**Live *A. muciniphila* isolate — confidence: weak.** One 9-completer arm (Depommier) was milder than pasteurized. One 58-person Chinese T2D RCT (Zhang 2025) was **null overall**; any weight/HbA1c movement was in a low-baseline subgroup and partly within-group. Colonisation seems to fail when the gut already has a lot of Akkermansia. That is a hypothesis with two small human tests, not a prescribing rule.
**Pasteurized *A. muciniphila* MucT — confidence: weak-to-moderate (context-specific).** Depommier 2019 is real: **verified** 3-month RCT, pasteurized 10^10/day, insulin sensitivity **+28.6%** and insulin **−34%** vs placebo, cholesterol **−8.7%**, LPS down, weight **−2.3 kg (P = 0.091)**. It is also tiny, imbalanced at baseline, and author-patented. The larger insulin-sensitivity RCT (**n = 142, 30 billion/day**) **failed its Matsuda primary**. The 2026 Maastricht trial is the strongest weight number: **~2 kg less regain** over 24 weeks after an 8-week diet (company-funded). Net: pasteurized MucT is the only form with repeated human RCTs; ITT metabolic effects are **small or absent** except after a diet and in people with **low baseline Akkermansia**.