What Actually Breaks a Fast (and What Doesn't)

Two meanings of breaking a fast

What actually breaks a fast depends on which fast you mean: anything containing calories ends a strict, water-only fast, while the insulin-based meaning counts a fast as broken only when something raises insulin enough to slow the release of stored fat (Jensen et al., Diabetes 1989). Those two tests agree on most of the menu and disagree on a short list of everyday favorites. Water passes both trivially. Black coffee carries under five calories a cup yet left fasting metabolic markers essentially unchanged at rest in the most direct trial available (Clinical Nutrition ESPEN, 2021), which is why it holds its place in both tiers. The interesting disagreements live in between, and that middle ground is where most real fasting questions get asked.

Everything below rules on candidates one at a time, sorted into the three tiers people actually fast by: keeps the fast, gray area, and ends the fast. Verdicts come first, mechanisms second, because nobody mid-fast wants an essay. One reading rule for the table: the Ruling column answers the stricter, strict-fast question, so anything marked keeps satisfies the metabolic test too.

The verdict table: keeps, gray, ends

Here is the whole question in one table, from water to whey. Four rows earn longer explanations afterward because their evidence is genuinely mixed, genuinely interesting, or both.

Three rows carry footnotes worth their citations. Diluted apple cider vinegar does more than stay neutral: Ostman's team reported that acetic acid served alongside a bread meal lowered both blood sugar and insulin responses in a dose-dependent pattern in healthy volunteers, though the trials tested vinegar generally rather than apple cider vinegar specifically, so appetite and fat-loss overclaims run ahead of the data. Plain tea carries almost no calories, and the closest trial evidence is a 2020 meta-analysis of 27 randomized trials of green tea totaling 2,194 participants that found slightly lower fasting blood sugar with no significant effect on fasting insulin, so treat it as indirect support for other plain teas. Sugar-free gum survives on thinner evidence, a 2015 study of fasted men chewing gum tracked gut hormones rather than any insulin jump, and 2020 laboratory-medicine work treated pre-test gum as a minor source of measurement variability, so no robust trial shows sugar-free gum raising insulin materially during a fast. MCT oil and butter in coffee simply bring calories, which ends a strict fast by definition, and their human evidence base is thinner than coffee's.

One practical trap hides in two green rows: flavored electrolyte mixes and some packet sweeteners bulk themselves out with sugar or maltodextrin, both of which carry calories. The ingredient label settles those cases faster than any study.

The four rulings that deserve more than one line

Black coffee and the caffeine asterisk

Black coffee earns its place honestly, with one asterisk that gets misattributed constantly. In a 2002 clamp study, Keijzers and colleagues found that acute intravenous caffeine lowered whole-body insulin sensitivity by about 15 percent, alongside a roughly fivefold rise in epinephrine. That result circulates as coffee raising insulin, which it never showed. The study infused pure caffeine into a vein, brewed coffee is a different exposure carrying chlorogenic acids and other compounds that may pull the other way, and the closest thing to a direct test of coffee itself, black coffee before bloodwork, showed no meaningful change in fasting metabolic markers at rest. So the calibrated verdict stands: black coffee keeps the fast, caffeine's downstream effects on sensitivity are real in isolation and unresolved in coffee, and the open question concerns what happens later in the day, not whether the fast survived. Individual responses vary everywhere on this page, which is why these are group-level findings rather than promises about your afternoon.

Artificial sweeteners: a genuinely two-sided file

Sweetener verdicts attract the loudest wrong answers in this topic, in both directions, so here is the two-sided version. A 2020 meta-analysis of randomized controlled trials in the American Journal of Clinical Nutrition found no acute insulin or blood sugar effect from low-energy sweeteners compared with water. The scarier mechanism, taste-triggered insulin release, exists but is small: a 2020 systematic review of 50 studies concluded that insulin rose in only about four in ten treatments tested, with a median increase around 2.5 micro-units, judged small against ordinary background fluctuation, and a companion meta-analysis estimated the pooled effect as statistically detectable but minor. The genuine caution signal is narrow. Pepino and colleagues reported in Diabetes Care that sucralose taken together with a carbohydrate load altered glycemic and insulin responses in people who did not habitually use sweeteners, a finding contested in published correspondence and tested with carbohydrates present rather than sucralose alone mid-fast. Neither camp gets its slogan. What the trials support is narrower: no calories, small contested signals, and no good answer yet for very long fasts. The practical trap again sits on the label, where some packet sweeteners carry maltodextrin fillers that add real calories.

Dairy: why a splash counts and a latte is a meal

Milk looks harmless because it barely moves blood sugar. Insulin tells a different story, and Holt's classic insulin-index study showed why: across 38 common foods, protein-rich foods and bakery products produced insulin responses disproportionately higher than their glycemic responses, so a food's effect on blood sugar alone does not predict its effect on insulin. Follow-up work by Nilsson and colleagues reached the same conclusion for milk proteins specifically, tracing their outsized insulin effect to gut hormones and amino acids rather than to the small lactose load. Lactose adds carbohydrate, butterfat adds calories, and together they mean a splash of cream in otherwise black coffee already nudges insulin enough to end a strict fast by definition. A latte is mostly warm milk. It is simply a meal wearing a cup.

BCAAs and protein powder: the breakers that look like drinks

This is the cleanest cut on the page. A randomized trial documented that swallowing just over 30 grams of branched-chain amino acids raised insulin far above placebo, an effect the investigators partly credited to gut hormones, and whey-loading trials reported the same strongly insulinotropic pattern for protein powder. Protein also calls in glucagon, the hormone that counterbalances insulin (peak glucagon roughly doubled after whey or casein drinks in Hoefle et al., 2015), but by the insulin definition the fast is still over. Blood sugar barely moved in either case, and that is the lesson: amino acids can raise insulin without tripping the blood sugar rise people use as their shortcut test. If your fast is meant to keep insulin low, many people save the shaker bottle for the eating window.

How to break a fast gently

Whatever ended the fast, the next decision matters more than usual. After a long stretch without food, many people notice their first meal seems to land hardest, a general experience rather than a measured law, so the size of that first insulin request is worth choosing on purpose. The house rule is simple: score the first meal, treat it as the most important score of the day, and aim it lower-insulin and protein-forward. Eggs with smoked salmon and avocado, chicken with roasted vegetables, Greek yogurt with toasted nuts, meals like these reopen digestion without asking for a big insulin response.

In RealFoods, scoring means a photo: the app estimates each meal's likely insulin impact from the meal itself, no device worn, nothing drawn from your bloodstream, and post-fast sensitivity is a reason to weight the first number more heavily than the ones that follow. For where that first meal sits in the larger arc of what your body does hour by hour, see our guide to the stages of fasting.

About autophagy, honestly

Many readers protect a fasting window for autophagy, the cellular cleanup process, so this page owes them the same honesty as every other section. Direct measurement of autophagy timing during human fasting has never been demonstrated, a limit the field states plainly, and the largest human trial to date illustrates it. Published in the Journal of Physiology in 2025 as an exploratory analysis, it randomized 121 people with obesity to standard care, calorie restriction, or intermittent fasting plus time-restricted eating for six months. Autophagic flux differed between the intermittent fasting group and standard care at six months (a post hoc comparison, P = 0.04, partly because flux tended to fall in the control group), while calorie restriction did not differ from standard care, so the authors describe a possible increase rather than a measured timeline. The familiar hour bands, 16 hours, 24 hours, circulate from animal-model extrapolation, not from measured human thresholds. Nothing in the table above, coffee included, has been shown to switch human autophagy off or on. Insulin responses to food are measurable today while autophagy timetables remain provisional, which is why this page stays anchored to the metabolic framing.

Where this page stops

This framework is metabolic only. It does not describe, rank, or rule on Ramadan or any religious observance, whose practices belong entirely to the traditions that keep them. Fasting there is not a fat-loss protocol, and this page does not grade it as one.

Medically ordered fasting is clinician territory too. Blood-test and pre-surgery windows, often water-only, were set for reasons outside this framework, so follow the instructions your clinician gave rather than the tiers above. And medication never bends around a fast: take prescribed medicine exactly as directed rather than adjusting it to protect a fasting window, and bring timing questions to the prescriber or pharmacist who knows your case. Formulation is fair game as education, gummy and chewable supplements commonly carry sugars and carrier carbohydrates such as maltodextrin while tablets are generally negligible, which is worth knowing when you plan your morning.

Finally, no calorie cutoff gets endorsed here, because none has earned it. The United States Food and Drug Administration allows products under five calories to round down to zero on labels, a rounding convention rather than physiology. The popular allowance of roughly 100 calories during a fast, sometimes called a dirty fast, has no research basis behind its number. Your strict fast ends when calories arrive, your fast by the insulin definition ends when insulin rises enough to slow the release of stored fat, and both are cleaner questions than any single figure.

Ready to stop guessing what to eat?

Common questions

Does coffee break a fast?

No, black coffee keeps the fast for most people. It carries under five calories a cup, and the most direct test available, black coffee before bloodwork, found no meaningful change in fasting metabolic markers at rest. The honest asterisk is caffeine itself, which lowered insulin sensitivity about 15 percent when infused intravenously in a 2002 clamp study, a different exposure from brewed coffee. Keep it black, because cream and milk end a strict fast.

Do artificial sweeteners break a fast?

Probably not, though the honest answer is two-sided. A meta-analysis of randomized trials found no acute insulin or blood sugar effect from low-energy sweeteners versus water, and reviews of taste-triggered insulin release concluded the effect exists but is tiny, with a median rise around 2.5 micro-units. The one solid caution is sucralose consumed together with a carbohydrate load, which altered responses in non-habitual users in a contested 2013 trial. Read packet labels too, some fillers such as maltodextrin carry calories.

Does gum break a fast?

Sugar-free gum is the mildest gray area on this page. A piece runs about 2 to 6 calories, the evidence base is thin, and no robust trial shows a material insulin effect during a fast. The exception is maltitol-sweetened gum, which behaves close to sugar, so read the sweetener on the label.

How many calories will break a fast?

There is no validated cutoff. The popular allowance of roughly 100 calories, sometimes called a dirty fast, has no research basis, and treating products under five calories as zero is a Food and Drug Administration labeling rounding convention rather than physiology. Calories end a strict fast by definition, while the insulin-based definition counts a fast as broken only when something raises insulin enough to slow the release of stored fat (Jensen et al., Diabetes 1989).

Do BCAAs or protein powder break a fast?

Yes, both. A randomized trial documented oral branched-chain amino acids raising insulin far above placebo, partly through gut hormones, and whey protein shows the same strongly insulinotropic pattern. Protein also calls in glucagon, the hormone that counterbalances insulin (Hoefle et al., 2015), but by the insulin definition the fast is still over. They barely move blood sugar, which is the trap, amino acids can raise insulin enough to break a fast by the insulin definition without producing the blood sugar rise people expect to catch them.

What should you eat to break a fast?

Something lower-insulin and protein-forward, in a modest portion. After a long fast, the first meal is the one worth choosing most deliberately, so eggs with smoked salmon, chicken with roasted vegetables, or Greek yogurt with nuts reopen digestion without asking for a large insulin response. Score that first meal, treat it as the day's most important one, and let variety return with later meals.

Keep reading: the stages of fasting, how to know if your body is burning fat, or how to lower insulin naturally.