---
title: "Will AI replace doctors? What the trials and the bankruptcy record show (2026)"
description: "Will AI replace doctors? US physician jobs are projected to grow, the AI trials worked through physicians, and the AI-first clinic went bankrupt."
url: "https://theinternetninja.com/blog/will-ai-replace-doctors-what-the-trials-show-2026/"
updated: "2026-10-31"
---
# Will AI replace doctors? What the trials and the bankruptcy record show (2026)
By Maneesh Behera, Head of AI Deployment. Published 2026-10-31. The Internet Ninja, analysis.
> The record on doctors does not show AI taking the job. The US outlook projects physician jobs growing, the best-measured clinical AI gains came with a physician acting on the output, and the company that sold AI as a doctor substitute went bankrupt.
## Case files this argument rests on
- [AI-ECG mortality alert: a 15,965-patient randomized trial in Taiwan cut 90-day all-cause mortality from 4.3% to 3.6% (HR 0.83)](https://theinternetninja.com/stories/ai-ecg-alert-rct-taiwan-cuts-90-day-mortality-3-6-vs-4-3pct-nature-medicine-2024/)
- [An OpenAI reasoning model reanalyzed 376 unsolved rare-disease cases at Boston Children's. Experts confirmed 18 new diagnoses.](https://theinternetninja.com/stories/boston-childrens-manton-openai-o3-deep-research-rare-disease-reanalysis-18-diagnoses-376-cases-nejm-ai-2026/)
- [Babylon Health, the AI-first primary-care company, collapsed into bankruptcy in 2023 and sold its UK service for $620,000](https://theinternetninja.com/stories/babylon-health-ai-first-primary-care-spac-collapse-chapter-7-bankruptcy-2023/)
If you are a doctor, or training to be one, you have heard the forecast. The models pass the exams, read the scans and write the notes, so what is left for the physician?
There is a record to test that against, and it points the other way. The two best-measured AI results in TIN's files worked through a physician, and the company that sold AI as a doctor substitute is in a bankruptcy court.
Start with the jobs outlook. US employment of physicians and surgeons "is projected to grow 4 percent from 2025 to 2035" ([source](https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm)). That is growth, not decline.
## What "will AI replace doctors" actually asks
Will AI replace doctors asks whether software can take over the work a patient trusts a physician with: deciding what is wrong, choosing what to do about it, and answering for the result.
Getting the answer right on a test is the narrow part, and models already do it. The question is whether anyone has measured AI doing the rest without a physician. This post tests that against three TIN case files.
## Will doctors be replaced by AI in 10 years?
Not on the official US outlook, which does not name AI as a driver. It expects physician demand to rise because patients are getting older and sicker.
The Bureau of Labor Statistics counts 862,800 physician and surgeon jobs in 2025 and projects 894,100 in 2035 ([source](https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm)). It projects about 22,100 openings a year over the decade ([source](https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm)).
The stated reason: "An increasing number of older adults, who are more likely than younger people to experience health problems and to need complex care, is expected to drive overall employment growth" ([source](https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm)). These are forecasts, not measurements, and TIN did not check them.
## Does AI work better with a doctor or without one?
With one, on the strongest trial TIN holds. In Taiwan, an AI read routine ECGs and alerted the treating physician to high-risk patients. The physician still decided what care to give.
The trial randomized 39 physicians and 15,965 patients ([source](https://pubmed.ncbi.nlm.nih.gov/38684860/)). Within 90 days, 3.6% of patients in the alert arm died, against 4.3% in control, a hazard ratio of 0.83 ([source](https://pubmed.ncbi.nlm.nih.gov/38684860/)).
The benefit sat in the patients the model flagged high risk, with a hazard ratio of 0.69 ([source](https://pubmed.ncbi.nlm.nih.gov/38684860/)). It is one trial at one hospital system, not yet replicated elsewhere. The AI did not treat anyone. It changed who got a physician's attention, and that is where the lives were saved.
## Can AI diagnose what doctors missed?
Sometimes, and physicians still made the diagnosis. At Boston Children's, a reasoning model reanalyzed rare-disease cases that specialists had already failed to solve.
Across 376 unsolved cases, physicians established 18 new diagnoses, an additional yield of 4.8% ([source](https://europepmc.org/article/MED/42529462)). Each one went through expert variant review and clinical lab confirmation before it counted.
Seven of the 18 were "rediscoveries," answers a team elsewhere already had but had not shared ([source](https://www.nbcnews.com/tech/innovation/ai-boston-childrens-hospital-diagnose-rare-diseases-kids-openai-rcna350387)). The study was retrospective. The model widened the search, and the clinicians decided what was true.
## Will AI replace primary care doctors?
The best-known attempt says no. Babylon Health built a primary-care business on a symptom-checker chatbot and sold it as a substitute for seeing a doctor.
At a 2018 launch, Babylon said the average pass mark for doctors on a GP exam was 72 percent and its AI scored 82 percent ([source](https://www.cnbc.com/2018/06/28/babylon-claims-its-ai-can-diagnose-patients-better-than-doctors.html)). Correspondence in The Lancet found the study did "not offer convincing evidence" the system could beat doctors "in any realistic situation" ([source](https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(18)32819-8/fulltext)).
In August 2023 Babylon filed for Chapter 7 in Delaware ([source](https://www.courtlistener.com/docket/67680026/babylon-inc/)). Its UK telemedicine service was "sold for just $620,000" ([source](https://hlth.com/insights/news/bankrupt-digital-health-company-babylon-sells-u-k-assets-for-just-620-000-2023-09-20)). An exam score was never the job.
## Will AI replace nurses and doctors?
Not on the US outlook for either. Registered nurse employment "is projected to grow 6 percent from 2025 to 2035," with about 180,800 openings a year ([source](https://www.bls.gov/ooh/healthcare/registered-nurses.htm)).
TIN holds no case file that measures AI doing nursing work, so this post makes no claim about it beyond the forecast.
## The proof
TIN's case file on the Taiwan trial checks every figure against the paper and independent outlets, and names its limit as one unreplicated trial: [the AI-ECG mortality alert RCT](https://theinternetninja.com/stories/ai-ecg-alert-rct-taiwan-cuts-90-day-mortality-3-6-vs-4-3pct-nature-medicine-2024/).
TIN's case file on Boston Children's reads the four-cohort table from the peer-reviewed abstract and prints what the number does not mean: [18 new diagnoses in 376 unsolved cases](https://theinternetninja.com/stories/boston-childrens-manton-openai-o3-deep-research-rare-disease-reanalysis-18-diagnoses-376-cases-nejm-ai-2026/).
TIN's case file on Babylon traces the claim, the SPAC listing and the court filings to primary records: [the AI-first primary-care collapse](https://theinternetninja.com/stories/babylon-health-ai-first-primary-care-spac-collapse-chapter-7-bankruptcy-2023/).
Whether clinical AI improves patient outcomes at all is argued separately in [does clinical AI improve outcomes](https://theinternetninja.com/blog/does-clinical-ai-improve-outcomes-what-the-peer-reviewed-record-shows-2026/). None of these files counts physician jobs lost to AI, and this post claims no such number.
## What each record shows
| Record | What it tested | Result | What it means for doctors |
|---|---|---|---|
| BLS outlook 2025 to 2035 | Projected US physician jobs | Growth of 4 percent | Aging and chronic illness drive it, not AI |
| Taiwan AI-ECG trial | AI alerts to physicians | 90-day deaths 3.6% vs 4.3% | Gain came through the physician |
| Boston Children's | Model reanalysis of unsolved cases | 18 of 376 newly diagnosed | Clinicians confirmed every one |
| Babylon Health | AI chatbot as doctor substitute | Chapter 7, UK unit sold for $620,000 | The replacement bet failed |
## The bottom line
The forecast that AI replaces doctors treats medicine as answering questions, and models are already good at answers. The record says the measured gains arrive when a model changes what a physician looks at, and the physician acts and answers for it. The one company that removed the physician from the pitch is the one in bankruptcy court. The useful test for any clinical AI is not its exam score. It is whose decision it improves, and who carries the outcome when it is wrong.
## Sources
1. U.S. Bureau of Labor Statistics, "Occupational Outlook Handbook: Physicians and Surgeons," accessed 2026-10-31. https://www.bls.gov/ooh/healthcare/physicians-and-surgeons.htm
2. U.S. Bureau of Labor Statistics, "Occupational Outlook Handbook: Registered Nurses," accessed 2026-10-31. https://www.bls.gov/ooh/healthcare/registered-nurses.htm
3. Nature Medicine, via PubMed, "AI-enabled electrocardiography alert intervention and all-cause mortality: a pragmatic randomized clinical trial," 2024-04. https://pubmed.ncbi.nlm.nih.gov/38684860/
4. NEJM AI, via Europe PMC, Jaech, Cheatham et al., "LLM-Assisted Reanalysis of Unsolved Rare Disease Genomes Increases Diagnostic Yield," 2026-06-25. https://europepmc.org/article/MED/42529462
5. NBC News, Jared Perlo, "AI helped diagnose 18 children whose rare diseases had stumped doctors," 2026-06-18. https://www.nbcnews.com/tech/innovation/ai-boston-childrens-hospital-diagnose-rare-diseases-kids-openai-rcna350387
6. CNBC, "Babylon claims its AI can diagnose patients better than doctors, doctors call that 'dubious'," 2018-06-28. https://www.cnbc.com/2018/06/28/babylon-claims-its-ai-can-diagnose-patients-better-than-doctors.html
7. The Lancet, Fraser, Coiera and Wong, "Safety of patient-facing digital symptom checkers," 2018-11-06. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(18)32819-8/fulltext
8. CourtListener / RECAP, Babylon Inc., case 23-11127, Chapter 7, U.S. Bankruptcy Court, District of Delaware, filed 2023-08-09. https://www.courtlistener.com/docket/67680026/babylon-inc/
9. Forbes, syndicated on HLTH, "Bankrupt Digital Health Company Babylon Sells U.K. Assets For Just $620,000," 2023-09-20. https://hlth.com/insights/news/bankrupt-digital-health-company-babylon-sells-u-k-assets-for-just-620-000-2023-09-20
## Questions
**Will AI replace doctors?**
AI has not replaced doctors on the record so far. The US Bureau of Labor Statistics projects physician and surgeon employment to grow 4 percent from 2025 to 2035, and the best-measured clinical AI result, a 15,965-patient trial in Taiwan, worked by alerting physicians who then decided the care.
**Will doctors be replaced by AI in 10 years?**
Doctors are not expected to be replaced by AI in the next 10 years on the official US outlook. The Bureau of Labor Statistics projects about 22,100 physician and surgeon openings a year through 2035, driven by an aging population and chronic illness, and does not name AI as a factor.
**Will AI replace primary care doctors?**
AI did not replace primary care doctors in the best-known attempt. Babylon Health marketed a symptom checker as diagnosing better than doctors, a claim The Lancet said lacked convincing evidence, then filed for Chapter 7 in 2023 and sold its UK service for $620,000.
**Will AI replace nurses and doctors?**
AI is not projected to replace nurses and doctors on the US outlook. The Bureau of Labor Statistics projects registered nurse employment to grow 6 percent from 2025 to 2035, and TIN has no case file that measures AI doing nursing work.
This post is analysis, not proof. The evidence lives in the case files: https://theinternetninja.com/stories/