Only 18 of 63 popular "weight loss" foods have trial evidence that they help. None beats about 4.4 lb (2 kg).

Bioma Insights · I-002 · first published 9 October 2026 · data and review reports published alongside, free to reuse with a link to this page or to biomanutrient.app

In plain words

Every month people search for whether cottage cheese, chia seeds, lemon water or green tea will help them lose weight. We took 63 of these foods, drinks, diets and supplements, each searched about 1,000 times a month or more, and looked for the randomized trials behind each one.

• 18 of the 63 have trial evidence of weight loss. In the best available evidence, people lost a typical 2 lb (0.9 kg) (median) more than a comparison group, which was often a placebo or a similar diet without the food. The largest effect was 4.6 lb (2.1 kg), for psyllium fiber supplements in overweight adults.
• 20 were tested for weight, and the best available evidence found no meaningful weight loss. That includes chia seeds, avocado, dark chocolate, coconut oil, fish oil, kefir and quinoa. For a few of them, other reviews disagree, and we say which.
• 11 have never been tested as a way to lose weight. We found no randomized trial that gave them, as the thing being tested, to people trying to lose weight or with overweight and measured their weight. That includes cottage cheese, lemon water, popcorn, granola, tofu and coconut water. Some of them appear in trials for other purposes, which we list.
• 14 were tested only indirectly, for example as an extract in a capsule, as part of a mixed product, or with body fat rather than body weight measured. We count them separately rather than guess.

This is not a judgement on whether these foods are healthy, and some supplements on the list carry their own safety questions. It is about what the trials can support on weight. When a headline calls a food "good for weight loss", it has usually been tested much less than the headline suggests, and sometimes not as a weight-loss food at all.

For journalists: method, numbers and sources

What we counted

The 63 items are every "does X help you lose weight" question in Bioma's evidence reviews. Each was included because its question has about 1,000 or more estimated monthly searches in our keyword data (DataForSEO, Google, United States, English, September–October 2026); the estimate for each item is in the CSV. They are not a random sample of foods. They are the ones people ask about.

Each item has its own published review on Bioma that lists the searches run (PubMed, systematic-review registers, trial registries, retraction checks), the trials included and excluded, and what would change the answer. Links to all 63 are in the CSV published with this page.

How each item was labeled

Two independent readers applied the same written rules to each of the 63 reviews, without seeing each other's results. Both readers were AI language-model agents (Anthropic Claude Sonnet and Claude Opus), not people. They agreed on 54 of 63. The 9 items where they disagreed went into the "tested only indirectly" group.

We then gave this article, without access to our data, to two further AI reviewers, Anthropic Claude Opus and OpenAI GPT-5.5, and asked each to check every claim against PubMed and other public sources and to hunt for trials we had missed. Over five rounds they found trials for eight items that our reviews had not counted (tuna, Greek yogurt, banana products, prawns, tofu, corn, coconut water, granola bars) and conflicting meta-analyses for ginger, whey, coconut oil, grapefruit, chia, green tea and psyllium. As a result, five items our readers had agreed on (banana, tuna, collagen, pasta, Greek yogurt) were moved to "tested only indirectly", and five items (prawns, tofu, corn, coconut water, granola) stayed "not tested for weight loss" but gained notes on the trials that exist for other purposes. Every change is listed in the CSV, and the review reports are published with the data. No human dietitian has checked the labels.

• Weight loss: the best available evidence, usually the largest or most recent meta-analysis, found a statistically significant reduction in body weight. We report the pooled difference against control.
• No meaningful weight loss: the item was tested with body weight or BMI as an outcome, and the best available evidence found no statistically significant reduction, or a change its authors judged not clinically meaningful (coconut oil: 0.09 lb (0.04 kg) higher per extra 1 tsp (5 mL) a day). Creatine is in this group because its best evidence shows a small weight gain (0.89 kg, mostly lean mass and water), not a loss. This does not prove there is no effect; small effects may be missed.
• Not tested for weight loss: no randomized trial lasting at least four weeks in which the item itself was assigned as the intervention, rather than the comparison, to people trying to lose weight or with overweight or obesity, with body weight or BMI measured. Evidence about a wider group (dairy, fish, whole grains) does not count. Trials of the item for other purposes are listed below.
• Tested only indirectly: the item was studied for weight, but only as an extract, inside a combined or fortified product, or with fat mass instead of body weight as the outcome. This group includes the 9 items where our two readers disagreed and the 5 moved here after independent review.

The 18 items with evidence of weight loss

The median of the 17 effects reported in kilograms is 2 lb (0.92 kg), and stays 2 lb (0.92 kg) without the two single-trial rows (salmon, sparkling water). Apple cider vinegar (standardized units) is not in the median. The vinegar and apple cider vinegar rows draw partly on the same trials; the retracted trial is not in the vinegar review's weight pool.

The 20 items with no meaningful weight loss in the best available evidence

Avocado, beet juice, chia seeds, coconut oil, creatine, dark chocolate, fish oil, fruit juice (100%), ginger, grapefruit, kefir, nuts, peanut butter, quinoa, saffron, vitamin B6, water, watermelon, whey protein, whole wheat bread.

Where other reviews disagree:

• Ginger: the largest and most recent meta-analysis (36 RCTs, 2025, PMID 41101746) found no change in body weight; an earlier meta-analysis of 27 RCTs (2024, PMID 38261398) found 3.4 lb (1.52 kg).
• Coconut oil: the most recent dose-response meta-analysis (2025, PMID 40481535) found weight 0.09 lb (0.04 kg) higher per extra 1 tsp (5 mL) a day, which its authors judged not clinically meaningful; a meta-analysis comparing coconut oil with other oils (PMID 34786908) reported 1.7 lb (0.75 kg) lower weight.
• Whey protein: our review relies on the larger null analyses; a meta-analysis in overweight and obese adults (PMID 29087242) reported a significant reduction.
• Grapefruit: an early trial in 91 obese adults (PMID 16579728) found more weight loss with fresh grapefruit than placebo; two later trials (PMIDs 22304836, 21288350) and the pooled analysis found no difference.
• Dark chocolate: no overall effect, but one meta-analysis reported a significant reduction in a subgroup eating at least 1.1 oz (30 g) a day for 4–8 weeks.
• Vitamin B6: rests on one 8-week trial in 44 women; weight did not differ, though its authors report lower fat mass against placebo.
• Chia seeds: one meta-analysis (PMID 39299649) concludes there may be a weight benefit; the GRADE-assessed meta-analysis we rely on (PMID 39225983) found no significant difference.

The best evidence and PubMed ID for each item is in the CSV.

The 11 items not tested for weight loss

Almond milk, cabbage, coconut water, corn, cottage cheese, granola, lemon water, popcorn, shrimp, soursop, tofu.

Trials of some of these exist for other purposes, and a journalist searching PubMed will find them:

• Granola: a 12-week trial in 55 overweight adults compared two granola bars, with and without added oligofructose (PMID 29490721); granola was in both arms.
• Tofu: a 4-week crossover lipid trial in 42 men, tofu against lean meat (PMID 10694766), and a 4-week crossover of Kori-tofu against whey in 45 adults with raised cholesterol, where weight rose about equally in both arms (PMID 36615709).
• Shrimp: a 12-week crossover in 23 men, prawns against crab sticks; weight stayed stable (PMID 20196970).
• Corn: corn cereal as the comparison arm in a barley trial (PMID 40944189).
• Coconut water: acute rehydration trials after exercise (PMIDs 12056182, 17883020).
• Cabbage: trials of kimchi test a fermented product, not cabbage itself.
• Cottage cheese: only single-meal satiety studies.

We did not count evidence about a wider group either, because a headline about "cottage cheese for weight loss" makes a claim about cottage cheese.

The 14 items tested only indirectly

• Banana: green banana biomass and banana starch in people with diabetes or prediabetes (PMIDs 30887937, 20623003), not the fresh fruit.
• Collagen peptides: a meta-analysis of 12 placebo-controlled RCTs (PMID 35658958) found 2.6 lb (1.2 kg) less fat mass; body weight was not pooled.
• Greek yogurt: two trials in lean or exercising adults not trying to lose weight found no difference in body mass (PMIDs 31114790, 41470847). One 8-week trial of kashk, a related strained high-protein yogurt, in 70 overweight women on a low-energy diet found BMI fell more than in the control group (PMID 31987116).
• Coffee: trials of green coffee extract and caffeine, not brewed coffee.
• Ginger tea: ginger capsules, not tea.
• Lemon juice: a 7-day lemon detox diet.
• Pasta: a meta-analysis of pasta eaten within low-glycemic-index diets (PMID 29615407, −1.4 lb (−0.63 kg)). No trial tested pasta on its own.
• Pineapple: a bromelain enzyme supplement.
• Spinach: trials of a spinach-leaf (thylakoid) extract, for example one of 48 women; no trial of spinach as eaten.
• Sweet potato: a meal-replacement trial.
• Tuna: one 12-week RCT of selenium-enriched tuna nuggets in 120 patients with ischemic heart disease (PMID 42369918), which found 4 lb (1.8 kg) more weight loss than standard dietary care.
• Vitamin B12: given together with folic acid.
• Whole grain bread: evidence on whole grains in general, and a 12-week trial of whole-grain rye foods against refined wheat in 255 adults with overweight that found no difference (PMID 41861447); no trial of whole grain bread itself.
• Yogurt: trials compare yogurt with other foods or yogurt types rather than testing it for weight loss, for example a 34-person trial confounded by calcium and a 9-week trial against soy pudding with no weight difference (PMID 29179781).

Limits of this analysis

• The 63 items reflect what people search for, not a balanced sample of foods.
• The Bioma reviews were searched on 6–7 October 2026; each review page lists its exact searches. "Not tested for weight loss" means we and our reviewers did not find such a trial. Our independent reviewers found trials for eight items that our own reviews had not counted, which shows that searches miss things. A trial we missed would move an item into another group, and we will update this page if one is found.
• Most effects come from trials lasting weeks to a few months, often in people who were also eating fewer calories. They do not tell you what happens over years.
• Supplements, whole foods and whole diets are mixed in this list because people ask about all of them. The table says which is which.
• All labeling and checking was done by AI agents under written rules. The CSV gives each label from both readers, the reviewers' corrections, and the sentence from each Bioma review that supports the label, so every row can be checked by hand.

Reusing the data

The full table (63 rows: label, effect size, confidence interval, best evidence, PubMed ID, link to the Bioma review, both readers' labels and the independent reviewers' notes) is published with this page as a CSV and is free to use under CC BY 4.0. Please credit "Bioma" and link either to this page or to the Bioma home page, biomanutrient.app.

Questions or corrections: editorial@biomanutrient.app. If you find a trial we missed, we will update the page and say so here.