominence × 0.20) +
  (frequency × 0.15) +
  (contextual_framing × 0.15)
)

Rank and Position

Absolute Rank vs Local Rank

OneGlance uses absolute rank (reading order across the entire response), NOT local rank (position within a sub-category).

Example:

AI Response:

## Best CRM for Small Teams
1. HubSpot
2. Pipedrive
3. Freshsales

## Best CRM for Enterprise
1. Salesforce
2. Microsoft Dynamics
3. SAP CRM

Brand	Local Rank	Absolute Rank
HubSpot	#1 in Small Teams	#1 (first overall)
Pipedrive	#2 in Small Teams	#2
Freshsales	#3 in Small Teams	#3
Salesforce	#1 in Enterprise	#4 (fourth brand mentioned)
Microsoft Dynamics	#2 in Enterprise	#5
SAP CRM	#3 in Enterprise	#6

From packages/services/src/analysis/analysisPrompt.ts:129:

In this example, Salesforce’s LOCAL rank within “Best for Enterprise” is #1, but its ABSOLUTE rank in the full response is #4 (it is the 4th distinct brand listed overall).

Why Absolute Rank Matters

Users read responses sequentially. If your brand appears as “#1 for Enterprise” but is the 7th brand mentioned overall, most users may never reach it. Strategic implication: Aim for absolute top-3 positioning, not just top rank within a niche category.

isTopPick vs isTopThree

isTopPick:

Requires absolute rank #1 AND explicit superlative language
Examples: “best overall”, “top recommendation”, “our #1 pick”
Being listed first without superlatives → isTopPick = false

isTopThree:

True if absolute rank is 1, 2, or 3
No language requirement—purely positional

Sentiment Analysis

Sentiment measures the tone of AI mentions.

Sentiment Score (0-100)

From packages/services/src/analysis/analysisPrompt.ts:162:

Condition	Score Range
Response explicitly warns against or discourages the brand	0-20
Response notes significant drawbacks or unfavorable comparisons	21-40
Mention is purely factual/descriptive with zero evaluative language	41-59
Response uses favorable language WITH noted limitations	60-80
Response uses enthusiastic superlatives with NO caveats	81-100

Anti-inflation rules:

A score of 81+ requires EXPLICIT superlatives (“excellent”, “best”, “standout”)
“Good”, “solid”, “popular” → 60-75 range, NOT 80+
Being listed in a recommendation list does NOT automatically = positive sentiment
If response lists BOTH pros AND cons → sentiment cannot exceed 79

Sentiment Label

Automatic label based on score:

very_positive (81-100)
positive (60-80)
neutral (41-59)
negative (21-40)
very_negative (0-20)

Positives and Negatives Arrays

Text snippets explaining sentiment. Example from an 85-sentiment response:

{
  "sentiment": {
    "score": 85,
    "label": "very_positive",
    "positives": [
      "Excellent integration ecosystem",
      "Standout customer support",
      "Best-in-class mobile app"
    ],
    "negatives": []
  }
}

Example from a 45-sentiment response:

{
  "sentiment": {
    "score": 45,
    "label": "neutral",
    "positives": [],
    "negatives": []
  }
}

If sentiment is consistently low (<40) across multiple prompts, review the negatives array to identify recurring criticisms. These are signals for product or messaging improvements.

Recommendation Type

Classifies the strength of AI endorsement. From packages/services/src/analysis/analysisPrompt.ts:334:

Definition: Brand is explicitly named as the #1 choice with superlative language.Requirements:

Absolute rank #1
Language: “best overall”, “top recommendation”, “our #1 pick”
Must be positioned as top recommendation of the ENTIRE response (not just a sub-category)

Points: 100Example:

“For CRM tools, HubSpot is our top recommendation. It consistently delivers the best balance of features, usability, and support.”

Competitive Analysis

Competitor Extraction

From packages/services/src/analysis/analysisPrompt.ts:286: Included as competitors:

Brands directly compared to yours in the same response
Brands listed alongside yours in rankings
Brands mentioned in the same category discussion

Excluded:

Brands in a completely different section/topic
Generic category references (“CRM software” is not a competitor)
Your own brand (never listed as its own competitor)
Brands only in the user’s prompt (not AI’s response)

Competitor Deduplication

From packages/services/src/analysis/analysisPrompt.ts:302: Rule: Sub-products are consolidated under the parent brand. Examples:

“Zoho CRM” + “Zoho One” + “Bigin by Zoho” → name: “Zoho”
“Google Workspace” + “Gmail” + “Google Docs” → name: “Google”
“Salesforce Sales Cloud” + “Service Cloud” → name: “Salesforce”

Why: Prevents inflated competitor counts and provides clearer market landscape.

Competitor Metrics

Each competitor has:

Visibility

Competitor’s prominence in the response (same 5-dimension formula as your brand).

Sentiment

How positively/negatively the competitor is mentioned.

Rank Position

Competitor’s absolute rank in the response.

isRecommended

Boolean: Was the competitor explicitly recommended?

winsOver

Areas where the competitor beats your brand (per AI response).

losesTo

Areas where your brand beats the competitor.

Competitive Landscape Table

On the Dashboard, the Competitive Landscape card shows:

// From apps/web/src/app/(auth)/dashboard/page.tsx:178
<CompetitiveLandscape competitors={metrics.competitorData} />

Columns:

Brand: Competitor name (deduplicated)
Mentions: Number of responses where competitor appeared
Avg Rank: Average absolute position across all mentions
Sentiment: Average sentiment score

Use case: Quickly identify which competitors are most frequently mentioned and how they’re positioned relative to you.

Advanced Metrics

Brand Perception

Extracted from AI responses about how your brand is perceived.

Core Claims

Key statements AI providers make about your brand. Example:

{
  "perception": {
    "coreClaims": [
      "Intuitive interface designed for non-technical users",
      "Strong email marketing integration",
      "Generous free tier for startups"
    ]
  }
}

Strategic use: Verify these claims match your messaging. If AI has incorrect or outdated claims, publish updated content.

Differentiators

What AI says sets your brand apart from competitors. Example:

{
  "perception": {
    "differentiators": [
      "Visual pipeline management",
      "Mobile-first design",
      "Built-in calling and email tracking"
    ]
  }
}

Strategic use: Double down on differentiators that resonate. If AI doesn’t mention your key differentiators, increase content emphasizing them.

Best Known For

Single phrase summarizing brand perception. Examples:

“Ease of use and intuitive interface”
“Enterprise-grade security and compliance”
“Affordable pricing for small teams”

Pricing Perception

AI’s understanding of your pricing tier. Values:

premium: High-cost, enterprise-focused
mid_range: Balanced pricing
budget: Low-cost option
free: Freemium or fully free
not_mentioned: AI didn’t discuss pricing

If your pricing perception is inaccurate (e.g., AI thinks you’re premium when you’re mid-range), it may hurt recommendations. Publish pricing transparency content and get listed on comparison sites.

Risk Identification

Flags issues in AI responses that could harm your brand. Risk types (from packages/services/src/analysis/analysisPrompt.ts:353):

outdated_info

Definition: AI states something factually outdated about your brand.Example:

“HubSpot does not offer native LinkedIn integration.” (If you launched LinkedIn integration 6 months ago, this is outdated.)

Severity: Usually critical if materially incorrect.Action: Publish press releases, update review site listings, create content explicitly stating the correction.

factual_error

Definition: AI makes an incorrect claim (not just outdated—factually wrong).Example:

“Pipedrive is owned by Salesforce.” (Incorrect—Pipedrive is independent.)

Severity: criticalAction: File corrections with AI providers if possible (OpenAI, Google have feedback mechanisms). Increase authoritative content stating the fact.

brand_confusion

Definition: AI conflates your brand with another or attributes another brand’s features to yours.Example:

“Zoho CRM, formerly known as HubSpot…”

Severity: criticalAction: Severe brand positioning issue. Increase distinct branding in all content. Consider trademark enforcement if appropriate.

negative_association

Definition: AI associates your brand with a negative category or outcome.Example:

“CRM tools known for poor customer support include [YourBrand] and LegacyCRM.”

Severity: critical to warningAction: Investigate if the criticism is valid. If unfair, launch reputation management campaign.

missing_from_response

Definition: Your brand is absent from a response where it objectively should appear.Example: Prompt: “What are the best CRM tools?” AI lists 10 competitors but not your brand, despite being a major player.Severity: critical (indicates severe visibility gap)Action: Increase content marketing, get featured on major review sites, improve SEO for category terms.

Actions and Recommendations

Each analysis provides 3-5 specific, actionable recommendations. Priority levels:

critical: Brand is actively harmed—immediate action required
high: Major missed opportunity
medium: Optimization opportunity
low: Nice-to-have improvement

Example actions output:

{
  "actions": [
    {
      "priority": "critical",
      "recommendation": "Correct factual error: AI states [YourBrand] lacks mobile app, but you launched it in Q2 2024. Publish announcement and update all review site profiles."
    },
    {
      "priority": "high",
      "recommendation": "Brand is mentioned but ranks #6 absolute (behind 5 competitors) despite strong features. Publish comparison content: '[YourBrand] vs [Top Competitor]' targeting this prompt's keywords."
    },
    {
      "priority": "medium",
      "recommendation": "Sentiment is neutral (score 52) due to lack of evaluative language. Encourage customer reviews and case studies to generate more positive third-party content AI can reference."
    }
  ]
}

Using Metrics for Strategy

Scenario 1: Low Presence Rate (<40%)

Diagnosis: AI providers lack awareness of your brand. Action plan:

Content blitz: Publish 10-20 authoritative articles targeting your category
Review sites: Get listed on G2, Capterra, TrustRadius with customer reviews
PR campaign: Secure mentions in industry publications AI likely trains on
Backlink building: Increase domain authority so AI trusts your content
Re-run prompts monthly: Track presence rate improvement

Timeline: 3-6 months to see impact (AI models retrain slowly)

Scenario 2: High Presence (70%+) but Low GEO Scores (<50)

Diagnosis: AI knows your brand but doesn’t recommend it. Action plan:

Analyze sentiment and negatives: What criticisms are recurring?
Address product gaps: If AI mentions “lacks feature X”, prioritize building it
Messaging overhaul: If perception is off, update positioning across all channels
Competitive comparison content: Publish “[YourBrand] vs [Competitor]” emphasizing wins
Customer proof: Case studies, testimonials, and reviews that counter negative perceptions

Timeline: 1-3 months for messaging changes to propagate

Scenario 3: Strong in Some Prompts, Weak in Others

Diagnosis: Visibility is uneven across topics. Action plan:

Identify patterns: Group prompts by topic/buyer persona
Double down on strengths: Expand content for high-performing topics
Fill gaps: Create content specifically for low-performing topics
Test prompt wording: Sometimes rephrasing a prompt changes AI behavior

Example:

High GEO: Prompts about "small business CRM"
Low GEO: Prompts about "enterprise CRM"

Action: Launch enterprise-focused content series and case studies

Scenario 4: Competitor Consistently Outranks You

Diagnosis: One competitor dominates AI mentions. Action plan:

Study competitor’s sources: What sites does AI cite when mentioning them?
Get featured on those sources: Aim for G2, Gartner, review sites they dominate
Direct comparison content: Publish “Why [YourBrand] vs [Competitor]” content
Monitor competitor differentiators: See what AI says they do better, close those gaps
Track their GEO scores: Use OneGlance to add prompts mentioning the competitor

Scenario 5: Factual Errors in AI Responses

Diagnosis: AI has incorrect or outdated info. Action plan:

Document errors: Screenshot AI responses showing incorrect claims
Publish corrections: Blog posts, press releases explicitly correcting the info
Update review sites: Ensure G2, Capterra, etc. have current accurate info
File feedback with AI providers: OpenAI, Google, Anthropic have feedback forms
Re-run prompts monthly: Verify corrections propagate to AI models

AI model updates are infrequent (months). Don’t expect instant corrections. Consistent, authoritative content over 3-6 months is required.

Exporting and Reporting

Share metrics with stakeholders via exports.

Dashboard Export

From the Dashboard, click Export → JSON or CSV. Contents:

Aggregate stats (presence rate, avg rank, top competitor)
Impact metrics (total responses, recommendation rate, sentiment)
Brand perception (core claims, differentiators, pricing)
Source intelligence (top domains, citation counts)
Competitor data (visibility, sentiment, rank for each)

Prompts Export

From the Prompts page, click Export. Contents:

Prompt-level metrics (GEO score, sentiment, visibility, position)
Full AI responses for each prompt
Response-level metrics (per provider)
Sources and citations

Custom Reporting

For recurring executive reports, consider:

Export JSON: Richer data for custom analysis
Process in Python/R: Calculate custom metrics
Visualize: Use Tableau, Looker, or custom dashboards
Automate: Use OneGlance API to schedule exports

Next Steps

Managing Prompts

Refine your prompt strategy based on metric insights

Scheduling

Track metric trends over time with automated runs

Team Collaboration

Share metric reports and coordinate strategy across teams

API Reference

Programmatically access metrics for custom dashboards

Setup Guide - Initial workspace configuration
Analysis Concepts - How OneGlance calculates metrics
Workspace API - Programmatic access to metrics

Get Started

Core Concepts

Guides

​Overview

​Core Metrics Explained

​GEO Score (Generative Engine Optimization)

Visibility (25%)

Rank (25%)

Sentiment (25%)

Recommendation (25%)

​Interpreting GEO Score Ranges

​Detailed Metrics Breakdown

​Presence Metrics

​Presence Rate

​Mention Count

​Visibility Score (0-100)

​Rank and Position

​Absolute Rank vs Local Rank

​Why Absolute Rank Matters

​isTopPick vs isTopThree

​Sentiment Analysis

​Sentiment Score (0-100)

​Sentiment Label

​Positives and Negatives Arrays

​Recommendation Type

​Competitive Analysis

​Competitor Extraction

​Competitor Deduplication

​Competitor Metrics

Visibility

Sentiment

Rank Position

isRecommended

winsOver

losesTo

​Competitive Landscape Table

​Advanced Metrics

​Brand Perception

​Core Claims

​Differentiators

​Best Known For

​Pricing Perception

​Risk Identification

​Actions and Recommendations

​Using Metrics for Strategy

​Scenario 1: Low Presence Rate (<40%)

​Scenario 2: High Presence (70%+) but Low GEO Scores (<50)

​Scenario 3: Strong in Some Prompts, Weak in Others

​Scenario 4: Competitor Consistently Outranks You

​Scenario 5: Factual Errors in AI Responses

​Exporting and Reporting

​Dashboard Export

​Prompts Export

​Custom Reporting

​Next Steps

Managing Prompts

Scheduling

Team Collaboration

API Reference

​Related Concepts

Build docs developers (and LLMs) love

Overview

Core Metrics Explained

GEO Score (Generative Engine Optimization)

Interpreting GEO Score Ranges

Detailed Metrics Breakdown

Presence Metrics

Presence Rate

Mention Count

Visibility Score (0-100)

Rank and Position

Absolute Rank vs Local Rank

Why Absolute Rank Matters

isTopPick vs isTopThree

Sentiment Analysis

Sentiment Score (0-100)

Sentiment Label

Positives and Negatives Arrays

Recommendation Type

Competitive Analysis

Competitor Extraction

Competitor Deduplication

Competitor Metrics

Competitive Landscape Table

Advanced Metrics

Brand Perception

Core Claims

Differentiators

Best Known For

Pricing Perception

Risk Identification

Actions and Recommendations

Using Metrics for Strategy

Scenario 1: Low Presence Rate (<40%)

Scenario 2: High Presence (70%+) but Low GEO Scores (<50)

Scenario 3: Strong in Some Prompts, Weak in Others

Scenario 4: Competitor Consistently Outranks You

Scenario 5: Factual Errors in AI Responses

Exporting and Reporting

Dashboard Export

Prompts Export

Custom Reporting

Next Steps

Related Concepts