AI comparison methodology
How sources, votes, scores and caching fit together.
Sources and model opinions
Supported free manufacturer pages and Wikipedia provide source material, with retrieval dates and limitations. A source link is not independent verification. Scores, recommendations and reasons are subjective AI outputs, not lab tests, customer reviews or verified product ratings.
Consistent criteria
Products use a shared category framework. Unknown categories can receive a framework before evaluation; all voters then use the same dimensions. Product existence and missing specifications are never established using model memory. Browsing tools are disabled for the comparison models.
Votes versus fit scores
Each successful independent model chooses one option. Equal vote counts are reported honestly as a tie. A weighted fit score is a separate calculation from available dimension scores. Adjusting weights never rewrites votes or base scores, and missing values do not become zero.
Why models disagree
Models may value trade-offs differently and may make mistakes or hold outdated knowledge. A numerical score is not a measurement. Examine the source dates, uncertainty and reasons before relying on a result.
Shared cache and published snapshots
Products, evaluations, choices, reports and narrative translations can be reused. Cache identities account for model, prompt, product facts and framework versions; product order is normalized where appropriate. A published report is a durable snapshot and does not silently change when cache contents change. Dates describe the saved analysis, not current prices or stock.
Corrections
Report a problem through Contact & corrections, including the public report URL where possible. Check exact region, edition and current manufacturer information before buying. No purchase, medical, legal or financial advice is provided.