diff --git a/domains/health/no-peer-reviewed-evidence-of-durable-physician-upskilling-from-ai-exposure-as-of-mid-2026.md b/domains/health/no-peer-reviewed-evidence-of-durable-physician-upskilling-from-ai-exposure-as-of-mid-2026.md
index 7584e2421..c615c4a69 100644
--- a/domains/health/no-peer-reviewed-evidence-of-durable-physician-upskilling-from-ai-exposure-as-of-mid-2026.md
+++ b/domains/health/no-peer-reviewed-evidence-of-durable-physician-upskilling-from-ai-exposure-as-of-mid-2026.md
@@ -16,3 +16,10 @@ related: ["human-in-the-loop-clinical-ai-degrades-to-worse-than-ai-alone-because
 # No peer-reviewed evidence of durable physician upskilling from AI exposure as of mid-2026
 
 The Heudel et al. scoping review examined literature through August 2025 across colonoscopy, radiology, pathology, and cytology. Authors conclude: 'empirical studies consistently demonstrate that AI can inadvertently impair physicians' performance.' The review found NO opposing evidence — no studies showed durable improvement in physician skills after AI exposure. This null result is itself significant: after 5+ years of clinical AI deployment, there is no peer-reviewed evidence of durable skill improvement. The authors searched for counter-evidence and found none. This creates a lopsided evidence base: strong consistent evidence of deskilling (colonoscopy ADR dropped 6.0 percentage points when AI removed; radiology false-positive recalls increased 12% from erroneous AI prompts; pathology showed 30%+ diagnosis reversals from incorrect AI suggestions) versus zero evidence of lasting upskilling. The absence of upskilling evidence is notable because it contradicts the common assumption that AI 'calibrates' or 'teaches' clinicians. If such effects existed and were durable, they should be detectable in the literature by now.
+
+
+## Supporting Evidence
+
+**Source:** Savardi et al., Insights into Imaging, PMC11780016, Jan 2025
+
+Savardi et al. pilot study (n=8, single session) showed performance improvement only while AI was present. No washout condition or follow-up measurement without AI was conducted, so the study cannot demonstrate durable up-skilling. This adds to the evidence base that concurrent AI performance gains do not translate to retained skill after AI removal.