Help & Knowledge Base
Everything you need to document a case, read a report, and get around the app.
Getting Started
1. Sign in. Use your work email and a password, or "Continue with Google" / "Continue with LinkedIn." All three land you in the same account system.
2. Complete your profile (first time only). Every new account — however you signed in — is asked for a first name, last name, and medical registration number before it can reach any patient data. This is a one-time step.
3. Read the testing-phase warning. Above the patient form you'll see a persistent amber notice: this build has not completed a full compliance review, so use fictional or de-identified test cases only — no real patient names, dates of birth, or other identifying details.
Three Ways to Document a Case
Every field in the intake form — HPI, past history, medications, examination, investigations — can be filled in whichever of these ways suits you, and you can mix methods within the same case.
Type Directly
Dictate a Single Field (Push-to-Talk)
Each field has a small microphone icon next to its label. Click it once to start recording, click again to stop — this is push-to-talk, not continuous listening, so nothing is captured before you click or after you stop.
Once you stop, the audio is transcribed and the field fills in automatically. A status line under the field shows the raw transcript so you can check it against what you actually said before moving on.
Dictate the Whole Encounter
The "Dictate Whole Encounter" button records a single continuous session — useful for dictating the whole history in one go rather than field by field, or for recording an actual doctor–patient conversation.
When the recording includes more than one speaker, the transcript is automatically split and labeled by speaker (e.g. "Doctor:" / "Patient:") so you can see who said what before anything is filled in. If the audio doesn't support that split, you still get a single clean transcript — nothing is lost, it's just not speaker-separated.
Either way, stopping the recording opens the Encounter Review screen before anything is committed to the form.
Upload an Audio Recording
Already have a recording — from a separate dictaphone, a phone voice memo, or any other source? Upload it instead of dictating live. Accepted formats: MP3, WAV, M4A, WebM, and OGG.
The uploaded file goes through the exact same transcription and speaker-labeling pipeline as live "Dictate Whole Encounter" recordings, and opens the same Encounter Review screen afterward.
Reviewing a Dictated or Uploaded Encounter
Neither "Dictate Whole Encounter" nor an audio upload fills in the form directly. Both open a review screen first, showing the transcript (speaker-labeled where available) alongside any red flags the transcription step noticed automatically.
Use this screen to correct anything the transcription got wrong before it's distributed into the actual clinical fields. Nothing reaches the intake form, and nothing is analyzed, until you confirm from this screen.
Reading Your Report
Click "Generate Report" once the intake fields are filled in (however you got them there). The report is built from several distinct sections — understanding what each one is checking makes it much faster to review:
Reasoning Trace
A step-by-step account of how the report was built: syndrome recognition, framework evaluation, red-flag extraction. This is what makes the tool a transparent reasoning trail rather than a black box — you can see the working, not just the answer.
Triage / NEWS2 & Red Flags
An automatically calculated NEWS2 early-warning score from the vitals you entered, plus any red flags and the Emergency Action Plan if the case looks acute.
Worst-First Safety Net
A separate, deterministic check — not dependent on the AI getting everything right — that screens specifically for the most dangerous diagnosis a given presentation could represent, even if it isn't the most likely one.
Ranked Differential Diagnoses
The list of possible diagnoses, ordered by likelihood given the presentation, each with the guideline or reasoning it's based on. In Educational Mode, the top-ranked differential also shows a "Further Reading" list of real citations pulled live from PubMed.
Drug & Allergy Safety Check
Cross-checks any suggested medications against the current medications and allergies you entered — flagged separately from the main reasoning, again so it doesn't depend on the AI remembering to check.
Suggested Investigations & Management
Recommended next investigations and a management plan, evaluated against the clinical guideline sets applicable to the region(s) you've selected.
AI Voice Briefing (Read Aloud)
Click "Read Aloud" on a generated report to have it read back to you — critical alerts, triage category, ranked differentials, investigations, management, and the disclaimer, in that order. It's a manual button, not automatic — nothing plays until you click it.
It's a play/pause toggle: click once to start, click again to pause and resume where you left off. Clinical abbreviations are expanded for speech (e.g. "IV" is read as "intravenous," not "four" — except after a word like "Stage" or "Class," where "IV" correctly means the Roman numeral).
Loading a different report, generating a new one, or resetting the form automatically stops whatever is currently playing — so you never end up hearing a previous patient's report by mistake.
Auto Documentation
Below the report, the Auto Documentation section drafts a referral letter, a SOAP note, or a discharge summary directly from the encounter — each editable in place before you copy, download, or sign it (there's a signature pad built in for referral letters).
You can add extra document types, duplicate a card to draft a second version, or delete ones you don't need for a given case.
History & Calibration
History lists your past analyses, scoped to your account only — nobody else can see them. Clicking a past record reloads the full report; deleting one is permanent and asks for confirmation first, since it can't be undone.
Decision Calibration lets you record what you actually decided against what the AI suggested (agreed or overridden), building a simple agreement-rate view over time — useful both as a personal audit trail and for spotting where the tool tends to get overridden.
Account & Security
Sign in with email/password, Google, or LinkedIn. First-time sign-ins of any kind are asked to confirm a first name, last name, and medical registration number before reaching any patient data.
Forgot your password? Use the "Forgot password?" link on the sign-in screen — it emails you a reset link.
Every session is tied to your account, encrypted at rest, and every view or deletion is logged to an audit trail. See the About page for the fuller picture on data handling and current limitations.
FAQ & Troubleshooting
Can I use real patient names or identifiable details?
My dictation didn't pick up a second speaker — is that broken?
Read Aloud is still reading an old report.
Where do the "Further Reading" citations come from?
Still stuck, or found something that doesn't match this page? support@aethelq.com