UndetectedGPT vs. Undetectable AI
How do these similarly named services compare on matched Originality.ai scores and overall results?
Across 33 matched rewrites with numeric originality.ai scores, UndetectedGPT scored higher on 32, Undetectable AI on 0, and 1 tied. Their November composite scores were 50.65 and 12.61 out of 100, respectively.
Results on the same source texts
Each row compares both entries on the same source and includes pairs with two numeric scores. Meaning uses recorded embedding cosine. Detector scores reflect reported human-likelihood.
| Metric | UndetectedGPT higher | Undetectable AI higher | Tied | Measured pairs |
|---|---|---|---|---|
| Readability | 12 | 13 | 8 | 33 / 33 |
| Meaning preservation | 13 | 20 | 0 | 33 / 33 |
| GPTZero | 4 | 0 | 29 | 33 / 33 |
| ZeroGPT | 12 | 3 | 18 | 33 / 33 |
| Copyleaks | 25 | 0 | 8 | 33 / 33 |
| Originality.ai | 32 | 0 | 1 | 33 / 33 |
| Pangram | 32 | 0 | 1 | 33 / 33 |
| Grammarly | 32 | 0 | 1 | 33 / 33 |
| QuillBot | 33 | 0 | 0 | 33 / 33 |
Read the matched rewrites
These examples show the largest measured gaps in originality.ai, with one example per writing category.
- Academic essay ? S24
UndetectedGPT rewrite (32.0%) ? Undetectable AI rewrite (4.0%) - Application essay ? S03
UndetectedGPT rewrite (14.0%) ? Undetectable AI rewrite (2.0%) - Business email ? S18
UndetectedGPT rewrite (12.0%) ? Undetectable AI rewrite (2.0%)
Scope and source
Both entries use historical published commercial rewrites. The comparison uses the recorded outputs in the November 2026 dataset, version 2026-11-v1. See the methodology for metric definitions and limitations.
Model profiles: UndetectedGPT and Undetectable AI.