A journal club in the AI age is less like following a recipe and more like discovering your oven has started suggesting substitutions while quietly inventing paprika.

That, at least, is the delightful little crisis sitting inside Jacqueline Baras Shreibati's Reimagining Osler's Journal Club for the AI Age, a 2026 piece in the Journal of the American College of Cardiology (DOI: 10.1016/j.jacc.2026.01.039). The paper has no PubMed abstract, which feels almost poetic: a paper about rethinking how doctors discuss papers arrives as a citation-shaped artifact, asking the tribe to gather around and infer meaning. The humans do enjoy rituals.

The Ancient Ceremony of Reading Together

Journal clubs are old medical technology. Not "stethoscope old," but close enough to smell like library dust. Sir William Osler formalized one at McGill in 1875, partly so physicians could share access to medical periodicals they could not all afford individually. Behold: the original group subscription hack.

For more than a century, the format has stayed recognizable. A paper is chosen. A trainee presents it. Everyone asks whether the statistics are suspicious, whether the endpoint was meaningful, and whether the authors have committed the sacred sin of overclaiming. Then coffee happens.

Shreibati's title suggests a useful provocation: if medicine is entering an AI-heavy era, the journal club cannot remain only a paper-dissection potluck. It may need to become a training ground for judging algorithms, datasets, deployment claims, model drift, bias, privacy, user experience, and the strangely human art of knowing when a machine is bluffing with excellent grammar.

The New Guest at the Table Keeps Autocompleting the Menu

AI in medicine is not one thing. It is image classifiers, risk models, clinical note generators, triage tools, ambient scribes, chatbots, workflow software, and sometimes a spreadsheet wearing a lab coat. A 2024 review by Raza, Venkatesh, and Kvedar described generative AI and large language models as promising but hard to implement safely, especially because health data are messy, private, and distributed across systems like socks after laundry (DOI: 10.1038/s41746-023-00988-4).

This is where the journal club becomes more than academic calisthenics. A cardiology group discussing an AI paper now has to ask: What population trained the model? Who was left out? Was it tested outside the hospital where it was born? Does it help clinicians, or does it merely create a second inbox with better font choices? The humans call this "implementation science." An alien might call it "discovering that tools must survive contact with Tuesday morning clinic."

Tools like pdfb2.io fit into this ecosystem in a quieter way: before anyone can argue intelligently about an AI paper, they still need to handle the PDFs, extract pages, split files, and keep sensitive documents local. Even in the future, paperwork remains undefeated.

The Machine Is Clever, But Please Watch It Near the Scissors

The evidence base around medical AI is expanding fast, but it is not a parade of clean victories. A 2025 systematic review and meta-analysis in npj Digital Medicine examined 83 studies comparing generative AI with physicians on diagnostic tasks. The pooled diagnostic accuracy for AI models was 52.1%, with models performing worse than expert physicians overall (DOI: 10.1038/s41746-025-01543-z). That is not useless. It is also not "replace the doctor and let the chatbot hold the pager."

The more sensible interpretation is stranger and more interesting: AI may become a teaching partner, second reader, summarizer, question generator, or pattern spotter. It may help junior clinicians practice reasoning. It may expose uncertainty. It may also hallucinate, which is the polite technical phrase for "confidently making things up while wearing a tie."

This makes the AI-age journal club valuable. It can teach clinicians not just to read the paper, but to interrogate the entire machine-shaped ecosystem around it.

Medical Education Needs a Bigger Checklist

A 2024 BEME scoping review mapped 278 publications on AI in medical education and found uses across admissions, teaching, assessment, and clinical reasoning, while also flagging the need for ethical guidance (DOI: 10.1080/0142159X.2024.2314198). Translation from human academic dialect: everyone is experimenting, nobody has fully agreed on the house rules, and the blender is already plugged in.

Physicians' attitudes matter too. A 2025 mixed-methods study found that clinicians' views of AI were shaped more by experience and engagement than demographics, and that familiarity may reduce concerns (DOI: 10.2196/74187). This is exactly the kind of cultural problem journal clubs were built to handle. They create a room where skepticism is not sabotage. It is quality control with snacks.

Osler, But With Model Cards

The best version of Shreibati's idea is not "let AI summarize papers so doctors can stop reading." That would be like buying a treadmill and asking it to exercise on your behalf. The better version is a journal club that uses AI as both subject and tool: generate critique questions, compare interpretations, surface missing context, then make the humans defend the final judgment.

The future journal club might ask for the CONSORT diagram, the calibration curve, the subgroup performance, the prompt design, the external validation, the regulatory path, the privacy risk, and whether the model fails gracefully or falls down the stairs holding a clipboard.

A fine ritual, really. The humans gather. The machine speaks. The humans ask follow-up questions. Civilization inches forward, cautiously, caffeinated.

References

1. Shreibati JB. Reimagining Osler's Journal Club for the AI Age. Journal of the American College of Cardiology. 2026. DOI: 10.1016/j.jacc.2026.01.039

2. Raza MM, Venkatesh KP, Kvedar JC. Generative AI and large language models in health care: pathways to implementation. npj Digital Medicine. 2024;7:62. DOI: 10.1038/s41746-023-00988-4

3. Takita H, Kabata D, Walston SL, et al. A systematic review and meta-analysis of diagnostic performance comparison between generative AI and physicians. npj Digital Medicine. 2025;8:175. DOI: 10.1038/s41746-025-01543-z

4. Gordon M, Daniel M, Ajiboye A, et al. A scoping review of artificial intelligence in medical education: BEME Guide No. 84. Medical Teacher. 2024;46(4):446-470. DOI: 10.1080/0142159X.2024.2314198

5. Physicians' Attitudes Toward Artificial Intelligence in Medicine: Mixed Methods Survey and Interview Study. Journal of Medical Internet Research. 2025;27:e74187. DOI: 10.2196/74187

Disclaimer: This blog post is a simplified summary of published research for educational purposes. The accompanying illustration is artistic and does not depict actual model architectures, data, or experimental results. Always refer to the original paper for technical details.