How Is the BVS Score Calculated?
BVS (Brand Visibility Score) is a 0–100 score that measures a brand's AI visibility. It scores answers from 5 AI platforms on 5 dimensions (brand mention, semantic match, citation position, citation depth and source link) and weights them into one number.
Last updated: 2026-11-01 · Author: llmion
Formula Overview
Dimension weights in the two modes (search mode is the BVS v2 default):
| Dimension | Search mode (v2) | Memory mode (v1) |
|---|---|---|
| D1 Brand Mention | 25% | 30% |
| D2 Semantic Match | 15% | 25% |
| D3 Citation Position | 15% | 20% |
| D4 Citation Depth | 20% | 15% |
| D5 Source Link | 25% | 10% |
Five Scoring Dimensions
D1 Brand Mention
Whether the AI response mentions your brand name or aliases.
Scoring
With multiple samples per question and platform, D1 = mention rate × 100 (e.g. mentioned in 2 of 3 samples = 67). With a single sample, mentioned = 100, not mentioned = 0. Case- and width-insensitive, supports alias matching.
How to improve
Ensure your brand name appears in your core page titles, H1 headings, and within the first 200 words. Keep brand name spelling consistent.
D2 Semantic Match
How similar the context around your brand mention is to your core concepts (Entity terms). Only the 300 characters before and after the first mention are compared, not the whole response.
Scoring
Uses semantic vector comparison (text-embedding-3-small). Similarity ≥ 0.85 = 100, 0.70-0.85 = 60-100 (linear), 0.50-0.70 = 20-60, < 0.50 = 0. No brand mention at all = 0.
How to improve
Include your target Entity terms at least twice within the first 200 words of core pages. H1 and meta description should also contain these terms.
D3 Citation Position
Where your brand name first appears in the AI response.
Scoring
First 1/3 = 100 points, middle 1/3 = 60 points, last 1/3 = 25 points, not present = 0.
How to improve
Add a 'direct answer' paragraph at the top of your pages — 2-3 sentences giving the core conclusion. AI prefers citing content where direct answers are found at the top.
D4 Citation Depth
Whether AI actively recommends your brand, lists it as one of several options, or mentions it in passing, and whether the tone is positive, neutral or negative.
Scoring
Direct recommendation = 100, parallel mention = 65, indirect reference = 35, not mentioned = 0. Determined by AI analysis of response context. If the AI frames your brand negatively (drawbacks, warnings, advice against it), the score is halved (e.g. a negative parallel mention = 33); the dashboard question table flags it with a red "!".
How to improve
Add clear position statements to core pages (e.g., 'For clients with X needs, we are the best choice'). Use assertive language rather than neutral descriptions. When flagged negative, open the raw answer to see which drawback or controversy the AI cites and address it publicly (updated facts, an FAQ clarification, third-party reviews).
D5 Source Link
Whether the sources the AI actually cited after searching the web include your website (subdomains count).
Scoring
Cited = 100 points, not cited = 0 points. Scored on every platform (ChatGPT, Gemini and Claude run with web search; Perplexity and Google AI Overviews cite natively). Answers produced without a live search (memory mode) have no sources, so that question falls back to v1 weights.
How to improve
Make your own pages the source AI search cites: a direct answer at the top, concrete numbers, a real last-updated date, and Article / FAQPage structured data. Retrieval-based platforms favour recently updated, crawlable pages when ranking sources.
Platform Weighting
Final BVS is weighted across five platforms; when Google AI Overviews is not monitored, its weight is redistributed proportionally to the other four.
The world's most-used AI conversation platform, with the highest influence on brand recommendations. Monitored in live web-search mode.
The generative summary at the top of Google results, with the largest traffic impact. When it is not monitored, its weight is redistributed proportionally to the other four platforms.
A native AI search engine that cites sources on every answer, contributing most directly to traffic.
Google's AI assistant, monitored with Google Search grounding so it reflects your brand in Google's index.
Prefers professional content with a clear stance; strong influence in B2B. Monitored with the web search tool.
Competitor Scores and Share of Voice
Competitors are scored from the same AI answers as your brand, but with only the three rule-based dimensions: D1 brand presence, D3 position and D5 source links (whether the competitor's domain is among the cited sources). The weights are renormalised per BVS version to sum to 100%. D2 semantic match and D4 citation depth are not computed, so the dashboard labels these as "simplified BVS (no semantics or depth)" — not the same scale as your full BVS; treat the ranking as indicative.
Share of Voice records, for every AI answer, whether your brand and each competitor appear and which one appears first. Share = that name's appearances ÷ appearances of all tracked names; mention rate (share of answers that mention it) and first-appearance rate are also reported.
BVS Grade Definitions
BVS scores map to five grades, helping you quickly understand your brand's AI visibility status.
Strong
AI platforms actively and frequently recommend your brand as a top choice. Your brand holds a solid position on AI's recommendation list.
Visible
Your brand appears in most relevant AI responses, but may not always be the top recommendation. Continued optimization can push you to Strong.
Weak
Occasionally appears in AI responses, often in later positions or mentioned indirectly. Needs stronger structured content and Entity density.
Low
Rarely appears in AI responses, mostly as indirect associations. Needs to build AI-friendly content strategy from the ground up.
None
AI platforms almost never mention your brand when answering relevant questions. This is the most urgent state — but also represents the biggest growth opportunity.
Calculation Example
Suppose your brand is asked a question on ChatGPT and receives these scores:
D1 Brand Mention = 100 (mentioned)
D2 Semantic Match = 80 (high similarity)
D3 Citation Position = 60 (appeared in middle)
D4 Citation Depth = 65 (parallel mention)
D5 Source Link = 0 (your domain is not among the cited sources)
Question BVS = 100×0.25 + 80×0.15 + 60×0.15 + 65×0.20 + 0×0.25 = 25 + 12 + 9 + 13 + 0 = 59. Source links carry 25%, so a score is capped hard when your own site is not cited.
Frequently Asked Questions
What is the BVS score?
BVS (Brand Visibility Score) is llmion's 0–100 measure of a brand's AI visibility; the higher the score, the more often and more actively AI recommends your brand.
Every week llmion asks ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude your questions in live-search mode, scores each answer on brand mention, semantic match, citation position, citation depth and source link, and weights the platforms into a final score.
What is a good BVS score?
60 or above is Visible, and 80 or above is Strong, meaning AI actively recommends your brand as a first choice.
Below 20 is None, 20–39 is Low, and 40–59 is Weak. The grade thresholds are fixed and do not change when the algorithm is upgraded.
Why does the source link (D5) carry 25%?
AI platforms that search live cite their sources, and a citation of your site means the AI used your page as the basis for its recommendation.
The GEO study (Aggarwal et al., KDD 2024) found that adding citations, quotations and statistics raised content visibility in generative engine answers by 30–40% relative to the original. When an AI answers from memory without searching, there are no sources to check, so that question uses the v1 weights, where D5 is only 10%.
Why is the same question asked several times?
AI answers differ every time, and sampling several times reduces the score swings caused by a single answer.
On Pro and Agency, each question is sampled 3 times per platform every week and D1 = mention rate × 100, so being mentioned in 2 of 3 answers scores 67; Free and Starter sample once. Google AI Overviews is a summary on the search results page and is always sampled once.
What is the difference between search mode and memory mode?
In search mode the AI searches the web live before answering and cites sources; in memory mode it answers from training data only.
llmion monitors in search mode by default. When a platform's search is temporarily unavailable, that answer falls back to memory mode and is scored with the v1 weights. The two kinds of results are labeled separately and never mixed.
How is the score calculated without Google AI Overviews?
The 25% weight of Google AI Overviews is shared among the other 4 platforms in proportion to their weights.
ChatGPT then carries 40%, Perplexity about 26.7%, Gemini 20% and Claude about 13.3%, still 100% in total, so not monitoring one platform does not drag your score down.
Why are competitor scores on a different scale from mine?
Competitor scores are a simplified BVS that uses only 3 rule-based dimensions: brand mention, citation position and source link.
Semantic match and citation depth need extra AI analysis, which llmion runs only for your brand. The dashboard therefore labels competitors "Simplified BVS (no semantics or depth)"; treat their ranking as a reference and use share of voice to compare presence.
How often is BVS updated?
Monitoring runs and updates automatically every Monday at 02:00 UTC.
Paid plans also include 1–10 manual runs a month (depending on the plan), so you can check right after publishing. Four weeks after you publish content, the system also tracks the change in BVS automatically.
Methodology sources
BVS search mode and the source link dimension are based on the following research and official platform documentation:
- GEO: Generative Engine Optimization (Aggarwal et al., KDD 2024) — Research on generative engine optimization: adding citations, quotations and statistics raised content visibility by 30–40% relative to the original.
- Google Search Central: AI features and your website — Supporting links in AI Overviews come from pages that are indexed and eligible to be shown with a snippet.
- OpenAI: Overview of OpenAI crawlers — ChatGPT search reads web pages with OAI-SearchBot, which is configured separately from GPTBot, used for training.
- Gemini API: Grounding with Google Search — Gemini monitoring turns on Google Search grounding, and responses include their source pages.
- Claude: Web search tool — The web search tool used for Claude monitoring always includes citations in its answers.
- Perplexity API documentation — Perplexity answers are grounded in live web search and include built-in source citations.
Want to Know Your Brand's BVS Score?
The free plan includes a complete BVS score report. 5 minutes to set up, first report arrives next Monday.