Independent AI Commerce Research Platform
RecoScope is a live recommendation intelligence system that benchmarks product recommendations across ChatGPT, Claude, Gemini, and Perplexity, normalizes the results, and tracks how visibility changes over time.
Designed, built, and operated by Robert Hu.
447
Brands Tracked
4
AI Models
15
Categories
87
Benchmark Runs
How the system works
Collect
The same prompts, across every model
The same commercial-intent prompts are evaluated across multiple AI systems, in the same time window, so results are comparable rather than anecdotal.
Normalize
Raw responses become structured data
Raw recommendations are transformed into consistent brand, ranking, and category data, with brand names normalized so the same brand is counted as one.
Analyze
Longitudinal, human-reviewed findings
Longitudinal patterns, model differences, and recommendation changes are published through human-reviewed reports.
A Finding From The Data
AI models recommend brands based on different signals than marketplace search. A category-leading product can be almost entirely absent from AI recommendations.
In the running shoes benchmark, Nike leads in total mentions but no model ranks it in its top 3. Meanwhile ASICS and Brooks, with smaller marketplace share, dominate the AI picks. That gap is what the platform is built to measure.
See the running shoes benchmark →Marketplace Best Sellers
Nike
#1 in Running Shoes
Marketplace visibility: dominant
AI Recommendations
Nike
Not in any model’s top 3
AI visibility: nearly invisible
Top brands by AI mention volume
Sample benchmark output
Each report captures what every model recommended, in order, for the same set of prompts. Lawn Fertilizer, week of Apr 8, 2026.
| Agent | #1 | #2 | #3 |
|---|---|---|---|
| ChatGPT | Scotts Turf Builder | Jonathan Green | The Andersons |
| Claude | Scotts Turf Builder | Lesco | Milorganite |
| Gemini | Milorganite | The Andersons | Scotts Turf Builder |
| Perplexity | The Andersons | Milorganite | Scotts Turf Builder |
Live Benchmarks
Each category is benchmarked on a recurring schedule across ChatGPT, Claude, Gemini, and Perplexity, so the data shows how recommendations move, not just where they stand today. The tracker, research, and prompt pages are all outputs of the same shared longitudinal system.
How the models are evaluated
Each model is classified by commercial interest, so the data shows not just what AI recommends, but why different models diverge.
Classified by commercial interest
Independent models (Claude), search-grounded models (Perplexity), and commerce-influenced models (ChatGPT, Gemini) are separated to reveal how commercial integrations shift what gets recommended.
Standardized and comparable
The same prompts run across every model on a recurring schedule. Responses are parsed for brand mentions, rank position, and frequency to build comparable datasets over time.
Independent and integrity-first
No brand pays to influence rankings. Reports are published only after human review, and reflect organic model behavior at the time of testing.
RecoScope is an independent AI commerce benchmark designed, built, and operated by Robert Hu.
From system design and data architecture through methodology and published findings.
Why I built RecoScope