How we compare AI alternatives: our substitution rubric
Every comparison here is built on the same rubric: form factor, cost model, exit cost, licence tier, and hosting. We verify vendor facts on publication day, date every claim, publish no editorial scores, and name what we could not verify rather than filling the gap.
By the AI Alt Finder team
This page explains the method behind every comparison on this site, so you can judge whether our conclusions deserve any weight. It is deliberately short on philosophy and long on what we do and do not check.
The five axes, in the order we apply them
- Form factor: terminal, editor extension, standalone app, or API. Two tools with different form factors are rarely substitutes even when their features overlap.
- Cost model: seat, usage, credits, or hybrid. Being on the wrong model costs more than being with the wrong vendor.
- Exit cost: what you can export, on which plan, and whether the export functions without the vendor.
- Licence: what your specific tier permits, especially commercial use, and whether rights survive a downgrade.
- Hosting: who runs it, where the data goes, and who is responsible for keeping it patched.
Where our facts come from
Vendor pricing pages, vendor documentation, and machine-readable sources such as repository licence fields. Every product claim in a comparison here carries the date it was checked, and we re-check on publication day rather than trusting notes from a previous session.
Where a vendor page could not be read directly, we say so and describe structure rather than quoting a figure we did not verify. A missing number is better than a confident wrong one.
What we refuse to publish
- Editorial scores and star ratings. We have no first-party user votes, and an invented score out of five is a decoration pretending to be a measurement.
- Repository star counts and follower numbers, which are stale within weeks and tell you nothing about fitness for your task.
- Revenue and user-count figures that trace back to secondary blogs rather than the company or a filing.
- Feature matrices assembled from marketing pages. A tick in a grid hides the difference between a feature that exists and a feature that works.
How we handle products that change weekly
Some comparisons cover products only months old, where a rename or an acquisition can invalidate a page overnight. For those we do two things: we write about structure rather than benchmark results, because structure decays more slowly, and we put a dated freshness warning on the page itself.
We also record naming history where it exists, because a product that has been renamed several times leaves its documentation scattered under dead names, and that is genuinely useful to a reader searching for help.
Conflicts of interest, stated plainly
We are not affiliated with any vendor compared on this site, and brand names appear only to identify the products under discussion. Comparisons are not sponsored, and no vendor is given the ability to alter a verdict about itself.
If you build one of the tools discussed here and a fact is wrong or out of date, the fastest correction path is a submission with the vendor page that proves it. We would rather fix a date than defend one.