AI Can Audit Colonoscopy Quality From Recorded Video at Hospital Scale: AJG | August 2026

Introduction:

Colonoscopy quality is usually assessed using metrics such as ADR, but these provide limited insight into how individual procedures are actually performed. AI-CQ uses recorded colonoscopy video to automatically evaluate procedural quality after the examination.

Why was this study needed?

Manual video review is impractical at hospital scale.

ADR alone cannot capture withdrawal technique or polypectomy quality.

Automated auditing could enable continuous feedback across entire endoscopy units.

Key Findings:

AI-CQ analysed 18,597 colonoscopies from 55 attending endoscopists.

AI-derived withdrawal time correlated strongly with manual assessment (r=0.91).

AI-measured polyps per colonoscopy correlated with both ADR and serrated-lesion detection.

The system automatically assessed insertion and withdrawal times, polyp burden, and polypectomy technique.

Cold-snare use varied markedly between endoscopists, from 63% to 95%.

Unlike real-time CADe, AI-CQ functions after the procedure, enabling retrospective quality auditing and targeted feedback.

Whether AI-generated feedback ultimately improves colonoscopy quality or CRC outcomes remains unproven.

Bottom Line:

AI-CQ could transform colonoscopy quality assurance from periodic metric review to automated procedure-level auditing. The next question is whether this feedback actually improves endoscopist performance and patient outcomes.