Scoring methodology
One number, 5 criteria, published weights. If you disagree with the weights you can recompute the score yourself from the per-criterion ratings on every listing.
The formula
Each criterion is rated 0-5. The rating is divided by 5, multiplied by the criterion's weight, and the results are summed. If a criterion could not be assessed it is excluded and the total is rescaled across the remaining weight - so a missing criterion is never silently a zero.
score = round(sum(rating_i / 5 * weight_i) / sum(weight_i) * 100)The 5 criteria
| Criterion | Weight | What we ask |
|---|---|---|
| Detector pass rate | 40% | How many of the six detectors in the panel scored the output below the pass threshold, on the stated test date? |
| Meaning preserved | 25% | Does the rewritten text still say what the original said, including numbers, names, hedges and caveats? |
| Readability | 15% | Does a human reader find the output natural, or is it padded with synonyms and broken idioms? |
| Price and free tier | 10% | What does it cost, is the free tier large enough to judge the output, and are the limits visible before signing up? |
| Data handling | 10% | Is retention stated, is your text used for training, is an account required, can you delete your data? |
Why detection carries the most weight
Because it is the thing buyers in this category are actually buying, and the thing vendors are least accountable for. It is capped at 30% rather than higher for the opposite reason: a tool that defeats every detector while mangling the meaning of your text is not a good tool, and a score that says otherwise would be useless six months from now.
How detection is measured
Three fixed test texts, identical for every tool. Each is scored by all 6 detectors in the panel before and after humanizing. The criterion rating reflects the average drop across the panel, not the best single result - a tool that beats one detector and fails five does not score well here.
What the bands mean
| Score | Band |
|---|---|
| 92/100 | Excellent |
| 76/100 | Solid |
| 61/100 | Usable with caveats |
| 47/100 | Weak |
| 22/100 | Not recommended |
What cannot change a score
- Paying for any tier, add-on or sponsorship.
- Asking nicely, or asking with a lawyer.
- Being popular, being new, or being someone we like.
What can change it: a fact we got wrong, a product that changed, or a 180-day recheck. All three are free and all three are logged.