Y-AIV Score Methodology

The algorithm, data sources, sample limitations, and update mechanism behind the SilkGeo AI Visibility Score (Y-AIV) are fully transparent. We believe only explainable scores are valuable.

Score Overview

The Y-AIV Score is a composite score measuring brand visibility across AI platforms, ranging from 0 to 100. It is based on 7 dimensions, each with independent data sources and calculation methods.

DimensionWeightDescription
Mention Rate 20%Frequency of brand being mentioned in AI responses
Recommendation Rate 20%Frequency of AI proactively recommending the brand
Citation Rate 20%Frequency of AI responses citing brand-related sources
Brand Accuracy 15%Alignment between AI descriptions and brand facts
Share of Voice 10%Brand's share of AI mentions within its industry
Authority 10%Authority level of cited sources
Trend 5%Score change direction over the observation period

Scoring Formula

Y-AIV Score =
  Mention × 0.20
+ Recommendation × 0.20
+ Citation × 0.20
+ Accuracy × 0.15
+ SOV × 0.10
+ Authority × 0.10
+ Trend × 0.05

Each sub-dimension is normalized to a 0-100 score, then weighted and summed to produce the final Y-AIV Score.

Data Sources

AI Platforms
ChatGPT / Gemini / Claude / Perplexity / DeepSeek / Doubao / Grok / Copilot etc. 12+ platforms
Prompt Sources
Industry-relevant prompt library + real user queries + auto-generated
Data Period
Rolling 30-day window (Trend compares 7/30/90 days)
Sample Size
At least 500+ AI queries per brand per cycle

AI Platform Tiers

Prompt Categories

Dimension Calculation Methods

Mention Rate

Number of queries where the brand is mentioned in AI responses / total queries. Mention detection uses an Ensemble method: exact match + alias match + entity match + embedding similarity + LLM verification, outputting a confidence score.

Recommendation Rate

Number of queries where AI proactively recommends the brand / total recommendation-type queries. Recommendation ≠ mention: only counted when AI explicitly suggests the user choose the brand.

Citation Rate

Number of queries where AI responses cite brand-related sources / total queries. Citation detection extracts URLs from raw responses, normalizes domains, deduplicates, and classifies them.

Brand Accuracy

Compares AI brand descriptions with official brand information (website / Wikidata / Schema.org), calculating factual deviation rate. Deviation types include: founding date, product category, target audience, headquarters location, etc.

Share of Voice

Brand AI mentions / total AI mentions of all brands in the same industry. Can be sliced by industry / country / language / AI platform / prompt type.

Authority

Weighted authority score of cited sources. Source types include: Official Website / News / Wikipedia / Reddit / YouTube / Social / Review / Academic / Government, each with different base weights.

Trend

Change in current period score vs previous period score. Positive indicates upward trend, negative indicates decline.

Confidence & Limitations

Data Quality Assurance

Every core data record includes the following metadata:

Raw AI responses are permanently preserved, ensuring that historical data can be re-parsed when algorithms are upgraded in the future.

平台运行数据(实时)

以下数据由系统实时聚合,反映当前平台真实运行规模:

检测记录
关键词库规模
Prompt 库
品牌指数
AEO 报告
对比报表

数据来自站内实时聚合(/api/v1/unified/*),仅展示计数,不含任何关键词原文。

Measure Your AI Visibility with Y-AIV Score

Check your brand's AI visibility score for free

Free AI Visibility Check