AI Radiology Ranking: xAID Tops Its Own Teleradiology List

A vendor ranks the field and puts itself first

On September 30, 2026, xAID, a New York company that sells AI radiology software for teleradiology groups, published what it called the Top 7 AI Radiology Software Platforms for Teleradiology Groups in 2026. The list is meant to identify the tools best suited to high-volume reading that spans many facilities and time zones.1 xAID ranked itself first. Aidoc came second, followed by Viz.ai, Rad AI, Qure.ai, Annalise.ai and Lunit.2 The release went out through a press-release wire and was reposted almost word for word on a string of local-news and lifestyle sites.135 That spread makes it look like more outlets covered it than actually did.

Read it as marketing. It is not an independent evaluation. The company says it built the ranking from public information about each platform's regulatory clearances, clinical focus and deployment scope. It also says it scored the platforms on three things: how well they fit multi-site reading, how many findings they are cleared for, and how widely they are deployed.3 The scoring method and any weightings were not published. The one quality that decides first place, being built for distributed reading from day one, is the quality xAID claims for itself.2 Vendor-written "top platform" lists are common in this market. An Indian cloud-PACS company put itself at the top of an 11-item teleradiology software ranking.9 A teleradiology provider did the same in a top-10 list of service companies.10

The list still deserves a look, because its argument says something real about where radiology AI is heading.

The argument: workflow fit over algorithm count

xAID's main claim is that most radiology AI was built for one hospital's PACS and radiology department, then adapted for outside groups reading remotely.1 Teleradiology groups have different needs, according to the company. They handle shift changes and escalations from one facility to another. Each client hospital sets its own turnaround requirements. A single radiologist may cover ten or more sites in one overnight shift.13 A company spokesperson said groups are switching vendors once they see that triage tools designed for in-house departments don't fit this kind of work.3 The release offers no data on how often that switching happens.

The description of the competitors is accurate as far as it goes. The release calls Aidoc's suite of FDA-cleared triage algorithms one of the broadest in the category. It credits Viz.ai with the first FDA-cleared AI triage platform and notes its focus on stroke-network alerts.2 It describes Rad AI as a report-drafting tool that works alongside triage products rather than replacing them. It describes Qure.ai as strong in chest X-ray and CT in resource-constrained settings.2 Annalise.ai trained its algorithms on case archives from a large Australian radiology network. Lunit, listed on KOSDAQ, concentrates on chest and breast imaging.2

The release says all seven platforms share three traits:

• FDA clearance, or an equivalent regulatory pathway, for at least part of their algorithms
• A specific clinical or operational focus
• Design or adaptation for reading volume beyond a single hospital7

It names xAID, Aidoc and Viz.ai among the entries using cleared algorithms rather than research-only tools.2 The release does not list any of xAID's FDA clearances, and nothing else published about the company backs up that claim. Buyers should ask for clearance numbers before accepting it.

Where the ranking and other reporting differ

Other coverage paints a somewhat different picture of the competitors. The release presents Aidoc as having one of the widest cleared portfolios.2 An industry comparison credits Aidoc with 31 or more clearances across roughly 2,000 hospitals. The same comparison credits Viz.ai with more than 50 clearances at over 1,700 hospitals.18 Aidoc also received what was described as the first FDA clearance for a clinical AI device built on a foundation model. That device is a single body-CT triage product covering 14 conditions, cleared in January 2026.18 Lunit, ranked seventh, received a clearance for its digital breast tomosynthesis software in March 2026.15

The Rad AI description has a twist. xAID calls Rad AI a reporting companion, not a triage substitute.2 Another analysis says Rad AI's reporting tools are workflow software rather than a regulated diagnostic device, and estimates they reach about half of US imaging practices.18 Yet xAID's own marketing promotes AI-generated CT report drafts. It says a radiologist who reads 40 studies a day unaided can read 55 to 65 with those drafts.25 So xAID competes with Rad AI more directly than the ranking suggests.

The regulatory backdrop: lots of clearances, little scrutiny

The ranking's emphasis on FDA clearance makes sense given the numbers. The FDA's AI-enabled device list, updated in September with decisions through June 29, 2026, has 1,614 entries. Radiology accounts for 1,230 of them, about 76 percent.19 The FDA authorized 89 such devices in the second quarter of 2026 and 92 in the first, with radiology making up 74 to 75 percent each quarter.19 Annual totals have climbed from six authorizations in 2015 to 335 in 2025.14 The single most common product category is radiological computer-assisted detection and diagnosis. Computer-assisted triage, the category most relevant to teleradiology, is also common.11

The count needs context. The FDA itself says the list is not a complete inventory. It identifies devices through AI-related terms in public summaries.11 It also mixes standalone software with scanners and other hardware that have AI built in.12 The large equipment makers lead the count: GE HealthCare had 134 authorizations as of the latest update, partly through companies it acquired.19 Nearly all AI devices reach the market through the 510(k) pathway. That route relies on showing "substantial equivalence" to an existing device rather than on large clinical trials.1416 An FDA clearance proves a product passed that bar. It does not prove the product performs better in a busy overnight reading room.

For anyone looking ahead, the more important regulatory issue is generative AI. As of March 2026, no authorized device used generative AI or large language models.20 In June, the FDA gave breakthrough designation to two generative tools from Cognita and Aidoc that draft chest X-ray reports. Breakthrough designation is not marketing authorization.11 xAID itself notes that the FDA has not cleared any AI system to issue a final radiology report on its own. Every authorized device is assistive and works under a radiologist's review.26 For now, report drafting stays a support tool, and responsibility for the final read stays with a licensed physician.

Why teleradiology is where the buying happens

Pressure on the workforce explains why vendors are focusing on teleradiology groups. Health systems increasingly rely on teleradiology partners and pay premium labor costs to keep coverage, as imaging volume grows faster than staff.21 One projection puts the US shortfall at 17,000 to 42,000 radiologists, pathologists and psychiatrists by 2033.22 Another analysis expects radiologist supply to grow about 25.7 percent between 2023 and 2055. Imaging use could grow nearly as fast over the same period, so the gap would persist.30

Money is tight too. The 2026 Medicare Physician Fee Schedule added a permanent 2.5 percent "efficiency adjustment" to work RVUs, the units that set physician payment. It assumes technology has already made radiologists more productive.22 The shortage is worst in rural markets and on overnight and weekend shifts.28 Those are exactly the hours a single teleradiologist may spend covering many sites. North America made up about 41 percent of the teleradiology market in 2026.27

AI vendors are targeting large reading groups for that reason. The biggest US teleradiology networks already run dozens of triage models across thousands of facilities.910

Deployment risks the ranking leaves out

The list covers fit and clearances. It does not cover the governance problems that shape how hospitals actually deploy these tools. Models need to be validated on local scanners and patient populations and monitored over time, because performance can drop across sites.22 One analysis cites evidence that some vision-language models underdiagnose marginalized groups at higher rates than radiologists do.22 That risk grows when one tool serves dozens of client hospitals with different patients.

Liability is also unsettled. In a simulated study, about 75 percent of jurors found a radiologist at fault for missing a hemorrhage the AI had flagged, when the radiologist read the scan only once, after the flag. That fell to 53 percent when the radiologist read the image both before and after seeing the AI's output.30 Training lags as well. In a 2026 American Medical Association survey, more than a quarter of physicians said they had received no AI training.23

Bottom line

xAID's ranking is a vendor promoting itself, and it should be treated that way. Its core argument still holds: radiology AI is now judged on how it works in multi-site, overnight workflows, not only on how many algorithms a vendor has cleared. The underlying facts back that up. Radiology dominates FDA AI authorizations, regulators still require a radiologist to sign off on every read, and reimbursement and staffing pressure push hospitals toward remote reading. The winners in teleradiology will need both a broad set of verifiable clearances and workflow software that holds up across many hospitals and shifts. A self-published top-seven list doesn't show which vendors have both. Buyers who want to know should ask for clearance numbers, results validated at their own sites, and liability-aware reading protocols.