How We Score Brand Visibility

Overview

GEO Hub brand visibility scores measure how frequently an international brand is spontaneously recommended by China's major AI models when users ask category-relevant questions. Scores range from 0 to 100 per brand per model, where 100 means the brand was cited in all tested queries, and 0 means it was never cited.

Scores reflect organic AI citation frequency — not paid placement, not branded search, and not prompted mentions. The key question we answer for each brand is: "If a Chinese user asks an AI for brand recommendations in your category, will your brand come up?"

Models Tracked

We run queries across six Chinese AI platforms that collectively reach over 500 million monthly active users:

Query Framework

For each brand category, we use a standardized set of 12 queries designed to represent the types of questions real users ask when seeking brand recommendations. Queries are:

• Phrased in English (to test international brand positioning)
• Category-specific, not brand-specific (we never prompt with the brand name)
• Covering both direct recommendation requests and comparison/ranking requests
• Consistent across all models for fair comparison

Score Calculation

For each brand–model pair, we record whether the brand appears in the AI's response to each query. A brand is considered "cited" when it is mentioned as a recommendation or example — not merely in a disclaimer or off-topic context.

Scores are rounded to the nearest integer. A score of 75 means the brand appeared in 9 of 12 queries for that model. A score of 0 means the brand was never cited across all 12 queries — not that the AI has never heard of the brand.

Why Scores Vary Across Models

Each AI model has different training data composition and citation behavior. A brand like Tesla may score higher on DeepSeek (which has broad international knowledge) than on Wenxin (which is more anchored in Chinese-language media where Tesla's coverage has been more mixed). These differences are meaningful — they reflect each model's implicit brand associations, not measurement error.

Cross-model variation is one of the most actionable signals in GEO Hub data: a high Qwen score combined with a low Wenxin score suggests a brand is well-represented in Alibaba's e-commerce ecosystem but underrepresented in Baidu's Chinese-language content graph — a specific and fixable gap.

Limitations & Honest Caveats

Editorial Independence

GEO Hub does not accept payment to improve a brand's score, offer "featured" placements, or adjust scores based on commercial relationships. Scores reflect independent measurement only.

Brands are included based on their global relevance and China market presence. Inclusion or exclusion does not reflect editorial opinion about a brand's quality or strategy.

Update Cadence

Brand visibility scores are updated monthly. Each update re-runs the full query framework across all six models. Historical scores are archived so trend changes can be tracked over time.