Explore results →SCRIBE RESULTS / 22 SEPTEMBER 2026
Averages across 5 graded runs of 10 consultations using stored transcripts. Local preview — Hanah notes regenerated on 11 October 2026 with GPT-6.1 Sol, low reasoning: 50 notes across 10 consultations, five stored transcript variants each. Two consultations await new Hanah notes. WER and transcription evidence retain their original capture dates; other scribes retain their existing results. New notes were graded blind by Sonnet 5.5 medium, followed by an unblinded agent audit; provisional, without independent clinician validation. Hanah timings are estimates: prompted-note wait uses measured API generation duration + 2 seconds readiness allowance; time to a finalised note adds estimated review and correction time. Both exclude transcription finalisation and app overhead, pending a live rerun..
| Measure | Average | Run range | Rank |
|---|---|---|---|
| Time to a finalised noteWait from pressing stop to the graded note, plus review time: how long until the note can be finalised if it is reviewed straight away. | 1m 33s | 1m 30s–1m 35s | 1 of 6 |
| Notes ready to finalise as writtenShare of notes that captured every fact in the consultation's inventory in full with no hallucination flags, so nothing needed editing. | 30% | 20%–40% | 1 of 6 |
| 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. | 4% | 3%–5% | 1 of 6 |
| 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 24s | 1m 20s–1m 26s | 1 of 6 |
| Worst note, any runThe single slowest note to review and fix across every case and run. | 2m 36s | — | 1 of 6 |
| Time saved per noteTime to write the note yourself (8.1 min, the median in Apathy et al. 2023) minus review time. | 6m 42s | 6m 40s–6m 46s | 1 of 6 |
| Faster than writing it yourselfHow much quicker reviewing and fixing a note is than writing it yourself (8.1 min, the median in Apathy et al. 2023). Negative means slower. | 83% | 82%–83% | 1 of 6 |
| Errors per note (Weiner et al. standard)Relevant facts the note missed or captured only partly, plus every hallucination flag (a wrong detail counts twice, as in Weiner et al.), against an inventory of every fact said that relates to the patient's conditions, symptoms or signs (Weiner et al. 2020). Average per note, across 12 cases. | 1.22 | 1.10–1.30 | 1 of 6 |
| Note errors (Weiner et al. standard)Relevant facts the note missed or captured only partly, plus every hallucination flag (a wrong detail counts twice, as in Weiner et al.), against an inventory of every fact said that relates to the patient's conditions, symptoms or signs (Weiner et al. 2020). Total per run, across 12 cases. The flat count is in Hallucination flags. | 12.2 | 11.0–13.0 | 1 of 6 |
| Errors reaching the record, per 100 notes (estimate)Note errors (Weiner et al. standard) per 100 notes, times 70%: in Biro et al. (2025), GPs reviewing AI drafts left each planted error uncorrected 65–75% of the time. An estimate of what a typical review would let through, across 12 cases. | 85 | 77–91 | 1 of 6 |
| Fewer errors than doctor-written notesCompared with the 5.0 errors per note Weiner et al. (2020) found in physicians' notes checked against recordings of the visit. Our test counts more strictly than theirs, so this favours the doctors. 100% means no errors; below 0% means more errors than the doctors. Across 11 cases. | 76% | 74%–78% | 1 of 6 |
| Hallucination flags per note (Weiner et al. standard)Every clinical element the note documents that the consultation does not support, including wrong values, sides, doses and timings: Weiner et al.’s (2020) errors of commission. Paraphrase and clinical terms for what was said do not count. Average per note, across 12 cases. | 0.36 | 0.10–0.60 | 1 of 6 |
| Edit actions per noteMissing items, partial items and hallucination flags a clinician would need to fix, per note. | 1.22 | 1.10–1.30 | 1 of 6 |
| Clinical coverageShare of each consultation's fact inventory captured by the note (partial counts as half). | 98% | 97%–99% | 1 of 6 |
| Significant transcript errors per runMis-heard words that change clinical meaning, rated blind, across 12 cases. | 8.6 | 8.0–10.0 | 1 of 6 |
| Word error ratePooled word error rate per run against the human reference. Every word counts equally, including fillers. | 12.8% | 12.4%–13.7% | 1 of 6 |
| Stop to prompted noteMedian time from pressing stop to the note written to our instruction, per run. | 10s | 10s–11s | 1 of 6 |
| Data per consultAverage network data per consultation, per run. | 9.1 MB | 9.1 MB–9.3 MB | 1 of 6 |
| Avg RAM (stress test)Physical memory footprint (macOS `footprint`, matching Activity Monitor's Memory column) sampled every minute during one continuous 50-minute stress-test consultation — separate from the 12-case suite above. Every scribe recorded this visit at the same time on one machine, so absolute values reflect shared contention between them, not isolated real-world usage. Compare scribes to each other here, not to a device's spec sheet. | 1103 MB | 1067 MB–1178 MB | 2 of 6 |
| Avg CPU (stress test)CPU used per minute during the same continuous 50-minute stress-test consultation, measured as cumulative CPU-seconds consumed divided by wall-clock time (can exceed 100% when a scribe uses more than one core). Every scribe ran concurrently on one machine. | 28.6% | 26.2%–34.0% | 1 of 6 |
Definitions and weights are in the methodology. Compare Hanah head to head ↗
4 November 2026
The team is incredibly proactive and always available. They reached out and resolved my issue before I even had a chance to send a support email. I’ve honestly never experienced such an exceptional level of customer service!