Real AI parse accuracy (user-rated)
Real scores users gave us after reviewing parsed protocols, aggregated per-discipline. Transparency over marketing.
Updated in real-time as users submit ratings.
Rated
40
Success
63%
Partial
35%
Failed
3%
By discipline
| Discipline | Rated | Success | Partial | Failed |
|---|---|---|---|---|
| Life Science | 15 | 47% | 53% | 0% |
| Biology | 8 | 88% | 0% | 13% |
| Materials & Nanoscience | 6 | 50% | 50% | 0% |
| Chemistry & Chemical Engineering | 6 | 83% | 17% | 0% |
| Agriculture & Food Science | 3 | Too few samples (< 5) | ||
| Environmental & Earth Science | 1 | Too few samples (< 5) | ||
| Computer Science | 1 | Too few samples (< 5) | ||
Why we publish failures
Most AI products only show "95% accurate" hero numbers. We don't. Every parse is rated by a real user after they review the output — including the failures. Transparency is the only way you can decide whether to trust us with your literature reproduction.