SCRIBE RESULTS / 22 SEPTEMBER 2026
Preve
Averages across 5 graded runs of 11 consultations, the same audio and instruction for every scribe. 9 cases (no French or Hindi); no separate prompted-note step.
2m 12sTime to a signable note · ranked 5 of 7
18%Notes ready to sign as written · ranked 3 of 7
5.4Significant note errors per run · ranked 2 of 7
6.4Hallucination flags per run · ranked 4 of 7
17.6%Word error rate · ranked 5 of 7
Summary
Preve ranked 2 of 7 for significant note errors at 5.4 per run and for edit actions at 1.93 per note. Its clinical coverage was 92%, ranked 3 of 7.
Preve averaged 31.2 significant transcript errors per run, ranked 6 of 7, and 6.4 hallucination flags per run, ranked 4 of 7. Preve was tested on 9 cases, without French or Hindi, and has no separate prompted-note step.
Every measure
| Measure | Average | Run range | Rank |
|---|---|---|---|
| Time to a signable noteWait from pressing stop to the graded note, plus review time: how long until the note can be signed if it is reviewed straight away. | 2m 12s | 2m 10s–2m 16s | 5 of 7 |
| Notes ready to sign as writtenShare of notes that captured every checklist fact in full with no hallucination flags, so nothing needed editing. | 18% | 11%–33% | 3 of 7 |
| Editing effort0% means no editing required; 100% means writing the full note by hand. Fixes are weighted by the work they take: partial fact 1, missing fact 2, hallucination flag 3, and +2 when a fact is captured wrongly. | 18% | 17%–22% | 5 of 7 |
| Review time per noteEstimated clinician time to proofread and fix one note (190 wpm reading, 5 s to recall and 40 wpm to type each missing or partial fact, 12 s to verify and delete each hallucination flag, including a wrong fact before it is retyped). | 1m 51s | 1m 47s–1m 59s | 4 of 7 |
| Worst note, any runThe single slowest note to review and fix across every case and run. | 2m 45s | — | 2 of 7 |
| Time saved per noteEstimated time to write the note yourself (5 s to recall each checklist fact, typing it at 40 wpm) minus review time. | 2m 42s | 2m 23s–2m 49s | 3 of 7 |
| Significant note errors per runMissed facts, partial facts and hallucinations that a blind AI rater judged could change diagnosis, treatment, follow-up, safety or the legal record, across 11 cases. Significance is a judgement call; the flat count is in Hallucination flags. | 5.4 | 3.0–8.0 | 2 of 7 |
| Hallucination flags per runEvery claim in the note with no support in the consultation, counted flat: each one counts, however minor. No severity judgement. Across 11 cases. | 6.4 | 5.0–7.0 | 4 of 7 |
| Edit actions per noteMissing items, partial items and hallucination flags a clinician would need to fix, per note. | 1.93 | 1.56–2.63 | 2 of 7 |
| Clinical coverageShare of each case checklist captured by the note (partial counts as half). | 92% | 90%–94% | 3 of 7 |
| Significant transcript errors per runMis-heard words that change clinical meaning, rated blind, across 11 cases. | 31.3 | 31.0–32.0 | 6 of 7 |
| Word error ratePooled word error rate per run against the human reference. Every word counts equally, including fillers. | 17.6% | 17.4%–17.7% | 5 of 7 |
| Data per consultAverage network data per consultation, per run. | 17.0 MB | 16.9 MB–17.1 MB | 3 of 7 |
Definitions and weights are in the methodology. Compare Preve head to head ↗
Consultations
CONSULTATION