AI Narration Is No Gallbladder Safety Check

In laparoscopic gallbladder surgery, targeted prompts made AI narration sound more like a surgeon’s but did not make safety checks reliable.

• Specific prompts did not consistently improve coverage of procedural checklist items.
• The trained model found the critical safety view in 83% of positive clips but correctly rejected only 19% of negative clips.
• Only six clips showed the view, making these performance estimates uncertain.

AI may help draft notes for surgeon review, but fluent narration should not be treated as proof that the safety view is present.