How we work
What we will and will not tell you about sleep
Sleep is a subject with an unusual amount of confident nonsense written about it, much of it by companies selling something. We sell something too. So rather than asking you to assume good faith, this page sets out the rules every page on this site is written under, including the ones that cost us traffic.
In short
Every page here is written by the team that builds Tock. We do not employ a clinician and no page is medically reviewed, which is stated plainly rather than papered over. We do not invent citations, we do not claim to measure what we can only estimate, and we do not diagnose anything.
- Invented citations
- 0Invented citationsA fabricated study is unrecoverable
- Medical claims
- 0Medical claimsNo diagnosis, no promised outcome
- Small-sample figure
- 1Small-sample figureLabelled as such on both pages
The conflict of interest, stated first
This site exists to sell a subscription. Every page has a button on it. That is not a disclosure buried at the bottom of a page, it is the first thing worth saying, because everything else on this page is only meaningful if you know it.
The specific risk it creates is overclaiming: a sleep app has an obvious incentive to imply its sensors know more than they do, that its recommendations are stronger than the evidence is, and that a phone on a nightstand is a substitute for a clinic. The rules below exist because those are the exact temptations, and several of them have cost us pages that would have ranked.
What we will not write
These are hard rules rather than preferences, and two of them are enforced by the build rather than by anyone remembering.
Never, on any page
- No invented citations. A claim is either attributable to something a sleep researcher would recognise without a link, or it is written without one and hedged. We do not cite studies we have not read.
- No medical claims. Nothing here diagnoses, screens for, rules out or predicts any condition, and no page promises a health outcome. Tock does not detect sleep apnea and the glossary entry for it says so explicitly.
- No claiming to measure what we estimate. Sleep stages on a phone-only night are inferred from a microphone and an accelerometer. That is an estimate. The app enforces this in its own code and this site is held to the same line.
- No feature we do not ship. There is no Apple Watch app. The only true version of that claim is that Tock reads a Watch’s data through Apple Health, and every page that touches it says the second thing rather than the first.
- No comparison written to win. Every competitor page states what that product genuinely does better, every comparison table has rows the other side wins, and six of the thirteen verdicts send the reader elsewhere.
What we do have that other sleep sites do not
The counterweight to the conflict of interest above. Being the people who build the thing is a real advantage on exactly one axis, and it is worth being precise about which.
First-hand, and verifiable against the app
- Mechanism detail from inside the product. How the smart wake window actually works, on which two sensing channels, and what happens when it finds nothing. That is not summarised from anywhere; it is how the code behaves.
- The reasons behind design decisions, including the unflattering ones. Why the streak is a moon rather than a flame, why the sounds are generated sample by sample rather than looped, and what that costs.
- Honest limits, because we know where they are. Every feature page carries a "where it stops" section, and those are more specific than a reviewer working from the outside could write.
- Our own testing figures, labelled as what they are. The replay-testing number appears on exactly two pages and is described as small-sample internal testing every time it does. It is never called a study.
How sources are handled
Where a number comes from a professional body, it is quoted in the form that body published it and the body is named in the sentence. The sleep duration figures acrossthe sleep by age pages are the American Academy of Sleep Medicine consensus statements, given as ranges because that is how they were issued, and never collapsed into a single friendlier number.
Where something is well established but not attributable to one paper, it is written without a citation and phrased at the confidence the evidence supports. You will see a lot of "generally", "usually" and "for most people" on this site, and those are not filler. They are the difference between what is known about populations and what is true of you.
Where the evidence is genuinely thin, the page says so and does not sell around it. The clearest example is binaural beats, which is a page arguing that a category of sleep audio does not do what it claims, published by a company with a sleep audio library. It exists because writing the opposite would have been easier and would have been wrong.
What we do not have
The standard move for a health-adjacent site is a named clinician on the byline. We do not have one, and inventing one would be worse than saying so.
Gaps we are not papering over
- No medical reviewer. No page here has been reviewed by a clinician, and none of them claim to have been.
- No original research. We have not run a study. Our own replay testing is seventeen tracked nights, which is a useful signal and not evidence, and it is described that way wherever it appears.
- No sleep laboratory. Everything about staging on this site is compared against what a polysomnogram would show, and we have never run one.
- No public ratings to quote. The App Store listing has no ratings yet, so there is no aggregate rating in our structured data. Asserting one we cannot corroborate would be a trivial thing to fake and we have not.
How this is enforced rather than intended
A published standard that nothing checks is a marketing page about standards. Three of the rules above are enforced mechanically, and the build fails rather than warns when they break.
A post-build lint reads every rendered page on the site and fails on a duplicate title, a duplicate meta description, a page under the word floor, a missing canonical, an internal link that resolves to nothing, or an em dash in metadata. A schema check rejects content whose short answer runs past forty five words or whose description falls outside the range a search result will actually display. And every programmatic page derives its links from the same data it renders from, so a page cannot appear in a rail without existing.
None of that makes the writing true. It makes a specific set of failures impossible, which is a smaller claim and the only one worth making.
Tell us when we are wrong
The correction route, and the two pages that document the policy behind all of this.
Questions people ask
Who writes the content on this site?
The team that builds Tock, which is Wildframe. There is no freelance content operation behind it and nothing here is generated and published without a person deciding it is true. That means the mechanism detail is first-hand and it also means we have an obvious commercial interest, which is why the rules on this page exist.
Is this content medically reviewed?
No. We do not employ a clinician and no page here carries a medical review. That is a genuine limitation and it is why nothing on this site diagnoses, screens for or rules out any condition, and why anything that reads as a symptom is routed to a doctor rather than answered.
Why are there so few citations?
Because we only cite things we have actually verified. A claim is either attributable to something a sleep researcher would recognise without needing a link, or it is written without a citation and hedged accordingly. A fabricated reference on a health-adjacent site is unrecoverable, so the rule is absolute.
Do you correct pages when you get something wrong?
Yes, and we would rather hear about it than not. Email [email protected] with the page and what is wrong with it. Editorial pages carry a visible last-reviewed date that matches the dateModified in their structured data, so a correction is visible rather than silent.
The app is held to the same rule as the writing
Tock estimates sleep stages on a phone-only night and says so in the app rather than only here. Where it cannot know something, it does not draw it.