Claims that the gut microbiome plays a causal role in autism have become increasingly popular in research, the media, and funding initiatives. This hypothesis rests on what can be described as a “three-legged stool” of evidence: observational studies in humans, experiments in animals, and clinical trials using microbiome-based treatments. This talk critically examines each line of evidence and argues that, collectively, they fail to support this popular narrative.

Human studies comparing microbiomes between autistic and non-autistic individuals typically rely on small samples and a wide range of analytic choices, increasing the risk of false-positive findings. Consequently, the reported differences across studies are not only inconsistent but frequently contradict each other. More recently, larger studies using machine learning have also failed to identify a clear microbiome pattern associated with autism. Instead, they found that autism status explains only a very small proportion of differences in microbiome composition, far less than the natural variation seen between individuals or even within the same person over time.

Animal models fare no better and present major issues in terms of validity and interpretation. These studies ask: if the gut microbiome contributes to autism-related traits, can changing the microbiome produce or improve those traits in animals? However, it remains unclear whether these models accurately capture the risk factors involved in autism in humans, and whether the behaviours observed in mice meaningfully reflect features of autism in humans. This limits the conclusions that can be drawn from these models.

Clinical trials represent the most direct test of whether altering the gut microbiome can improve outcomes in autistic people. However, many studies testing interventions such as fecal microbiota transplantation and probiotics rely on weak designs that cannot determine whether treatments are responsible for observed improvements. This literature is dominated by open-label (where participants and researchers know who receives the treatment), single-arm (without a comparison group) studies using small samples, often involving only a few dozen participants. The few properly randomized, double-blind, placebo-controlled trials have not shown meaningful treatment effects, with observed improvements consistent with natural developmental changes rather than the treatment itself.

These lines of inquiry therefore failed to triangulate toward convincing evidence that the gut microbiome plays a causal role in autism. They also illustrate a broader pattern of research waste, in which repeated investment in small, methodologically weak studies generates research activity without producing reliable knowledge.

This highlights the importance of Open Research practices, including transparency, data sharing, and independent scrutiny, to identify weaknesses in evidence before unsupported claims become embedded in commercial products, public narratives, or expectations among vulnerable populations seeking answers.