An AI-designed microbiome-targeted supplement tested for glycemic control in type 2 diabetes — with the record stating expressly that no AI runs during the trial (ENBIOSIS, NCT07622628)
A randomized study of whether a microbiome-targeted nutritional product improves blood sugar control in adults with type 2 diabetes when added to usual stable treatment, compared with placebo over 24 weeks. The product was designed using AI before the study began, but a single fixed formulation is used for everyone in the active group and AI performs no personalization during the trial.
Trial overview (primary data)
- StatusRecruiting
- ConditionsType 2 Diabetes, Diabetes Mellitus Type 2
- InterventionsDIETARY_SUPPLEMENT: Microbiome-targeted oral food supplement, DIETARY_SUPPLEMENT: Matching placebo oral supplement
- SponsorENBIOSIS BIOTECHNOLOGIES
- Target enrollment100 participants
- Period2026-06-01 〜 2027-06-30
Key points
- A 24-week randomized study of whether a microbiome-targeted nutritional product improves blood sugar control in type 2 diabetes when added to usual treatment.
- The product was designed with AI before the study, but a single fixed formulation is used and the record states no AI personalization or dose adjustment occurs during the trial.
- The main question is whether HbA1c falls more from baseline to week 24 than under placebo.
- It is a 100-participant proof-of-concept study estimating treatment signal, safety, feasibility and parameters for a future confirmatory trial.
- Of the 803 trials this site holds as of 2026-08-31, 7 touch the gut microbiome. Sponsors with more than one trial number 135, and their records total 417 (52 percent).
- AI was used only in design before the trial; during it, everyone in the active arm receives one fixed formulation.
1Stating where the AI was used
What stands out in this record is that it bounds where AI was used. The product was designed with AI before the study began. During the study, however, the same fixed formulation goes to everyone in the active group, and AI is expressly not used for individual treatment decisions, dose adjustment or personalization. Saying AI was used invites an image of personalization; here it is not that.
Used at the design stage and not at the evaluation stage — the separation is declared from the outset.
2What fixing the formulation accomplishes
The record gives no reason, but it can be read as trial design. If the product varied by participant, an effect could not be separated into what came from the product and what came from the skill of the personalization. Giving everyone the same thing fixes the object of comparison to one. A trial evaluating something AI built and a trial evaluating a mechanism in which AI keeps running are different designs. This record is the former.
3The microbiome appears in 7 of 803 trials
Sorting the bodies of the 803 medical-AI clinical trials this site holds as of 2026-08-31 by keyword, only 7 touch the gut microbiome — a thin area against 33 for chatbots, 26 for documentation and 22 for wearables. The same sponsor has also registered a separate trial in ulcerative colitis.
Across the 803 records this site holds as of 2026-08-31, 135 sponsors have registered more than one trial, and the records belonging to them number 417, or 52 percent of the total. Reading one at a time makes it impossible to notice the same party running in parallel across different diseases.
4Used in design, or running throughout
What draws attention here is that the record limits where AI was used. The separation — used at design, not used during evaluation — is declared from the outset.
Why the separation is not stated in the record, though it reads plainly as trial design. A trial evaluating something AI built and a trial evaluating a mechanism in which AI keeps running are designed differently. This is the former, and the record states explicitly that AI is not used for individual treatment decisions, dose adjustment or product personalisation.
Why it matters
Products labelled as using AI differ greatly in which step actually used it. Something designed with AI and then fixed, and something judging continuously in operation, differ in how they are evaluated and how they are treated under regulation. Stating the boundary as this record does makes it possible to read afterwards what was actually tested.
FAQ
Where is AI used?
Why one fixed formulation for everyone?
Was an effect confirmed?
Sources (primary)
Source: ClinicalTrials.gov (U.S. NIH/NLM, public domain). This site does not provide medical advice. Verify the latest and exact details with the official source. This site is not endorsed or certified by the NIH/NLM.
- ClinicalTrials.gov (study record, original)
- NCT ID: NCT07622628