Methodology
How we test, and the rules that keep us honest
The method every stack must pass
Real work, real stakes
Every tool is used inside actual paid client projects by a working professional in that field. Never demo data, never marketing-page feature lists. Minimum two weeks per tool.
Head-to-head, same task
Competing tools get the identical brief. Same article, same client call, same deck. We keep the outputs and time logs.
Scored on four axes
Output quality, time actually saved (tracked, not estimated), setup friction, and value at the price a solo professional pays, with no enterprise-discount fantasy pricing.
Re-tested every 90 days
AI tools change monthly. Every stack page shows its last test date and next scheduled re-test. If a pick gets worse, it gets removed publicly, with a changelog note.
The honesty rules
- Commissions never touch rankings. The scoring is finished before we check whether a tool even has an affiliate program.
- We recommend $0-commission tools when they win. Claude is our top writing pick and pays us nothing. Every click on that link is revenue we chose not to make, and it's the best marketing we have.
- Every review has a "skip this if." If we can't name who shouldn't buy it, we haven't tested it hard enough to publish.
- Rejected tools are listed with reasons. What we said no to tells you more than what we said yes to.
- No sponsored placements. No "featured" slots. Ever. Vendors can't pay to appear here, and this rule has no price.
What we do with the money
Affiliate commissions fund the subscriptions (we pay full price for every tool we test), the testers' time, and the 90-day re-test cycle. Reader-supported means exactly that: you buying through our links is what keeps the testing independent of vendors.
Found a problem?
If a recommendation didn't hold up for you, we want to know. That's re-test signal. Email contact@provenstacks.com and it goes straight to the tester who wrote the page.