New this week: 6 head-to-head tests added, including Claude vs Gemini for long documents.
See what changed →The rules our writers and testers work to, written down so you can hold us to them.
These standards describe how we decide what to publish, how a ranking is produced, and what we do when we get something wrong. They apply to every page on this site, including pages that carry commercial links.
No company can buy a position, an inclusion, a badge or the removal of a criticism. We do not publish sponsored reviews, paid listings or vendor-supplied copy, and we do not accept payment to add links to existing pages.
Rankings are produced and locked before any commercial link is attached to a page. The person scoring a category is not the person who manages commercial relationships.
Every claim about a tool comes from one of three places: our own testing on a seat we paid for, the vendor’s own public documentation, or a named third-party source we link to. Where we could not verify something, the page says so explicitly rather than implying we checked.
Each category is scored against a fixed brief of forty tasks using six criteria: output quality, real monthly cost, speed under load, data handling, export and lock-in, and support and stability. The full process is documented on the how we test page.
Output scoring is blind. Results are stripped of branding and scored independently by two people, and disagreements are reconciled by re-running the task rather than by averaging.
Mistakes are corrected in place rather than quietly deleted. When a correction changes a ranking, price or recommendation, we note the change and the date on the page itself.
If you have spotted something wrong, please tell us with a link to the page and, where possible, a source. Pricing and feature corrections are usually live within one working day.
Prices and plan limits are re-checked weekly. The full test brief is re-run at least quarterly and whenever a tool ships a significant release. A tool is removed from a hub if it shuts down, or if it changes so much that the existing review would mislead a reader.
Any relationship beyond a standard affiliate programme — former client work, an investment, a personal connection to a founder — is disclosed on the relevant page. Where a conflict cannot be managed, we do not cover the tool.
We use AI tools for transcription, tidying notes and generating test inputs. We do not publish AI-written reviews, scores or verdicts. Every published judgement is made by a person who ran the tests.
Some outbound links earn a commission. They never affect position or score, they are labelled where they appear, and they are added only after scoring is complete. Full details are in the affiliate disclosure.
If you believe a page breaches these standards, write to us at hello@aialternatives.io with the page link and the specific issue. We aim to respond within two working days and will publish a correction where one is warranted.