The RecoScope Platform
RecoScope evaluates how leading AI systems recommend products, converts unstructured responses into normalized longitudinal data, and publishes human-reviewed evidence about AI-driven product discovery.
How the system works
Evaluate
The same commercial-intent prompts are tested across ChatGPT, Claude, Gemini, and Perplexity in the same time window, so results are comparable rather than anecdotal.
Capture
Each model's full response is preserved, along with the brands it named, the order it named them in, and which prompt and agent produced them.
Normalize
Brand references are cleaned and standardized, connected to categories, prompts, agents, and time periods, so the same brand is counted as one across models and runs.
Analyze
Cross-model agreement, visibility, rank movement, consistency, and recommendation differences are evaluated across the longitudinal dataset.
Review
Each run moves through a draft, reviewed, and published status. Only published runs are exposed on public surfaces.
Publish
Tracker reports, research findings, prompt-level pages, and private analysis outputs are all generated from the same structured dataset.
Operating scale
4
AI systems evaluated
15
Categories tracked
78
Published brand-runs
428
Brands measured
4,414
Recommendation mentions
March 2026
Operating since
Figures reflect published, public benchmark data and update as new runs are reviewed and published.
The data model
Every public output is derived from one normalized model. Each entity connects to the next, so a single benchmark run is fully traceable from raw response to published finding.
Categories
The product categories under evaluation (office chairs, running shoes, protein powder, and more), each with a benchmark cadence.
↓
Runs
A single benchmark of a category at a point in time. Runs carry a status and are only public once published.
↓
Agent responses
The raw output from each model for each prompt in a run, preserved separately from the structured data derived from it.
↓
Brand mentions
Every brand a model named, with its rank position, the raw string, and a normalized brand name for consistent counting.
↓
Run insights
The interpreted findings for a run: cross-model differences, common traits, gaps, and the key takeaway.
Decisions behind the platform
The system reflects a series of product and operating decisions, each made to keep the evidence comparable, auditable, and honest.
Compare models instead of treating AI as one channel
Recommendations diverge across ChatGPT, Claude, Gemini, and Perplexity. Measuring each separately, and classifying them by commercial interest, is the only way to see where and why they disagree.
Repeat standardized prompts
The same commercial-intent prompts run on a recurring schedule so the data reflects a controlled, repeatable process rather than one-off screenshots.
Preserve raw responses separately from normalized data
Full model outputs are kept alongside the structured mentions derived from them, so findings remain auditable and re-derivable.
Separate raw and normalized brand names
Storing both the raw string and a normalized brand name keeps the evidence intact while making counts consistent across spelling and formatting variants.
Track longitudinally
Categories are benchmarked over time so the system captures how recommendations move, not just where they stand on one day.
Separate public and private outputs
Public tracker reports and private brand analyses are generated from one dataset but gated differently, so private work never leaks onto public surfaces.
Require publication status before exposure
Public queries only return published runs. A run in draft or reviewed status is never rendered on an indexable page.
Balance freshness with caching
Public pages use incremental regeneration and per-route caching so the data stays current without paying a cold database query on every request.
Protect research integrity
No brand pays to influence rankings. Results reflect observed model behavior at the time of testing.
What the system produces
Four public outputs, one shared dataset.
Tracker →
Live category benchmarks with cross-model rankings, trend data, and movement over time.
Research →
Findings drawn from the dataset: cross-model patterns and how recommendations shift.
Methodology →
How prompts are run, parsed, normalized, and scored, and how models are classified.
Private analysis →
A per-brand visibility report, scored from the same data. See an example output.
What RecoScope does not claim
RecoScope does not reproduce model training data, explain every causal factor behind a recommendation, or provide universal rankings. It measures observed model outputs under a controlled and repeatable evaluation process, and reports what those outputs show.
RecoScope was designed, built, and is operated by Robert Hu to make AI product discovery measurable.
Why I built RecoScope