The source of every sentence in your report is visible.

No. Ssemora is decision-support software: it turns the conversation into a report, flags missing information and offers reasoned suggestions. It does not suggest a single diagnosis, does not produce probabilities or risk scores, and does not pre-select any suggestion. Responsibility for the report and for clinical decisions rests with the physician who approves the report.

Every statement extracted from the recording is linked to its character range in the transcript and its second in the audio recording. When you select a sentence, you see its source and verify it by listening to the audio. A statement with no basis in the recording is not shown in yellow: if it is a model interpretation it is marked grey, and if it is missing or uncertain it is marked red.

It works through the browser. On a computer or phone you can record the encounter or upload a ready-made audio file. You can add it to the home screen of your phone or tablet and open it like an app. Unfinished recordings are kept on the device; when the connection returns you can continue, send or delete them.

Yes, up to 6 photos per encounter: findings such as wounds, skin lesions and rashes, and documents such as ECG printouts, laboratory / test results, imaging reports or discharge summaries. The photo is downscaled on the device, and location and device information is deleted on the server. Only photos that the physician has confirmed as "no face/identity visible" go to AI analysis, if the institution setting is enabled; sentences that come from a photo are shown in the report with a "From image" label. Photos are also included in the Word and PDF output.

Each institution runs on its own isolated instance. Audio recordings are deleted after 7 days by default; by institution setting this can be extended to at most 30 days. After deletion, the transcript and report remain. National ID numbers are not stored in plain text.

Text sent to the AI model is always de-identified first: names, national ID numbers, file numbers, phone numbers, email addresses, addresses, institution names and exact dates are replaced with placeholders. In addition, one of three policies is chosen for each institution: cloud allowed, de-identified data only, or local model only; with the last one, the text never leaves for an external provider. Speech recognition runs locally by default.

Compliance is a property not of software but of the processing activity of the institution acting as data controller; for this reason we make no claim such as "KVKK-compliant" or certified. Ssemora provides tools that make it easier for the institution to meet its obligations: de-identification, retention period, audit log, sharing consent and its log, and per-patient export and deletion. Even though direct identifiers are removed, the risk of indirect identifiability in very small patient populations cannot theoretically be reduced to zero; this is why the AI policy is chosen at the institution level.

Sending to the HIS is designed to go through a connector written specifically for each institution; we do not currently publish a ready-made HIS connector. Instead, external systems can upload encounters and retrieve the physician-approved report through the REST API. Nothing is sent to the HIS without physician approval.

The physician approves the report. The approved report is locked and the action is written to the audit log. Every edit before approval is kept in the version history; you can compare versions and return to an unapproved version.