How we score
Every score comes from this rubric. Version 1.1, updated Nov 5, 2026.
Five criteria, 100 points
For each job, we rate every tool from 0 to 10 on five criteria. We weight them, add them up and multiply by 10.
Capability
How good the output or result is for this specific job, judged against the other tools in the same use case.
Reliability
How consistently it delivers: stability, fewer failed runs, clear limits, and how often the vendor changes plans or retires features.
Ease of use
How quickly a newcomer gets a good result without setup, jargon or workarounds.
Value
What you get for the price: free tier, entry plan, transparency of limits, and how fast usage allowances run out.
Momentum & trust
Active development, vendor track record, privacy and licensing clarity.
score = round( 10 × (0.30·capability + 0.20·reliability + 0.15·ease of use + 0.20·value + 0.15·momentum) )
What a score means
- 90+: our default suggestion for this job.
- 80–89: a strong option, with a trade-off to know about.
- 70–79: good for a specific type of user.
- Below 70: niche, or held back by price or reliability.
Scores compare tools within the same job, so one tool can score differently on different pages.
The weights are our editorial choice, not a scientific formula. Ratings move in half-point steps, so a gap of a few points between two tools is not a meaningful difference. Read the pros, cons and prices before you choose.
Badges
- 🏆 Top pick: our highest editorial score for that job.
- 🆓 Top free option: the highest value rating among tools with an ongoing free plan (not a trial or one-time credits).
- 💰 Good value: the highest value rating for that job, 8.5 or higher.
- 🌱 Beginner-friendly: the highest ease-of-use rating for that job, 9 or higher.
- ⚡ Highest capability rating: the highest capability rating for that job, 9.5 or higher (excluding the top pick). It is our rating, not a benchmark.
What we claim, and don't
- Scores are editorial judgments based on published features, pricing and public track record. They are not lab benchmarks.
- We don't run controlled head-to-head tests. Claims about speed or quality come from vendor docs and public reports.
- Prices come from vendor pages where possible. Each tool page lists the pricing sources we checked, the date and any caveats; whether a price is billed monthly or yearly is shown as unconfirmed unless the page says so. Feature descriptions summarise vendor claims and are not tested. A price we could not confirm on an official page is marked “Price not confirmed” and is never counted as within budget in Help me choose.
- Our checks can miss changes and can be wrong. Always confirm on the official site.
Independence
Rankings aren't paid for. UseWhich has no affiliate links or sponsored placements today; if any are added, they will be clearly labelled. Sponsored placements, if they ever exist, will be labelled and will never affect a rank or score.
Changelog
- v1.1, Nov 5, 2026: wording review after a source audit. Badges renamed so they do not imply tests (“Most powerful” is now “Highest capability rating”, “Best free” is now “Top free option”), the free-option badge now needs an ongoing free plan, and we added a note that the weights are our editorial choice. Two value ratings were lowered after price corrections (Stable Audio, Hailuo). Version 1.0 (Nov 3, 2026) was the first public version.