PulseAI, a Belfast medical technology company, said on 5 November 2026 that the US FDA has granted 510(k) clearance for its AI arrhythmia analysis model for wearables. The release gives no decision date or K number. The claim that inconclusive smartwatch ECGs fall from 20% to near-zero does not appear in it.
The company's real evidence is a 92% reduction in inconclusive readings against the Apple Watch's own algorithm, in a 296-participant study published in the journal Heart Rhythm. That study has an Irish link, because it was run at a teaching hospital in Ireland. Its authors include PulseAI directors and employees, and independent replication is limited.
- The company: PulseAI of Belfast, founded in 2020, announced the clearance in a press release dated 5 November 2026.
- Not yet confirmed by FDA data: the openFDA 510(k) dataset, last updated on 26 October 2026, shows no PulseAI record. The decision date and K number are not disclosed.
- The real figure: a 92% reduction in inconclusive results against the Apple Watch's native algorithm, in 296 participants (Heart Rhythm, peer reviewed).
- The headline figure: "20%" and "near-zero" are not in the company's release. "Near-zero" appears in a Tech Times headline.
- Ireland: the study's registered site is Beacon Hospital Research Institute, Dublin. No CE mark or Irish availability has been stated.
What did PulseAI announce?
The release, distributed through PR Newswire with a Belfast dateline of 5 November 2026, says the company has received "U.S. Food and Drug Administration (FDA) 510(k) clearance for its AI-powered arrhythmia analysis model for wearables". It describes a cloud service. Recordings are "submitted for automated analysis in the cloud, with structured results returned to the connected application for clinical review".
The release does not name a product, give a formal indication for use or list compatible devices. It says only that the technology "is designed for integration with compatible ECG devices and software platforms". So this is not a new feature that smartwatch owners can switch on today, and no device partner has been named.
Dr Alan Kennedy, CEO and founder of PulseAI, said: "With FDA clearance, our focus now is bringing this capability into the devices and platforms in use by millions every day." Professor Paul Bartholomew, Vice Chancellor of Ulster University, called the move from university spin-in to clearance "a significant achievement for the company". Ulster University is an investor through Innovation Ulster Limited, so that is not an independent view.
Has the FDA clearance been confirmed?
Not independently, at the time of writing. We searched openFDA, the FDA's public data service, for PulseAI as applicant and for all recent 510(k) decisions from Belfast. There was no match. That dataset was last updated on 26 October 2026, and it runs to decisions in late October, so a decision made or posted after that would not show. The filing could also sit under a different applicant name. We did not complete a search of the FDA's live 510(k) database.
A January 2026 report on the company's seed round said the money would support preparation for a future 510(k) submission, so any decision is likely to be recent. This is our reading, not a stated fact. A 510(k) is not an approval. The FDA describes it as a premarket submission showing that a device is "substantially equivalent" to a legally marketed device.
Where does "20% to near-zero" come from?
The phrase comes from a Tech Times article of 5 November, whose headline says "From 20% to Near-Zero". Its body says the rate fell from "roughly 20 percent to under one percent". The company's release contains none of these figures. It gives only the 92% reduction, with no starting rate.
Securities.io reports the study's first-attempt figures as 17.6% falling to 1.4%. Those two numbers do work out at a 92% reduction, but 1.4% is not under 1%, and 17.6% is not 20%. We could not read the full paper to check them, because the publisher's page was closed to us. The PubMed abstract confirms the 92% and not the underlying rates.
A baseline near 20% is not invented. Across other studies, device algorithms left between 12% and 26% of recordings inconclusive. But a fair summary of the PulseAI evidence is about 1% to 2% or lower, from a baseline of roughly 12% to 18%, in hospital patients.
| Source | Figure |
|---|---|
| PulseAI release, 5 November 2026 | 92% reduction against Apple Watch's native algorithm; no rates given |
| Heart Rhythm abstract, 296 participants | 92% reduction; no rates given |
| Securities.io, reporting the same study | 17.6% to 1.4% on first attempt |
| Tech Times, 5 November 2026 | "roughly 20 percent to under one percent" |
| Cardiovascular Digital Health Journal 2023, 206 patients, 824 ECGs, four devices | 16% with manufacturer algorithms, 1.2% with PulseAI |
| MedTech Dive 2023, Heart Rhythm conference data, 91 patients, 273 ECGs | 12% with Apple's ECG App, under 1% with PulseAI |
| BASEL Wearable Study, JACC Clinical Electrophysiology 2023, 201 patients (no PulseAI arm) | 17% to 26% across five devices; 18% for Apple Watch 6 |
How strong is the evidence?
The main paper, by Doggart and colleagues, is peer reviewed and appears in Heart Rhythm, published online on 7 November 2025. It is a prospective observational study, not a randomised trial. The same Apple Watch recordings were read by the native algorithm and by the AI. Its central point is about method: leaving inconclusive results out of the calculation flatters a device.
| Method and reporting framework | Sensitivity / specificity |
|---|---|
| Apple Watch, inconclusive results excluded | 96.1% / 97.9% |
| Apple Watch, inconclusive counted as incorrect | 78.1% / 81.0% |
| Apple Watch, up to three attempts allowed | 92.2% / 91.0% |
| PulseAI, inconclusive counted as incorrect | 98.4% / 96.1% |
| PulseAI, up to three attempts allowed | 98.4% / 96.6% |
Three cautions apply. First, the authors are not independent. The paper's disclosures say Dr Peter Doggart and Dr Alan Kennedy are directors and shareholders of PulseAI, two authors are employees and two are advisers. PulseAI Ltd is the lead sponsor on the trial registration. The disclosures say Apple Watch readings were verified by physicians blinded to PulseAI results, and that the protocol was preregistered as NCT05668377.
Second, these were hospital participants, so the results may not carry over to healthy people using a watch at home. Third, the most independent test is more modest. The 2023 University Hospital Basel study, funded by the Swiss Heart Foundation and the University of Basel, still had two PulseAI-linked co-authors. It found AI sensitivity of 88% and specificity of 95%, against 98% and 97% for cardiologists reading the same tracings. In that study, sensitivity improved significantly only for the Fitbit.
The release's subtitle also claims "up to 10x performance improvement". The body does not explain what was measured. It is a company claim, and we have not been able to tie it to a specific result beyond the inconclusive rate.
What does this mean for Irish readers?
The Irish link is real but narrow. The abstract says the study ran at a teaching hospital in Ireland, the registration lists Beacon Hospital Research Institute in Dublin, and one co-author is affiliated with Beacon Hospital. Other co-authors are based at Ulster University, Craigavon Area Hospital and Belfast Health and Social Care Trust.
FDA clearance has no legal effect in Europe. The UK government says EU device regulations have applied in EU member states and in Northern Ireland, and that "CE marking is required for the Northern Ireland market". Neither the release nor the company pages we read mention a CE mark, so availability in Ireland, north or south, is not known. For background, see our reports on Irish medtech and EU device rules and on Cork's Neurobell and its AI neonatal seizure monitor.
Frequently asked questions
Can I use this on my smartwatch now?
No consumer route has been announced. No compatible devices are named, and the release describes results going to a connected application for clinical review.
Does it remove inconclusive results altogether?
No. Published figures for the PulseAI model are about 2% or lower, in hospital settings. "Near-zero" is a headline description and not a study finding.
I got an inconclusive reading on my watch. What should I do?
This article is not medical advice. Talk to your GP or pharmacist about any symptoms or readings that worry you, and nobody should change treatment without their doctor or pharmacist.
Outlook
The next document to watch is the FDA's own database entry and decision summary, which should show the K number, decision date, indication for use and the evidence reviewed. After that, the questions are which device maker signs up first and whether PulseAI seeks a CE mark. Priya Life Science will update this report when the FDA record is available or the company confirms these details.
Sources
- PulseAI via PR Newswire (copy on Yahoo Finance), PulseAI Receives U.S. FDA 510(k) Clearance of AI-Powered Arrhythmia Analysis Model for Wearables, 5 November 2026
- Doggart P et al, Heart Rhythm, Accounting for inconclusive results and repeated testing, published online 7 November 2025 (PubMed abstract and disclosures)
- ClinicalTrials.gov, NCT05668377: Automated Diagnostic Performance of Smartwatch ECG for Arrhythmia Detection Using the PulseAI Neural Network, last updated 29 December 2022
- Weidlich S et al, Cardiovascular Digital Health Journal, Reducing the burden of inconclusive smart device single-lead ECG tracings via a novel artificial intelligence algorithm, 27 December 2023
- Mannhart D et al, JACC: Clinical Electrophysiology, Clinical Validation of 5 Direct-to-Consumer Wearable Smart Devices to Detect Atrial Fibrillation: BASEL Wearable Study, February 2023
- FDA, openFDA device 510(k) dataset, data last updated 26 October 2026, searched 6 November 2026
- FDA, 510(k) Premarket Notification database, definition of a 510(k), accessed 6 November 2026
- GOV.UK, Regulation of medical devices in Northern Ireland, updated 11 June 2026
- Tech Times, FDA Clears Belfast AI That Cuts Smartwatch ECG Inconclusives From 20% to Near-Zero, 5 November 2026
- Securities.io, Belfast's PulseAI Gains 510(k) Clearance for Wearable ECG AI, 5 November 2026
- MedTech Dive, report on the PulseAI algorithm and Apple Watch AFib detection, 23 May 2023
- Pulse 2.0, PulseAI: Seed Funding Secured To Advance AI-Driven ECG Diagnostics, 12 January 2026
Cover image: AI-generated illustration.
