New analyses from five randomized tandem trials show AI-assisted colonoscopy consistently finds more nonadvanced adenomas — on average 0.4 extra polyps per patient. That sounds like a win for cancer prevention.
But here’s the catch: more tiny polyps = more “high-risk” labels = more surveillance scopes.
Under current US-MSTF guidelines, the number of patients pushed into high-risk follow-up jumped nearly 70% simply because AI detected additional small lesions — many unlikely to progress to cancer.
Researchers warn that AI-driven over-classification could strain colonoscopy capacity without clear outcome benefit — fueling a debate now playing out in guideline committees on both sides of the Atlantic.
👉 How big is the burden?
👉 Are guideline tweaks the solution — or a safety risk?
👉 Could AI paradoxically increase costs while improving detection?
📌 Full article breaks down the numbers — and the trade-offs GI leaders must solve next.