See the output

Judge the report,
not the pitch.

Below is the structure of what a Cognitrion assessment actually produces — the same eight sections a clinician reviews, including the parts most software would rather not show you.

The report, section by section

This is a synthetic example. Every field shown is one the system genuinely produces; nothing here is illustrative filler.

Cognitrion™ Clinical Assessment Report

Sample output — for clinical review

§1 · Patient identification

Patient ID
CGN-SYN-7AF6-1BD9
Age
59
Assessed
12 August 2026
Assessor
Cognitrion™ v0.9 (research build)

§2 · Assessment profile

Task administered
Cookie Theft picture description (BDAE standard stimulus)
Pipeline stages completed
8 of 8
Assessment language
English

§3 · Linguistic analysis

Content words
37
Unique words
27
Fluency
Non-fluent
Disfluencies
10 (repetitions + partial attempts)
Long pauses
7 (timing-derived)
Transcription confidence
0.90

§4 · Classification

Transcortical Motor Aphasia
Severity: moderate Confidence: 90%

§5 · Clinical rationale — including what it could not decide alone

"Non-fluent, effortful output with relatively preserved auditory comprehension. Differential: connected speech alone cannot separate the minimal / non-fluent cluster (severe Broca · global · transcortical motor · mixed transcortical). Administer a repetition subtest, then comprehension — repetition impaired with good comprehension indicates Broca; repetition preserved with good comprehension indicates transcortical motor. Resolved: repetition preserved (accuracy 100%) and comprehension good → transcortical motor aphasia."

§6 · Evidence base invoked

Kertesz (2006), WAB-R Examiner's Manual — subtest thresholds for syndrome classification · Goodglass, Kaplan & Barresi (2001), BDAE-3 — Cookie Theft stimulus and conversational-speech rating · the classical Geschwind–Wernicke model.

§7 · Known limitations — printed on every report

  • Normative comparisons use simulated normative data generated by published statistical methods; the system is not yet validated against an external real-patient cohort.
  • Short-duration partial-word markers include known false positives on common function words — they are candidates for clinician review, not assertions.
  • Speech recognition strips filler tokens ("um", "uh") from the transcript text; the timing is recovered as pause markers instead.
  • Per-stimulus reaction-time anchoring is not yet wired end to end.

§8 · Clinician sign-off

☐ I would sign my name to this report as clinically defensible.

☐ I would not sign — the gaps I require addressed before sign-off are: ______________

Signature Date

Three things to look for

These are the parts clinicians tell us they check first.

§5

It refuses to guess

When spontaneous speech cannot separate two syndromes, the report says so and names the subtest that would decide it — rather than presenting one answer with false certainty.

§4

Confidence is explicit

Every classification carries a number. Below the configured threshold, the software will not finalise without a clinician's decision — approve, override, or defer.

§7

The limitations are printed, not hidden

Including the one that matters most: this has not yet been validated against an external real-patient cohort. You should know that before you read the classification, not after.

Read one for yourself

Request a de-identified sample report as a PDF, or ask for a walkthrough on a live assessment — intake to signed report, about thirty minutes.

Request the sample report → Email us instead
What we will not do: we will not show you a report from a real patient, and we will not quote accuracy figures from a study we have not run. What we can show you is the system working on synthetic cases with known answers, our full test evidence, and this format — reviewed by a clinician co-inventor. See the clinical disclaimer for our validation status, stated plainly.
For families Request a demo →