Start here →
ⓘ Educational information only — not medical advice. Nothing on this site diagnoses or treats any condition. In an emergency, call your local emergency number.
Medically reviewed · Peer-reviewed sources

AI in healthcare, evidence-checked.

Independent, medically reviewed guides to what AI actually does in medicine today — the real trial results, the real limits, and the questions you should ask before trusting any AI health tool.

✓ Peer-reviewed studies only ✓ Medically reviewed ✓ Clear about what we don't know
Screening workflow — illustrationAssisted review
Reading workload
−44.3% (MASAI RCT)
Human readers
100% retained
Evidence
RCT· cited
1614
FDA-authorized AI devices (Jun 2026)
20
% More cancers detected — MASAI interim RCT
34
% Correct diagnosis listed first — older checkers
100
% of claims cited
Where AI actually helps today

Three places AI is genuinely useful in medicine

Not hype — these are the areas where peer-reviewed evidence and regulators agree AI already earns its keep.

🩻

Medical imaging

AI assists radiologists by flagging suspicious findings in mammograms, chest X-rays and retina scans. In trials, AI plus a human reader caught more cancers than standard reading alone.

→ AI Radiology guide · trial data cited
📋

Documentation & admin

Ambient scribes turn doctor-patient conversations into clinical notes, giving clinicians back hours spent on paperwork — one of the most measurable wins so far.

→ AI in Healthcare guide
🧬

Research & drug discovery

Tools like AlphaFold predict protein structures, and AI-designed drug candidates have reached clinical trials. Promising — but a long way from a pill on your shelf.

→ AI in Healthcare guide
Medical review board

Who checks our work

Every guide is reviewed by a clinician before publication, against the peer-reviewed literature — never against vendor marketing.

AK
Dr. Anita Kumar
Radiologist · Reviewer

Checks every imaging claim against the underlying trials and their limitations.

JB
Dr. James Bennett
General practitioner · Reviewer

Reviews patient-facing claims, red flags and safety wording on every page.

MR
Maya Roberts, MSc
Health data scientist · Editor

Verifies statistics and study methods, and checks that sources say what we claim.

Placeholder reviewers — replace names and credentials before public launch. See About for our full methodology.

Quick answers

Frequently asked questions

Can AI diagnose cancer?+
AI is not approved to diagnose cancer on its own. In screening, AI assists radiologists — for example by flagging suspicious mammograms — and trials show it can improve detection when combined with a human reader. Diagnosis always remains with a qualified clinician. See the radiology guide for the trial evidence.
How accurate are AI symptom checkers?+
Accuracy varies widely. Older studies found checkers listed the correct diagnosis first 34% of the time and gave appropriate triage advice 57% of the time; newer AI tools are improving but still make mistakes. They are not a substitute for a clinician, especially with serious or worsening symptoms. See the symptom checker guide.
How many FDA-approved AI medical devices are there?+
Through the end of June 2026: 1,614 AI-enabled devices, with 1,230 (76%) in radiology. Most were cleared via the 510(k) pathway rather than new clinical trials. The healthcare overview explains what authorization actually means.
Will AI replace doctors?+
No credible evidence suggests that. The realistic picture is augmentation: AI handles repetitive tasks such as screening and documentation while clinicians keep responsibility for diagnosis and care.
Can I trust ChatGPT with medical questions?+
Not for decisions about your health. General AI assistants can be useful for background information, but they can also produce confident, wrong answers. Always verify health information with a clinician or an official health authority. See the symptom checker guide.

Know what the evidence says

Ten minutes of reading now beats trusting an app you don't understand. Start with the overview.

Read the overview