If you remember one thing from this site, make it this one.
The most impressive numbers in metabolic psychiatry come from studies of bipolar disorder and schizophrenia. Those numbers then get repeated in articles about depression, by people who either have not read the studies or are hoping you will not.
What the confusion looks like
The Stanford pilot study is the usual source. Twenty-three people with bipolar disorder or schizophrenia, four months on a ketogenic diet, and a reported 32% reduction in symptom scores among the schizophrenia participants.
That is a genuine finding, published in a real journal. It is also a finding about schizophrenia.
By the time it has travelled through a few articles it has become “keto reduces psychiatric symptoms by 32%”, and then “a diet that cuts depression by a third”. Nobody necessarily lied. The condition just quietly fell out of the sentence.
Why it is not a small distinction
Depression, bipolar disorder and schizophrenia are different illnesses. They have different diagnostic criteria, different courses, different biology as far as anyone can tell, and — most practically — different treatments. Lithium and antipsychotics are mainstays in bipolar disorder and schizophrenia; they are not first-line treatments for depression. Antidepressants, given alone to someone with undiagnosed bipolar disorder, can trigger a manic episode.
Nothing transfers automatically between them. A result in one tells you to go and run the study in the other. It does not tell you the answer.
There is also a reason to expect the ketogenic findings in bipolar disorder specifically: several mood stabilisers used in that condition began as anticonvulsants, and the ketogenic diet is an epilepsy treatment. The thread connecting them may simply not run to depression at all.
What exists for depression
One randomised trial, published February 2026, in 88 people with treatment-resistant depression. The ketogenic group improved by about two points more on a depression questionnaire at six weeks. By twelve weeks the difference was no longer significant, and the authors called the clinical relevance uncertain.
One case series of three patients.
One trial registered in Basel that has not reported.
That is the complete evidence base in depression. It is thin, and it deserves to be described as thin.
How to check an article yourself
Four questions, and they take about a minute:
- Which condition were the participants diagnosed with? If the article does not say, that is itself informative.
- How many people? Three is a case series. Twenty-three is a pilot. Eighty-eight is a small trial.
- Was there a control group? If everyone received the diet, improvement could be the diet, the attention, the expectation, or time.
- How long did the effect last? A difference at six weeks that is gone at twelve is a different thing from a durable one.
Every study this site relies on is listed with those answers already filled in on the evidence page.
Where to go next
General education, not a diagnosis and not a treatment. If your mood has been low for weeks, your GP is the next step.
Explore the guides