Editorial standards
How the numbers are made.
Every table, score and estimate in this library has a method. This page is the method.
Competitor data
Competitor names, entry prices, engine counts and feature flags are compiled from public pricing pages and product documentation in November 2026. Prices change; every comparison page says so. We do not use review-site averages as a score.
Editorial scores
Scores are 1–10 and weighted for a team that has to ship: engine coverage 25%, citation depth 25%, content workflow 20%, value for money 15%, time to value 15%. The weights are applied consistently across every comparison page.
ROI model
Manual checks assume one check per prompt per month at three minutes and $65/hour. The model is labelled wherever it appears. It is a planning tool, not a promise.
Benchmarks
Targets such as 25–40% share of answer at 90 days are planning ranges, not measured medians. We say so on the page. Replace them with your own baseline as soon as you have one.
Freshness
Every page carries a published and updated date. When a product fact changes, the page is regenerated and the date moves. The library is generated from a single product profile, which is why prices do not drift between pages.
AI use
The library is machine-generated from structured data and reviewed for accuracy before publication. The product profile, not the model, is the source of truth for prices and capabilities. Pages that need a human judgement — legal, medical, financial — carry a review note in the source data.