Diagnosis Pad
Diagnosis Pad is a medical AI assistant and clinical scribe that runs entirely offline on the practitioner's own device. It transcribes consultations live, surfaces three differentials and three recommendations, then drafts the clinical note.
What is Diagnosis Pad?
Diagnosis Pad is a medical AI assistant and clinical scribe for healthcare professionals, published for iOS, macOS and Android and listed under the Medical category on both app stores. Its defining choice is where the computation happens: every inference runs offline, on the practitioner's own device, with nothing sent to the cloud for processing. The founder, Dr. Nora (MBBS FRACGP DFSRH), a practising family doctor and medical AI researcher, presents this as a direct answer to competing tools that upload complete audio recordings of consultations. The inference engine is supplied by Inferencer, and a sibling product, Treatment Pad, covers aesthetic photography.
During a session the app produces four things. It transcribes the ambient audio of the consultation, clinician and patient alike. It surfaces the three most likely differential diagnoses as the conversation progresses, framed as a guard against missing a possibility. It generates the three leading recommendations, together with the questions worth asking to reach the right treatment. Finally it drafts a clinical note summarising the transcript, either live or at the end, depending on your settings.
The vendor also makes a transparency claim that is unusual in this market: the engine is described as deterministic, without the random heuristics of typical generative chatbots, and thousands of reverse-engineered terms are exposed so that its knowledge base and reasoning can be audited. The clinical encyclopedia is advertised as holding over 6,000 terms on the website, though both app stores say over 5,000.
Two claims deserve caution. The training figures quoted, up to 9 trillion tokens and over 680,000 hours of speech, together with a "Patent Pending" mention, come with no source. And no clinical performance figure is published anywhere: no accuracy, no sensitivity, no specificity, no study.
Hardware requirements are steep, calling for an M4-class iPad or Mac, an A17 Pro iPhone on iOS 17.6, or a Snapdragon 8 Elite Android device, with a 2.8 GB download. A separate Enterprise offer runs the models on-premises with custom electronic health record integration, and image intelligence for OCR and computer vision is announced as coming.
What it does
- Transcribe a live consultation from ambient audio, capturing both clinician and patient
- Surface the three most likely differential diagnoses as the session unfolds
- Suggest the three leading recommendations, with the questions worth asking next
- Draft a clinical note summarising the consultation, either live or once it ends
- Explain any differential or recommendation through a context-aware medical terminology base
- Search an offline medical encyclopedia without leaving the app or the premises
- Apply an automatic deletion policy to everything the app stores on the device
When to use Diagnosis Pad / When not to
A quick filter to help you decide if Diagnosis Pad is the right fit.
When to use Diagnosis Pad
- Family physicians and other primary-care doctors who take a verbal history at every consultation
- Clinicians working without reliable connectivity: home visits, pre-hospital care, remote practice
- Practices bound by strict medical confidentiality, where no recording may leave the device
- Practitioners who would rather audit an AI's reasoning than accept it, using the built-in terminology base
- Healthcare organisations wanting an on-premises deployment wired into their own patient records
When not to use Diagnosis Pad
- Patients and the general public: the app is written for health professionals, not for self-diagnosis
- Anyone on older or mid-range hardware, since an M4, A17 Pro or Snapdragon 8 Elite class device is required
- Teams expecting to connect through an API or a browser: there is neither a public API nor a web version
- Buyers who need formal compliance paperwork, as no DPA, subprocessor list or device certification is published
- Practitioners working in a language other than English or Arabic, the only two the app declares
How to use Diagnosis Pad
A typical end-to-end flow, from setup to results.
- Check your hardware first: an M4-class iPad or Mac, an iPhone with an A17 Pro chip on iOS 17.6 or later, or an Android device around the Snapdragon 8 Elite
- Download the app from the App Store or Google Play through the site's download page, allowing roughly 2.8 GB
- Grant microphone access, which is requested solely to run the transcription
- Open Settings to set the automatic deletion policy and choose what is generated live rather than after the session
- Tap Start Session as the consultation begins, and speak normally
- Let the app transcribe the ambient audio of both clinician and patient
- Follow the three differentials and three recommendations as they appear during the session
- Tap any differential or recommendation to open the context-aware terminology base and see the reasoning behind it
- Pause and resume the session as the consultation requires
- Collect the clinical note at the end, edit it from the Essentials tier, or share it through your operating system
Pros & Cons
Pros
- Offline architecture: nothing leaves the device for processing, which is decisive under medical confidentiality
- Works with no connectivity at all, from home visits to pre-hospital care and poorly covered areas
- The free tier is genuinely usable, keeping every analysis feature and limiting only session length and history
- Entry pricing is low for the clinical scribe market, at $9.99 per month, with a $29.99 ceiling
- A rare transparency effort: a deterministic engine, detailed reasoning, and an auditable terminology base
- AES-256 encryption at rest plus an automatic deletion policy the practitioner configures
- An on-premises Enterprise route exists for organisations that cannot externalise anything
Cons
- No clinical performance figure is published at all: the only numbers given are training volumes, without source or study
- HIPAA, GDPR, NHS and PIPEDA compliance appears in the compliance page title, while the body claims only an intention
- No DPA, no subprocessor list, no DPO, no Article 27 representative and no stated retention period
- The publisher's legal identity and address are absent from the site, which carries no legal notice at all
- Hardware demands are severe and recent, with a 2.8 GB app that rules out most of the installed device base
- No web version, no public API and no developer documentation; a single support email is the only contact channel
- Communication is inconsistent: 6,000 versus 5,000 encyclopedia terms, and structured data that omits Android, quotes only a free price and links a missing image
Pricing & Plans
A permanent free plan is available under the name Basic, with no expiry date. The cheapest paid entry point is the Essentials plan at 9.99 USD per month, rising to 29.99 USD per month for Professional. Prices are displayed in US dollars and converted to the local currency by Apple or Google at checkout. The Enterprise offer is not publicly priced.
- on-device private AI
- unlimited processing
- transcriptions
- differentials
- guidance
- clinical notes
- detailed reasoning and session sharing
- but sessions are time limited and no history is kept
- everything in Basic
- plus no session time limits and the ability to edit transcripts
- still no session history
- everything in Essentials
- plus the ability to review past sessions
- private on-premises AI running on the customer's own servers
- with custom electronic health record integration
Data, GDPR & hosting
A consolidated view of how Diagnosis Pad handles your data.
GDPR overview
Implementation is thin, and the wording matters. GDPR is named only in the title tag and meta description of the compliance page; the body of that page claims nothing more than an intention, namely that Diagnosis Pad "aims to be the strongest compliant AI application with international health data policies". No legal basis is cited, no data-subject right is described, and no procedure is offered to exercise one. There is no DPO, no Article 27 representative, no data processing agreement, no subprocessor list and no stated retention period. What the vendor does provide is architectural: processing stays on the device, local storage is AES-256 encrypted, and nothing reaches its servers. One inconsistency deserves noting, as the app claims to use no third-party tools while the website itself runs Google Analytics.
Who owns the data?
The terms and conditions are silent on data ownership: no clause assigns rights over your recordings, transcripts or notes to the publisher, and none grants it a licence. In practice the question is settled by architecture rather than by contract, since sessions are stored locally and the vendor states it receives nothing. Two routes still move data outward, and both are started by you: the share function, which exports through the operating system to a destination you choose, and the feedback function, which sends session data to the publisher for analysis after you have reviewed and edited it. Device backups to iCloud or Google Backup may also copy data into your own account.
Reuse rights
Nothing in the terms restricts what you do with the transcripts, differentials and notes the app produces: no permission is required and no licence is reserved by the publisher. Sharing is a built-in function available from the free tier onwards, and it is you who picks the destination through the operating system. Editing a transcript unlocks at the Essentials tier, and reviewing past sessions only at Professional; below that, no session history is kept at all. The single contractual limit is behavioural rather than proprietary: you must use the app lawfully and responsibly, and you remain responsible for verifying anything it produces before acting on it.
Data retention & training
Hosting summary
There is effectively no vendor-side hosting to describe. Inference and storage both happen on the practitioner's device, which is why no hosting country or region appears anywhere on the site: it follows from the architecture rather than from an omission. Local storage is encrypted with AES-256. Data can nevertheless leave the device by two routes that lead to your own accounts rather than the vendor's: iCloud or Google Backup if device backups are enabled, and the share function you trigger yourself. The only data reaching the publisher is what you send through the feedback function, plus anonymised crash logs if your operating system settings allow them. Subscription checks with Apple or Google happen periodically and carry no personal data. The website itself is a different matter: it sits behind Cloudflare, on an anycast address that identifies no real hosting country, and it runs Google Analytics. No processing jurisdiction is declared, though the app publisher is established in the United Kingdom according to the Google Play listing.
Things to keep in mind
Risks and trade-offs to weigh before adopting Diagnosis Pad.
- A wrong or incomplete differential can steer a clinical decision, and the terms place that responsibility entirely on the practitioner, with the software supplied as is and no warranty
- No clinical performance data exists to calibrate trust: no accuracy, sensitivity or specificity figure, no study and no external evaluation is published anywhere
- No certification or medical-device status is claimed, so the tool sits outside the regulatory framework a buyer might assume applies to diagnostic support
- Automation bias is a real risk: a continuous stream of three differentials can anchor the clinician's thinking and quietly impoverish the history-taking it was meant to support
- A deterministic engine is reproducible, not correct, and the two are easy to confuse when the same answer keeps coming back
- The app records the ambient audio of a consultation, and patient consent remains the practitioner's duty since the tool does not manage it
- Health data can still leave the device without anyone deciding so: iCloud or Google Backup may copy it to a third-party account, and the feedback function sends real session data to the publisher with no stated retention period
Setup & Integrations
Technical difficulty
Setup is consumer-grade: download from the App Store or Google Play, grant microphone access, and optionally set the deletion policy and choose what is generated live. No server, no account creation and no configuration is mentioned. The real obstacle is hardware rather than technique, since an M4-class Apple device, an A17 Pro iPhone or a Snapdragon 8 Elite Android is required, along with a 2.8 GB download. The Enterprise on-premises option is an entirely different proposition, involving a server deployment and a custom health record integration project.
Deployment
Apps stores
Integrations
Supported languages
Behind Diagnosis Pad
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Diagnosis Pad.
Frequently asked questions
Does any consultation data leave my device?
Does Diagnosis Pad make a diagnosis?
Is there a free plan, and what does it leave out?
What does the paid version cost?
Which devices can actually run it?
Is there a web version or an API?
Is my data used to train the models?
How long is data kept?
Is Diagnosis Pad a certified medical device?
Who is behind the tool, and how do I reach support?
Should you pick Diagnosis Pad?
Diagnosis Pad makes an unusually clean promise: medical inference that never leaves the consulting room, verifiable by architecture rather than by policy. For a clinician who cannot let patient audio reach a third-party server, that is worth more than any assurance page, and the permanent free tier makes the promise testable at no cost. Pricing is modest for the clinical scribe market, and the transparency effort around a deterministic engine, detailed reasoning and an auditable terminology base is genuinely uncommon.
The reservations are equally clear, and in a medical context they matter. No clinical performance data is published at all: the only figures offered are training volumes, quoted without source, alongside a "Patent Pending" mention. No certification or regulatory status is claimed, which is coherent with a product presented as supplementary information rather than a device, but it leaves a buyer with nothing external to lean on. The compliance page is the weakest link: its title announces HIPAA, GDPR, NHS and PIPEDA, while its body only says the app aims to be compliant, and none of the usual artefacts, a DPA, a subprocessor list, a named data protection contact or a retention period, exists to back it.
The publisher is equally hard to see. The site carries no legal notice, no company name beyond the brand and no postal address; the identity surfaces only through the app store listings. Add steep hardware requirements and a young footprint, with the domain registered at the end of 2024, and the picture is of a promising, privacy-first tool that still has its evidence and its paperwork to produce. Try it on the free tier, and verify the compliance claims yourself before it touches real patient data.
- Choosing a selection results in a full page refresh.
- Opens in a new window.