Guideai automation tools

How We Test AI Software

The practical editorial process behind evidique reviews, from account sourcing and test tasks to scoring, corrections, and affiliate disclosure.

Published Jul 1, 2026 Updated Jul 8, 2026

Accounts and access

We record whether we used a free trial, paid account, or vendor-provided access. Vendor access never guarantees coverage or a positive result. A future article will label the access method beside its test notes.

Test tasks

Each review starts with tasks drawn from the reader’s actual job: ship a TypeScript change, audit a small site, build a landing page, or connect two services. We write the acceptance checks before touching the tool.

Scoring

We score seven dimensions from 0 to 10. Output quality carries 25% of the final score, ease of use 20%, features and value 15% each, reliability and integrations 10% each, and documentation 5%.

Pricing checks

An editor checks the official pricing page and records the date. We never call that data real-time. Old articles get a fresh check before a meaningful update.

Sponsors cannot buy a score, category win, or editorial approval. Commercial links carry clear disclosure and the appropriate link attributes.

AI in our workflow

AI may help with transcription, outlines, or copy checks. A named human editor remains responsible for every claim, test record, and correction.

Corrections and updates

Send a specific correction to the address on our contact page. We verify it, update the visible date, and note material changes. Old reviews are re-tested when a product changes enough to alter the verdict.