June 28, 2026
MRI or A-I-oh no?
I used Claude Code to get a second opinion on my MRI
Man asks AI to double-check his MRI and the internet instantly calls in the adults
TLDR: A man used Claude Code to review his shoulder MRI after doubting a clinic’s fast treatment plan, and the AI disagreed sharply with the doctor’s diagnosis. Commenters were split between concern about the clinic and disbelief that anyone would trust a chatbot over a real second opinion.
A sore shoulder, a rushed clinic visit, and suddenly the internet had a brand-new medical soap opera. The writer says doctors found a serious shoulder tear on his MRI and quickly moved into treatment, but after getting uneasy about the clinic’s choices, he fed the scan files to Claude Code for a DIY AI second opinion. Plot twist: the AI reportedly came back saying the tendon looked basically intact. Yes, the human doctor saw a big tear; the chatbot saw no clear tear at all. And just like that, the comments section turned into a full-blown courtroom drama.
The strongest reaction was basically: absolutely not. One commenter flatly said they would not trust Claude with image-based diagnosis at all, while another delivered the internet’s most obvious but brutal comeback: maybe get an actual second opinion from another doctor. Ouch. A radiologist chimed in with the classic professional shrug — impossible to judge without the full scan — which only added to the tension. Meanwhile, another commenter joked that this is every doctor’s new nightmare: first patients googled their symptoms, now they arrive with chatbot receipts.
That’s where the real drama lives. Some readers were alarmed by the clinic’s treatment choices, especially after the article said one therapy may not be recommended and another injection looked suspiciously like homeopathy. But others were more worried by the bigger trend: people trusting AI to read medical images before a trained specialist. The vibe was equal parts consumer watchdog, tech panic, and “we are so cooked.”
Key Points
- •The author received an MRI-based diagnosis of a Grade III partial-thickness tear of the subscapularis tendon after experiencing right shoulder pain for several weeks.
- •After obtaining the MRI report and treatment list, the author used GPT 5.5 Pro, which flagged concerns about shockwave therapy and a Traumeel injection.
- •The MRI data was a standard DICOM export of a few hundred files totaling about 266 MB.
- •Using Opus 4.8 inside Claude Code, the author conducted an AI review of the MRI with minimal clinical context, and the model reported the tendon as intact.
- •A later arbitration run by Claude, using the human report and additional context, concluded that evidence favored no discrete partial- or full-thickness tear, instead indicating mild insertional tendinosis.