We Scored 12 Health Trends This Week and Read Every Anchor Study — Only 2 Held Up

Every week we sweep five sources for what the internet is actually chasing in health and self-optimisation, score each topic on five factors, and pick which ones become videos. This week the sweep ran as 16 agents: three scouts, then one adversarial verifier per candidate whose only job was to falsify the angle before it reached a script. Thirty-two raw candidates became twelve scored ones. All twelve anchor studies were pulled and read. All twelve held.

What did not hold were the framings. Five of the twelve angles on the board are not what the scout proposed, because the verifier proved the original version wrong — and that is the part worth publishing.

Liver detox: the claim is on every bottle and in no trial

A 2026 American Journal of Gastroenterology analysis took the twenty best-selling liver-cleanse products on Amazon — 1,420,584 units and $38.8 million a year — and found that every single one promised to eliminate toxins or provide a detox, while 85% claimed to enhance liver function. No trial measured toxin elimination, because that is not a measurable clinical endpoint.

The scout wanted us to say milk thistle does nothing. That is falsifiable on camera and we would deserve the correction. What is defensible is narrower and more damning: the label makes a claim the evidence base was never designed to support.

Sea moss: the giveaway is the denominator

"92 of the 102 minerals your body needs" fails before you reach the biology. The body has an established requirement for roughly sixteen minerals. The denominator is invented, and the ninety-two is an element count rather than a nutrition fact.

That does not make sea moss inert — which is exactly why we did not file it as a debunk. It is a potent iodine source. A 2026 European Journal of Nutrition trial measured habitual seaweed consumers at a median 658 µg of iodine a day, above the tolerable upper limit; six weeks after they stopped, intake was 189 µg and their TSH had fallen significantly, most in those who had been highest. You are buying one mineral in a dose you are not tracking, not ninety-two.

The magnesium scare: the number survived, the conclusion did not

We planned to debunk "70% of people are magnesium deficient". Then we found a serum-based NHANES 2021-2023 figure in The Journal of Nutrition: 67.8% of US adults fall below the 0.85 mmol/L cutoff for chronic latent magnesium deficiency. So the frightening statistic now has a citation.

What does not survive is what gets built on top of it. That cutoff is a research threshold rather than the clinical reference interval, which is precisely what "latent" means. Printing the 67.8% without that sentence would be the same scare-marketing we are criticising.

What we scored highly and refused to make

Two of the top three never became videos. Creatine-and-the-brain and ashwagandha-and-cortisol both already exist as videos we published in the last eleven days, and a duplicate inside a fortnight costs more in fatigue than a fresher, lower-scoring topic gains. The highest-scoring topic is not automatically the one that gets made.

Three more sit on the board that we have not published a verdict on yet. The board shows their scores anyway, because a research desk that only publishes its conclusions is a marketing department.

Why the whole board is public

Everyone in this category publishes verdicts. Almost nobody publishes the scoring that produced them, the topics that were rejected, or the reason. The board carries the real Ahrefs volume for each topic, the anchor study with its PMID, our 0-50 score across five factors, and whether the popular version of the claim survives its own evidence. Where our own ranking page contradicted a verdict, we fixed the page before publishing the verdict.

This is analysis, not medical advice.