Does what I eat change how I feel?

This page is about research, not about how you should eat. It reports what trials found and hands you to the tools, not a plan. If your relationship with food feels strained in any direction, start at Food & feelings — start here instead — that's the safeguarding page, this is the evidence page, and the check-in tool is the one you actually use day to day.

The trial behind most of the "diet helps depression" headlines is the SMILES trial (Supporting the Modification of lifestyle In Lowered Emotional States — Jacka et al., BMC Medicine, 2017). It's worth reading exactly what it did, because the headline number travels further than the design does.

• 12 weeks, single-blind, randomized. 67 people enrolled — 33 in the diet group, 34 in the control group.
• The control group wasn't "no intervention" — it was a matched social-support protocol, same visit schedule, same length.
• 55 of the 67 participants were already receiving psychotherapy, medication, or both. The trial tested diet added to treatment — never diet instead of it. Nothing on this page is a reason to stop, start, or change a prescribed medication — that decision belongs with the prescriber who started it.
• The diet group improved significantly more on a standard depression scale (Cohen's d = −1.16, a large effect).
• Remission: 32.3% (diet) vs. 8.0% (control). Number needed to treat: 4.1 — but the confidence interval on that is wide, 2.3 to 27.8, meaning the true effect could be much smaller than it looks.

Promising, not Well-supported. One small trial, however striking, isn't yet a large, replicated evidence base.

A site built on evidence-honesty can't cite SMILES's headline number and leave out the published challenge to it. Molendijk et al. (2018) raised questions about whether undisclosed recruitment practices could explain an effect size this large. Jacka and colleagues published a reply the same year defending the trial's conduct. Neither side has the final word — the exchange is public, and citing only the number while skipping the argument about it would be its own kind of misleading.

A handful of smaller trials have tested similar territory since:

• HELFIMED (Parletta et al., 2019) — a Mediterranean-style diet plus fish oil, with encouraging but smaller-scale results.
• AMMEND (Bayes et al., 2022) — a Mediterranean diet trial specifically in young men.
• PREDIDEP (Cabrera-Suárez et al., 2023) — continuing the same general line of inquiry.
• Large prospective cohorts link overall dietary pattern to later depression risk.

Reverse causation deserves its own paragraph, because it's the honest reading of most of this literature and almost no popular article says it out loud. Cohort studies can show that a dietary pattern and depression travel together over time; they cannot show which one moved first. Depression itself reliably changes how people eat — appetite, motivation to cook, and food choice all shift when someone is depressed. Some of what looks like "poor diet causes depression" is very plausibly "depression causes poor diet," running in the direction nobody wants to hear.

This is the most-mangled fact in wellness writing, so it's worth being exact. Roughly 90–95% of the body's serotonin is made in the gut, by cells called enterochromaffin cells. That serotonin does not cross into the brain — the blood–brain barrier blocks it. Brain serotonin is made separately, inside the brain, from the amino acid tryptophan, which does cross. Any sentence claiming that eating well "raises brain serotonin" by way of gut serotonin is not describing a real pathway.

The real route runs through the vagus nerve, immune and inflammatory signaling, and metabolites produced by gut bacteria — a message sent toward the brain, not a delivery of the molecule itself.

Fermented foods and overall microbiome diversity have some supporting evidence for mood (Wastyk, Gardner, et al., Stanford, 2021) Promising. Plainly stated: evidence for probiotic supplements specifically improving mood is weaker than the marketing around them suggests.

Tryptophan is the precursor to brain serotonin, and unlike gut serotonin, it genuinely does cross the blood–brain barrier — but through a transporter (LAT1) it shares with several other amino acids it has to compete against. A protein-rich meal raises tryptophan and its competitors at the same time, so brain tryptophan barely moves. This is why "turkey makes you sleepy" is folklore, not physiology.

Carbohydrate does something more useful here: it raises insulin, which clears the competing amino acids from circulation, so proportionally more tryptophan gets through. The counter-intuitive result is that the carbohydrate in a meal matters more than the protein for this specific pathway. Promising — acute tryptophan-depletion research supports this route being real, but no trial has shown that adjusting dietary tryptophan treats depression. This is a mechanism, not a treatment.

Dopamine precursors (tyrosine, phenylalanine) compete for transport the same way. Tyrosine has modest evidence for cognitive performance under acute stress; for mood specifically, the evidence is weak. Anecdotal. Tryptophan shows up in eggs, oats, dairy, poultry, seeds, and legumes — named descriptively, not as a list to eat more of, since no amount or daily target is established.

5-HTP and SAMe both sit downstream of this same pathway and carry a real serotonin-syndrome risk with other serotonergic medication — they belong on the supplements page, not here.

This section covers whole-diet patterns only. For specific compounds — magnesium, omega-3 doses, St John's Wort, and what each interacts with — see the dedicated Supplements page. Whole-diet evidence is consistently stronger than single-nutrient evidence, which is exactly why that page exists separately rather than as a list bolted onto this one.

"Bad oils cause inflammation, which damages mitochondria, which causes low mood" is a coherent-sounding hypothesis with almost no human trial evidence at any link in the chain. Human trials of linoleic acid intake do not consistently show increased inflammatory markers — the seed-oil claim is far more confident online than it is in the literature.

Inflammation and depression are genuinely associated in the research; that part is real. But association is not the same as the causal chain above, and anti-inflammatory diets have not been shown to treat depression. Anecdotal for the whole idea as currently popularized.

Three separate claims, three different evidence levels — keeping them apart matters.

1. Drug-resistant epilepsy Well-supported — long-established, clinician-supervised, especially in children; classic ketogenic, MCT, and modified Atkins protocols are Cochrane-reviewed. This is real and medically supervised.
2. Mood or bipolar disorder Anecdotal — small pilot and case-series work only. Interesting, not established.
3. Cancer — no clinical evidence of benefit. Preclinical work exists; human outcome evidence does not. A ketogenic diet is not a cancer treatment, and delaying or substituting oncology care to try one is dangerous.

Therapeutic ketogenic diets are medically supervised, not self-started, and are incompatible with eating-disorder recovery — the restriction involved can trigger relapse. See Food & feelings — start here if that's closer to your situation.

Not a meal plan, a list of foods to eat or avoid, a calorie or macro target, or anything framed around weight or appearance. This page tracks one outcome only — mood — and reports what randomized trials found about it, honestly including the trial that came up short (vitamin D) and the challenge to the trial that came up strongest (SMILES). Nothing above is an instruction.