The 5W AI Visibility Index measures how brands are understood, described, and presented by the five major AI engines that now serve as the first research step for a growing share of buyers, investors, and partners. It is produced by 5W, a PR, creative and AI Visibility agency, and this is the methodology.
The three scoring methods at a glance
| Method | Used in | How it scores |
|---|
| A: 5W AI Visibility Index | Brand studies using this rubric | Mention, Accuracy, Sentiment, Completeness, Consistency; 0-100 composite |
| B: AI Visibility Index Franchise | Franchise studies in the Research Index | Frequency 40%, Rank 25%, Industry Accuracy 20%, Factual Accuracy 15% |
| C: Citation Share Index | Citation Share Index studies | Citation Frequency 40%, Cross-Engine Breadth 20%, Query-Type Breadth 20%, Extractability 15%, Crawl Access 5% |
Method versions and fieldwork dates
Everything-PR research uses more than one scoring method. Each study names the method it uses and the dates its prompts were run, so results can be compared like for like. Three versions are in use:
Method A: 5W AI Visibility Index scoring. The five-dimension rubric described on this page (Mention, Accuracy, Sentiment, Completeness, Consistency), scored per prompt per engine and converted to a 0 to 100 composite.
Method B: AI Visibility Index Franchise composite. The entity-level composite used in the franchise studies listed in the Everything-PR Research Index, weighted Frequency 40%, Rank 25%, Industry Accuracy 20%, and Factual Accuracy 15%.
Method C: Citation Share Index scoring. The modeled, directional composite used in the Citation Share Index studies, weighted Citation Frequency 40%, Cross-Engine Breadth 20%, Query-Type Breadth 20%, Extractability 15%, and Crawl Access 5%.
What every study should state. The method version (A, B, or C), the number of prompts, the engine panel, the measurement window with start and end dates, whether figures are observed or modeled, and the study's limitations. Scores from different method versions should not be compared directly. Example: the Indian Brand AI Visibility Index used Method C with a May 1 to May 28, 2026 measurement window.
Engine panel
All studies cover five engines: ChatGPT (GPT-4o, no web browsing), Claude (Sonnet, standard mode), Perplexity (default mode, web search enabled), Gemini Advanced (standard mode), and Google AI Overviews (via Google Search). Each engine is queried independently. Sessions are cleared between queries to prevent conversation context contamination.
Prompt design
Each brand is evaluated across four prompt categories:
Category A: Business description (15 prompts). "What is [brand]?" "What does [brand] do?" "Who founded [brand] and when?" Plus 12 variants. Tests entity completeness and accuracy.
Category B: Founder and leadership (12 prompts). "Who is [founder name]?" "What has [founder] said about [topic]?" Plus 10 variants. Tests named-person citation, the signal that compounds most over time.
Category C: Competitive positioning (18 prompts). "[Brand] vs [Competitor], what's the difference?" "What are the alternatives to [brand]?" "Which [category] company should I use?" Plus 15 variants. Tests whether the brand appears in competitive context and how it is positioned relative to peers.
Category D: Buyer intent (15 prompts). "Should I use [brand]?" "Is [brand] reputable?" "I need help with [use case], who should I hire?" Plus 12 variants. Tests the high-stakes queries where AI visibility has the most direct revenue impact.
Scoring (Method A)
Each response is scored across five dimensions:
Mention (0 to 2): Not mentioned = 0. Mentioned in passing = 1. Named and described = 2.
Accuracy (0 to 3): Not mentioned = 0. Significant errors = 1. Mostly accurate = 2. Fully accurate = 3.
Sentiment (0 to 2): Absent or negative = 0. Neutral = 1. Positive or recommended = 2.
Completeness (0 to 2): Missing key facts = 0. Basic description = 1. Complete with key differentiators = 2.
Consistency (0 to 1): Significant inconsistency across engines = 0. Consistent = 1.
Maximum score per prompt per engine: 10. Maximum total across 60 prompts and 5 engines: 3,000. Composite AI Visibility Score = (total score / 3,000) × 100.
Visibility tiers
Tier 1 (75 to 100): Complete, accurate, positive entity model. Appears consistently in category, competitive, and buyer-intent queries. Named founder citation established. Genuine AI visibility leadership.
Tier 2 (50 to 74): Mostly accurate but significant gaps in category positioning, competitive context, or founder citation. Common in brands with strong press coverage but weak entity infrastructure.
Tier 3 (25 to 49): Partial brand knowledge, accurate on basic identity, weak on positioning and leadership. Common in brands with earned media but no entity and schema layer.
Tier 4 (0 to 24): The "billion-dollar ghost" tier. AI engines don't know the brand, describe it incorrectly, or return hedged, uncertain answers that foreground competitors. Common in brands that have grown quickly without investing in communications infrastructure that reaches AI training data.
The methodology is updated annually to reflect changes in AI engine behavior, new engines entering the panel, and improvements in prompt design based on accumulated study data. Custom studies are available through 5W, the AI visibility agency that gets brands in the chatbox. See where 5W ranks in Best AI Visibility Agencies.
Related research: 5W IPO AI Visibility Index: The 25 Companies · Citation Share: The New Discoverability KPI · Everything-PR Research Index