Medical study places without entrance exam restrictions — in Europe and worldwide

Personally guided by Marcel Kloos, founder of Medschool Experts, from your first consultation to the start of your career.
Book a free consultation
  • The right university for you — and a secure placement
  • A study place abroad with no entrance exam, internationally recognised
  • Support with enrolment, preparation and finding accommodation
  • Support with your university transfer
  • Guidance all the way to your first job as a doctor
Quick survey · anonymous
What was the main reason your studies came to a halt?
One question, ten seconds. Your answers feed directly into our advising — and into the next analysis here in the magazine.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Anonymous. We use your answers internally only, to improve our advising — never shared with third parties.
Magazine

AI in healthcare: AI discharge summaries outscore doctor-written ones in London study

AI in healthcare: AI discharge summaries outscore doctor-written ones in London study

20.11.2026

8 min read Lesezeit

Short answer

AI in healthcare: in a London study of 1,294 admissions, 89.7% of AI-drafted discharge summaries were signed off vs 67.1% written by doctors.

AI in healthcare passed a practical test this week. On 17 November 2026, researchers at Imperial College London and Imperial College Healthcare NHS Trust reported that 89.7% of AI-drafted discharge summaries were good enough to sign off with no or minor edits. For summaries written by doctors, the figure was 67.1%.

Twelve resident doctors rated summaries from 1,294 real admissions without knowing which version came from the AI. For anyone studying medicine now, the result is a preview of the job. Future doctors will likely spend less time writing discharge summaries and more time checking them.

Short answer: In a blinded London study published on 17 November 2026, AI-drafted discharge summaries were signed off with no or minor edits in 89.7% of cases, against 67.1% for doctor-written ones. The AI still missed key information in 23.5% of drafts, so close review by a doctor remains essential.

Last reviewed: November 2026.

What did the 2026 Imperial study on AI discharge summaries find?

On 17 November 2026, researchers at Imperial College London and Imperial College Healthcare NHS Trust reported that 89.7% of AI-drafted hospital discharge summaries could be signed off with no or minor edits, compared with 67.1% of summaries written by doctors. Twelve resident doctors rated summaries from 1,294 real admissions at five London hospitals without knowing which version came from the AI.

The admissions date from 1 September 2022 to 30 September 2023. Of the 1,294 summaries, 649 were the original doctor-written versions and 645 were produced by GPT-4o mini, using a two-stage prompt that clinicians helped design. Some cases were reviewed twice, giving 1,442 evaluations in total. Most stays were short, with a median of three days.

The gap held across the board. Half of the AI drafts (50.8%) needed no changes at all, against a quarter (25.5%) of the doctor-written ones. AI summaries were also more likely to score highly on accuracy, completeness, tone and clarity, with adjusted differences of 19.4 to 29.1 percentage points. Perfect scores in every domain went to 18.8% of AI summaries and 8.2% of human ones.

That is a large margin.

How safe are AI-generated discharge summaries in NHS hospitals?

In the November 2026 London study, AI-generated discharge summaries were less often rated as potentially harmful than doctor-written ones: 26.2% versus 45.1% for the clinical summary section, and 26.8% versus 51.5% for the plan and requested actions. Hallucinations appeared in 6.0% of AI drafts and 6.9% of human summaries.

Omissions tell a similar story. Key clinical information was missing from 23.5% of AI summaries and from 39.8% of the doctor-written ones. Read those numbers twice, because they cut both ways. The AI did better than tired humans under time pressure, yet almost one in four drafts still left out something that mattered.

The researchers concluded that the AI summaries met the current standard of care only with close clinician oversight. Review stays mandatory.

AI in healthcare: AI discharge summaries study, London 2026AI in healthcare discharge summary study, November 2026: 1,294 admissions rated by 12 resident doctors. 89.7% of AI summaries signed off with no or minor edits versus 67.1% of doctor-written. 50.8% of AI drafts needed no changes versus 25.5%. Key information omitted in 23.5% versus 39.8%. Hallucinations in 6.0% versus 6.9%.AI IN HEALTHCARE · NOVEMBER 2026AI discharge summariesvs doctor-written1,294real admissions, 5 London hospitals,rated blind by 12 resident doctors89.7%AI drafts signed off with no orminor edits (doctors: 67.1%)50.8%AI drafts needing no changesat all (doctors: 25.5%)23.5%AI drafts missing key clinicalinformation (doctors: 39.8%)6.0%AI drafts with a hallucination(doctor-written: 6.9%)medschool-experts.com
AI in healthcare: how AI-drafted discharge summaries compared with doctor-written ones in a 2026 London study of 1,294 admissions.

How did resident doctors co-design the AI tool at Imperial College Healthcare?

Resident doctors at Imperial College Healthcare NHS Trust helped build the discharge summary tool before the large 2026 comparison. In an earlier development phase published in July 2026, four doctors in their first and second postgraduate years refined the prompt over three rounds, and the share of drafts they would sign rose from 53% to 82%, then to 89% on a test set.

The earlier phase was small. It covered 52 inpatient stays, with nine held back for testing. The work ran inside the trust's secure data environment, so patient records never left NHS control. The five hospitals involved serve more than 1.3 million patients a year.

Imperial has built this kind of involvement into its research culture. Its Resident Doctor Research Group, set up in 2025, has seen 25 resident doctors named as authors on published research since launch. No coding background is required to join.

What does AI in healthcare mean for junior doctors' workload in 2026?

For resident doctors in the UK in 2026, AI in healthcare is moving the discharge summary from a writing task to a checking task. The Imperial study measured quality, not time saved, so the workload gain is still unproven. What changes is the job itself: reviewing a draft against the full record instead of assembling it from scratch.

Anyone who has worked a ward knows who writes discharge summaries. It is usually the most junior doctor on the team, often late in a shift, piecing together results and medication changes from scattered notes. That matters on the ward. A good first draft could give that doctor time back. A bad one that gets waved through could send a GP the wrong plan.

The UK government is moving in the same direction. A discharge tool is in development at Chelsea and Westminster NHS Trust, to run on the NHS Federated Data Platform, and every draft must be fully reviewed by the responsible clinician before use. Ambient AI scribes raise related questions, which we covered in our piece on AI scribes and the NHS workforce debate.

How are other countries using AI documentation in healthcare in 2026?

AI documentation tools are spreading fast outside the UK in 2026. In the United States, 81% of physicians used AI in their practice in 2026, up from 38% in 2023. Canada enrolled 12,019 primary care clinicians in a national AI scribe programme, and 40% of Australian GPs reported using AI scribes in November 2025.

CountryAI documentation in practiceKey figure
United KingdomAI-drafted discharge summaries tested on 1,294 admissions in five London hospitals (published 17 November 2026)89.7% signed off with no or minor edits vs 67.1% for doctor-written
United StatesPhysician use of AI, including clinical documentation (March 2026)81% of physicians use AI, up from 38% in 2023
CanadaNational AI scribe programme for primary care (results June 2026)12,019 clinicians enrolled, 1.11 hours a week saved in working hours
AustraliaGP use of AI scribes (November 2025)40% of GPs, up from 22% in August 2024

Adoption is broad. The Canadian results give the clearest picture of time. Canada Health Infoway reported in June 2026 that more than 7,400 of the enrolled clinicians were active users by February 2026. They saved 1.11 hours a week during working hours and nearly an hour a week after hours. Same-day note completion rose from 55.9% to nearly 75%.

American doctors are using AI widely but warily. In the same March 2026 survey, 88% said they were concerned about losing skills, especially doctors with ten years or less in practice. We looked at that tension in our article on the AI training gap for doctors.

{{cta}}

Which skills do medical students need for AI in healthcare?

Medical students training in 2026 need the clinical judgement to check an AI draft against the real patient record. The London study shows why: even the better AI summaries left out key information in 23.5% of cases. Spotting what is missing requires a doctor who knows what should be there.

In practice, that means a few concrete abilities:

  • Reading a draft against source data, including results, medication charts and ward notes
  • Medication reconciliation, since treatment and medication changes are a known weak spot in discharge summaries
  • Clear clinical writing, so you can correct a draft fast and in plain language for the GP
  • Knowing how language models fail, from confident hallucinations to silent omissions
  • Basic data governance, including why secure data environments exist

The fundamentals still count. Anatomy, physiology, pharmacology and history taking matter more, because they are what lets you see an error. If you want a sense of how quickly AI is reaching patients as well as doctors, read our report on AI in medicine and a 200,000-person health study.

Students who study medicine in Europe in English-taught programmes train in hospitals that are digitising quickly, and they rotate through wards where these tools are arriving. You can compare programmes across the largest selection of universities worldwide on our university finder.

The author's view

When I started dental medicine at Medical University Sofia in 2015, at 24, AI tools that draft clinical documents did not exist in everyday care. Five years after my graduation in 2021, a London hospital trust has tested them on more than 1,200 real admissions. That pace still surprises me.

I have helped many hundreds of students start medical school abroad, and I read this study as good news for them. The doctors who will do well are the ones who know their medicine well enough to catch what a draft leaves out.

That skill is built in lectures, on wards and through years of practice, wherever you study.

Summary

  • On 17 November 2026, researchers in London reported a blinded comparison of AI-drafted and doctor-written discharge summaries from 1,294 admissions at five hospitals.
  • Twelve resident doctors signed off 89.7% of AI drafts with no or minor edits, against 67.1% of doctor-written summaries.
  • AI drafts missed key information less often (23.5% vs 39.8%) and had a similar hallucination rate (6.0% vs 6.9%).
  • Resident doctors co-designed the tool, and close clinician oversight remains a condition of safe use.
  • The United States, Canada and Australia are adopting AI documentation quickly, with 81% of US physicians using AI in 2026.

Conclusion

The London results suggest the discharge summary, one of the most familiar tasks in a junior doctor's week, is about to change shape. The first draft will increasingly come from software, and the doctor who signs it remains responsible for every line.

For students, the best preparation is a strong clinical foundation plus the habit of checking every draft against the record. If you are planning to study medicine in Europe and want to talk through programmes and next steps, Marcel or another experienced consultant on our team will help you in a free consultation. Book your consultation here.

Frequently asked questions about AI discharge summaries

What is a discharge summary?

A discharge summary is the document a hospital doctor writes when a patient leaves hospital. It records the diagnosis, key test results, treatment, medication changes and the follow-up plan, and it goes to the patient's GP. In the UK it is usually written by the most junior doctor on the ward team, which is one reason AI drafting tools are being tested on it first.

Were AI discharge summaries better than doctors' in the 2026 London study?

In the study published on 17 November 2026, resident doctors signed off 89.7% of AI-drafted discharge summaries with no or minor edits, compared with 67.1% of doctor-written ones. AI drafts also scored higher for accuracy, completeness, tone and clarity. The researchers stressed that the AI summaries met the standard of care only with close clinician oversight.

Do AI discharge summaries make mistakes?

Yes. In the November 2026 London study, 6.0% of AI-drafted discharge summaries contained a hallucination, close to the 6.9% seen in doctor-written ones. Key clinical information was missing from 23.5% of AI drafts, compared with 39.8% of human summaries. Every AI draft therefore still needs to be checked by a doctor against the full patient record before it is sent.

Will AI replace junior doctors?

Nothing in the 2026 London study on AI discharge summaries points to replacing junior doctors. The AI produced drafts, and resident doctors reviewed, corrected and signed them. The work shifts from writing discharge summaries to checking them, which still requires clinical knowledge, judgement and accountability. In the United States, 81% of physicians already use AI in their practice alongside their clinical work.

Which AI skills should medical students learn in 2026?

Medical students in 2026 benefit most from strong clinical fundamentals combined with the habit of verifying AI output. That means reading a draft against results and medication charts, spotting omissions, understanding how language models hallucinate, and writing clear corrections for GPs. These skills build on anatomy, pharmacology and clinical reasoning rather than replacing them.

Your question · answer by email
Still have a question?
Write it here. I'll respond by email and, if it helps others, include it anonymously in the Q&A for this article.
Thanks — your question has arrived. You will get an answer by email, usually within two working days.
That didn't work, unfortunately. Please try again or contact us using the contact form.
We use your email address solely to reply to your question. No sharing with third parties, no newsletter.

What students say about Medschool Experts

Over 500 placements at 25+ partner universities across 10+ countries. See how our students experienced their path into medical school abroad in our student stories.

Read student stories

About the author

Marcel Kloos

Founder and owner of Medschool Experts, federally recognised dentist

Marcel Kloos is the founder and owner of Medschool Experts, a placement agency for medicine, dentistry, veterinary medicine and pharmacy degrees abroad. Since 2021 the team has secured more than 500 study places at 25+ partner universities in Europe and worldwide.

Study medicine abroad: Marcel's own path

Born in Stuttgart, Marcel started his degree late: at 24 he enrolled at the Medical University of Sofia, Faculty of Dental Medicine, and graduated in 2021 from the English-language programme. He went through the same steps his clients face today — nostrification of school certificates, entrance interviews, finding a flat in Sofia and later the recognition of a foreign degree. He is a federally recognised dentist.

What Marcel does at Medschool Experts

Today Marcel lives in Bern and runs Medschool Experts full-time. He visits every partner university in person, negotiates placement agreements directly with faculties and deans, and leads the free consultations in which applicants learn at which university they can start next semester. His YouTube channel on studying medicine abroad has helped thousands of applicants compare universities, costs and recognition.

Topics Marcel writes about

  • Study medicine abroad independent of your high school diploma grades: universities, costs and admission
  • Study dentistry, veterinary medicine or pharmacy in Europe in English
  • Tuition fees and living costs at medical schools in Europe and worldwide
  • Recognition of foreign medical degrees in the EU, Switzerland, the UK and the US
  • Transferring between medical schools and returning home into a higher semester

All articles by Marcel Kloos are based on first-hand visits to the universities and conversations with current students. Book a free consultation to discuss your own path.

More articles worth reading

Medical School Tuition Statistics 2026: Median Fees in 38 Countries

Countries

9 min read Lesezeit

Medical School Tuition Statistics 2026: Median Fees in 38 Countries

Marcel Kloos

Marcel Kloos

22.11.2026

Read article
Oldest Universities Where You Can Study Medicine: Founding Dates from 1088

Universities

8 min read Lesezeit

Oldest Universities Where You Can Study Medicine: Founding Dates from 1088

Marcel Kloos

Marcel Kloos

22.11.2026

Read article
English-Taught Medical Schools 2026: 455 Programmes in 49 Countries

Medical Studies Abroad

7 min read Lesezeit

English-Taught Medical Schools 2026: 455 Programmes in 49 Countries

Marcel Kloos

Marcel Kloos

22.11.2026

Read article
View all articles
Portrait photo of Ahmad, Medschool Experts client

Ahmad

Portrait photo of Sabrina, Medschool Experts client

Sabrina

Portrait photo of Rami, Medschool Experts client

Rami

Portrait photo of Noah, Medschool Experts client

Noah

500+ students placed.
Now it's your turn!

No daydreaming, no empty promises. Just a clear plan — for the start of your medical degree.

Book a free consultation
  • 40 minutes, no obligation