Cortisol Face: The Sign Clinicians Are Told to Ignore

"You're Not Ugly, You Just Have Cortisol Face"

That sentence — some version of it — is the engine of one of the most durable wellness trends of the past two years. The format is always the same: two photographs of the same person, before and after. In the first, the face is round, the jawline soft, the eyes a little swollen. In the second, everything is sharper. The caption explains that the difference is cortisol, and that you can fix yours.

The commercial layer arrived almost immediately. Trade coverage in WWD dated the trend's arrival in the beauty industry to July 2024 and framed it, with admirable candor, as "an opportunity for brands to tap into stress-relief products."²⁹ By January 2025, the same reporter was citing retail-analytics data showing that the "cortisol mocktail" was averaging 1.4 million weekly TikTok views, up roughly 20,000% year over year, and that a single supplement brand built on the concept had become the fourth most-viewed wellness brand on the platform.²⁹ An investment banking director quoted in that piece explained the naming strategy better than any critic could: "Seeing the word stress stresses people out. If you use the word cortisol, one, it sounds more legit, and two, it doesn't create anxiety."²⁹

I want to be precise about what I am and am not disputing. Cortisol is a real hormone. Cushing's syndrome is a real disease. Facial rounding is a real sign of it. Facial puffiness is real, measurable, and varies within a single day. None of that is in question.

What I'm disputing is the causal chain that connects them — that everyday psychological stress raises your cortisol enough to remodel your face, and that a supplement can lower it back down. That chain has been assembled out of true parts. The parts do not connect.

The Trend Is Named After the Least Useful Sign

Start with the disease. Endogenous Cushing's syndrome — the kind that isn't caused by steroid medication — occurs at roughly two to three new cases per million people per year according to the Endocrine Society's clinical practice guideline.¹ Population registries in Denmark and Sweden put it slightly lower, between 1.2 and 2.4 per million annually. A family physician working a full career will statistically diagnose somewhere between zero and one case.

Facial rounding is genuinely common within that tiny population — reported in roughly 81% to 90% of patients across cohorts and registries. But that's the wrong direction of inference, and the guideline says so explicitly.

Table 1 of the 2008 Endocrine Society guideline sorts the physical signs of Cushing's into two columns.¹ The first is titled, in effect, features that best discriminate: easy bruising, facial plethora, proximal myopathy, and reddish-purple striae wider than a centimetre. The second column is features common in the general population and less discriminatory. In that column: dorsocervical fat pad, obesity, thin skin, peripheral edema, fatigue, weight gain, insomnia — and facial fullness.

The discriminating facial sign is redness. Not puffiness. The trend is named after the item in the second column.

It gets worse for the visual claim. No likelihood ratio has ever been published for isolated facial rounding as a diagnostic sign, and — this is the part that should stop anyone doing before-and-after photo analysis — nobody has ever measured how common a round face is in the general population. A 2019 review of Cushing's screening lists general-population prevalence figures for hypertension, obesity, acne and striae, and lists "round face" as simply unclear. The denominator for the entire viral claim does not exist.

And when someone actually tested whether you can see Cushing's, the answer was no.

To be fair to the underlying biology, the morphology is real when the disease is real. A 2026 study using three-dimensional photogrammetry compared 42 patients with confirmed Cushing's against 42 controls matched for age, sex and body mass index, and found genuine mid- and lower-face fat accumulation beyond what BMI explained, with a machine-learning classifier reaching 91.7% accuracy.⁴

Read that carefully. It took 3D imaging, BMI-matched controls and a trained classifier to detect the thing. That is simultaneously the best evidence that cortisol-driven facial change exists and the best evidence that you cannot see it in a bathroom mirror. An earlier computer-vision attempt with BMI-matched controls classified female patients correctly only 57% of the time — barely better than a coin.

Chronic Stress Doesn't Make Cortisol High. It Makes It Flat.

Here is where the trend doesn't just overreach — it runs backwards.

The foundational meta-analysis on chronic stress and the hypothalamic-pituitary-adrenal axis pooled 107 independent studies and 8,521 people.⁶ Its central finding, stated in the title as a question, is that cortisol goes up at the onset of a stressor and then comes back down as time passes. The characteristic chronic-stress pattern is lower morning cortisol, modestly higher evening cortisol, and a flattened daily rhythm — with effect sizes of d=0.39 for the flattened slope and d=0.31 for total daily output. Small effects, and the direction is not "high."

Then there's hair cortisol, which is the best available measure of chronic exposure because it integrates months of hormone deposition rather than capturing one moment.

So: how stressed you feel and how much cortisol is actually in your tissue are, across thousands of people, essentially unrelated. Which is worth sitting with, because "I feel stressed, therefore my cortisol is high" is the entire premise of the self-diagnosis.

Two more inconvenient facts. First, obesity is associated with normal or low circulating cortisol, despite higher overall production and turnover. Second, burnout — the state most of the audience for this content would identify with — is classically associated with a flattened rhythm and sometimes low cortisol, not a high one.

What about the daily cortisol readings people are being sold on? They aren't stable enough to be a personal trait. Across three multi-wave cohorts, day-to-day fluctuation accounted for 78% to 89% of the variance in the cortisol awakening response, and person-level reliability over more than a year was at or below 0.13.⁸ For comparison, in the same dataset, extraversion had a reliability of 0.77. To get a merely respectable estimate of someone's typical daily cortisol output, the authors calculated you'd need five visits, three months apart, three days each, six samples per day.

One Percent Versus Forty-Three Percent

Let's put numbers on the competing explanations for a rounder face.

Across 146 cohorts and more than 34,000 people, the correlation between hair cortisol and BMI is about r=0.10.¹⁰ That is roughly one percent of the variance in body weight. Other large datasets land in the same place — a cohort of 2,527 British adults found correlations of 0.10 for BMI and 0.08 for waist circumference. All of it cross-sectional, with the direction of causation unresolved, since obesity itself alters cortisol metabolism.

Now the boring variable. When researchers had raters judge facial adiposity from photographs and regressed it on measured BMI, body mass explained 43% of the variance in how fat a face looks.¹¹ Follow-up work using 179 facial landmark points found that heavier women have measurably wider faces (width-to-height r=0.36) and squarer jawlines (cheek-to-jaw r=−0.29). That is precisely the shape people are photographing and calling cortisol face.

And the threshold for noticing is low. A psychophysical study using two-alternative forced choice on faces morphed along a 31-step BMI continuum found the just-noticeable difference for facial weight change was 1.33 kg/m² — which the authors convert to 7.8 pounds for a woman of average height, about three and a half kilos.¹²

Gravity, Sleep, Alcohol, and the Things a Clinician Actually Checks

If not cortisol, then what? The honest answer is that morning facial puffiness is mostly plumbing.

Gravity. Fluid pools in your legs all day under gravity and redistributes upward when you lie down. This is well quantified, mostly out of sleep-apnoea research rather than endocrinology: overnight leg fluid loss of roughly 130 to 240 millilitres has been measured in patient populations, and experimentally shifting fluid rostrally in awake volunteers narrows the upper airway within five minutes. In one study, overnight change in leg fluid volume correlated with change in neck circumference at r=−0.79, and with how long the person had been sitting the previous day at r=−0.59.¹³ I'll be straight about the gap: no study has measured a specific fluid volume landing in the tissue around your eyes, and the widely repeated claim that morning puffiness resolves in exactly 30 to 60 minutes has no primary source behind it. The mechanism is solid; the facial-specific numbers aren't published.

Sodium — and this one surprised me. Eight separate literature searches for controlled evidence that dietary salt produces visible facial swelling in healthy people returned nothing. The best-controlled experiment points the opposite way: in a randomised metabolic-ward study, healthy men were fed sodium chloride across a range from 50 to 550 milliequivalents per day.¹⁶ Plasma volume rose dose-dependently by about 315 mL at the highest intake — but total body water did not increase, and body mass did not increase. High sodium moved fluid out of the interstitium and into the vasculature, which is the wrong vector for puffy tissue. Long-duration balance studies have since shown that sodium is stored osmotically inactive in skin and muscle and mobilised without parallel weight change. Any "cut salt, drop two kilos of water" claim is inconsistent with the controlled feeding data.

Alcohol. The diuresis is smaller than folklore suggests — a crossover trial of one litre of weak beer found a net urine difference of about 158 mL versus alcohol-free, and when participants were already dehydrated the diuresis disappeared entirely. What alcohol demonstrably does to a face isn't oedema, it's vascular: in 82,737 women followed for 14 years with 4,945 clinician-diagnosed incident cases, alcohol intake raised rosacea risk in a clean dose-response, from HR 1.12 at one to four grams a day to HR 1.53 at 30 grams or more, P for trend below 0.0001.¹⁷ Facial redness, not facial swelling. Which, given that plethora is the discriminating Cushing's sign and puffiness isn't, is a nice irony.

And then the actual differential. If a clinician sees persistent facial swelling, the list they work through is: hypothyroidism (non-pitting, doughy, lateral eyebrow loss — a TSH settles it); nephrotic syndrome (periorbital oedema worst on waking, resolving through the day — a urine dipstick settles it); ACE-inhibitor angioedema (asymmetric, no itch, mostly within the first ten weeks, with substantially higher risk in Black patients); obstructive sleep apnoea; sialadenosis from bulimia, which widens the jawline rather than the eye region; dermatomyositis, where the periorbital change is violaceous; and superior vena cava syndrome, where facial oedema appears in about 82% of cases and roughly two-thirds are malignant. None of these are on a supplement label.

Every Mass-Market Cortisol Kit Is the One Test the Guideline Rejects

Suppose the trend convinces you to check. The consumer options are a finger-prick-plus-saliva kit at around $99, single-timepoint venous blood tests at $79 to $89, or an annual blood draw inside a $365-a-year panel. Every one of the mass-market products is a single-timepoint measurement.

The Endocrine Society guideline recommends against using a random serum cortisol as a diagnostic test at all, and recommends against widespread screening outside defined risk groups.¹ What it does endorse are three tests, each requiring either timing discipline or repetition: the 1 mg overnight dexamethasone suppression test, 24-hour urinary free cortisol, and late-night salivary cortisol collected twice. A hierarchical meta-analysis of 139 studies and 14,140 participants put their sensitivities at 98.6%, 94.0% and 95.8% respectively.¹⁸ None of them is a random midday draw.

The variability is the reason. Salivary cortisol has within-subject reference change values in the range of 96% to 245% — meaning a result has to roughly double or triple before the change is distinguishable from noise. Twenty-four-hour urinary free cortisol has an intra-patient coefficient of variation of 52% in people with active, moderate-to-severe Cushing's disease, and in that same series it did not correlate with clinical features.¹⁹ In general-population samples, the strongest predictor of a 24-hour urinary cortisol result was how much fluid the person drank. Venepuncture itself induces a cortisol response; so does a finger-stick. The act of sampling perturbs the thing being sampled.

False positives cluster in exactly the population most likely to buy a test: late-night salivary cortisol specificity drops to about 85% by immunoassay in obese subjects, roughly 20% of men over 60 have at least one elevated late-night value, and up to half of women taking oral contraceptives have a falsely positive dexamethasone suppression test.

As for oversight — there isn't any. These are laboratory-developed tests run in certified labs, not FDA-cleared devices. The FDA's own consumer page notes such tests "typically do not have the FDA's independent assurance of the analytical validity, clinical validity, or clear communication of test results." The rule that would have changed that was vacated in full by a federal court in March 2025, the agency did not appeal, and the vacatur was formally implemented.²⁰ As of today there is no premarket review of at-home cortisol tests and no pending rule that would create one.

Adrenal Fatigue Was Reviewed Out of Existence in 2016. The Shelf Didn't Notice.

The construct underneath most of this — that stress "exhausts" the adrenal glands — was systematically reviewed a decade ago. Screening 3,470 articles down to 58 studies, the authors concluded, verbatim: "This systematic review proves that there is no substantiation that 'adrenal fatigue' is an actual medical condition. Therefore, adrenal fatigue is still a myth."²¹ The Endocrine Society's patient library says the same thing and adds that no test can detect it.

Which brings us to what's actually being sold. Ashwagandha is the anchor ingredient, and it is not a niche product — US retail scanner data put mainstream-channel ashwagandha sales at $144.5 million in 2024, up 13.8% year over year and the third-bestselling herb in the country.

The pooled picture is genuinely inconsistent. One 2026 systematic review and meta-analysis of 23 trials and 1,706 patients reports a cortisol effect of SMD −1.18 — and also a signal on thyroxine, which is its own conversation. A 2025 review found no significant impact on perceived stress at all (SMD −0.355, 95% CI −1.188 to 0.47, P=0.40), with risk of bias rated generally moderate.²³ A 2022 review concluded the reduction in acute stress was modest and "largely influenced by studies with a high risk of bias," with no significant effect on fatigue and inconsistent anxiety findings. One review in this literature reports standardised mean differences around −6, which is roughly six standard deviations, and should be read as a red flag rather than a result.

The rest of the cabinet is thinner still. Phosphatidylserine's cortisol claim traces to a 1990 study of eight healthy men given bovine-derived extract intravenously before exercise; today's products are oral soy capsules, and the FDA's 2003 qualified-health-claim letter concluded there is "very little scientific evidence" for the claim. The pivotal rhodiola trials all test one proprietary extract with recurring authorship, and a Cochrane-style review found every included study at high or unclear risk of bias — no systematic review has assessed cortisol at all. There is no cortisol-specific trial of holy basil. And "cortisol cocktails" have an empty evidence base rather than a weak one: full-text searches of the biomedical literature return no clinical study of the drink. Worth knowing if you make one: cream of tartar is potassium bitartrate at roughly 16,500 mg of potassium per 100 g, which matters for anyone with kidney disease or on ACE inhibitors, ARBs or potassium-sparing diuretics.

Now the part I find genuinely remarkable. In 2004 the Federal Trade Commission sued the marketers of CortiSlim and CortiStress over cortisol claims. The active blend was L-theanine, magnolia bark, magnesium and B vitamins. The settlement announced in January 2007 had seven marketers surrendering assets worth at least $12 million, including $8.4 million in cash, and the FTC simultaneously warned more than 25 websites that it was "not aware of any competent and reliable scientific evidence" for cortisol claims.²⁶ It has never retracted that.

Read the back of a cortisol supplement in 2026 and you are reading substantially the same shopping list. And in June 2026 the FTC filed a fresh complaint alleging unsubstantiated claims that products "lower, reduce, or regulate cortisol" — the pivotal study relied on reportedly having no placebo control and ten participants — then moved to hold the original CortiSlim formulator in contempt of the 2005 order.²⁶ Same hormone, same district court, twenty-one years apart.

In the UK, the advertising regulator upheld a ruling in June 2026 against a supplement brand whose ads promised to "calm cortisol… so you can beat puffiness, stubborn weight, and low energy" and said "goodbye to cortisol face," including a woman in a lab coat asserting that high cortisol makes your body hold water "right in your face."²⁷ The ads were flagged by the regulator's automated monitoring system, not by a consumer complaint.

And the manual interventions? There is exactly one randomised trial of facial gua sha in existence, 34 women randomised to a roller or a gua sha tool.²⁸ Both arms improved. There was no sham arm and no untreated arm, so every within-group improvement is uninterpretable, and the trial used a post hoc power analysis. It did not measure puffiness, oedema, fluid, or lymphatic flow. A 2025 systematic review of facial massage devices framed itself explicitly around gua sha and jade rollers — and not one study of either met its inclusion criteria.

The Harm Isn't Wasted Money. It's Iatrogenic.

I don't think this trend's main cost is a $44 jar of magnesium. Three other things worry me more.

Products sold to lower cortisol have been found to contain corticosteroids. A blinded, duplicate assay of 12 over-the-counter "adrenal support" supplements found detectable triiodothyronine in all twelve, pregnenolone in 42%, budesonide in 25%, androstenedione in 17%, and cortisol itself in 8%.²⁵ That is the strongest single safety finding in this entire file. Steroid exposure is also the far more common route to a genuinely rounded face: iatrogenic Cushing's occurs at roughly 550 cases per million person-years among oral steroid users, versus two to three per million per year for the endogenous disease, with a five-year risk climbing from 1.1% at low cumulative doses to 47.1% at the highest.⁵

Ashwagandha injures livers. The NIH LiverTox monograph rates it likelihood score B, "likely cause of clinically apparent liver injury," and notes that rare instances of fatal injury or emergency transplant have been reported, particularly in people with pre-existing liver disease. A case series from Iceland and the US network described five patients developing jaundice two to twelve weeks after starting, with a cholestatic or mixed pattern.²⁴ An Indian series of 23 patients included three with acute-on-chronic liver failure, all of whom died. Denmark banned it from food supplements in 2023 after concluding no safe lower intake limit could be established.

And real Cushing's is already dangerously slow to find. A meta-analysis of 5,367 patients across 44 studies put the mean time to diagnosis at 34 months, with no improvement between the pre- and post-2000 eras.³ Patient-reported data has people seeing an average of 4.6 physicians. In the European registry, 45% of deaths occurred within 90 days of starting treatment. A trend that trains millions of people to attribute facial change to ordinary stress, and a parallel one that trains them to attribute it to a rare tumour, are both capable of adding noise to a signal that is already too quiet.

There is a counter-case, and it deserves airtime. In 2024 a 21-year-old woman in Hertfordshire saw a social-media checklist of Cushing's symptoms — "I was reading a checklist of my life," she told a UK broadcaster — after having been previously misdiagnosed with PCOS. She raised it herself at an endocrinology appointment. Her cortisol came back at 1,334 nmol/L against a normal of roughly 135. Imaging found a 17-centimetre cancerous mass growing into her liver. She was told the surgery saved her life. Social media can push a rare diagnosis into public awareness. It can also turn normal physiologic variation into a cascade of unnecessary endocrine workups. Both are happening.

One last note on the evidence base for the trend itself, because it says something. Two independent searches across the major dermatology and endocrinology journals found no editorial, viewpoint, letter or correspondence about "cortisol face" anywhere in the peer-reviewed literature, and no position statement from any dermatology or endocrine society. The only primary research is a conference abstract presented in March 2026, reporting that cortisol content on social platforms is dominated by "emotionally charged, fear-based, and aspirational marketing" with inconsistent platform enforcement — and it reports no sample size and no percentages.³⁰ Meanwhile the single most-quoted number in the coverage, "nearly 800 million views," is uncited at its source and has been attached to three different hashtags on its way through three outlets. The profession's response has happened in trade press and continuing-education slides, not in journals. That silence is part of why the trend has had such a clear run.

A Real Hormone, a Real Disease, and a Chain That Doesn't Connect