Editorial process
Every guide on this site passes through the same seven stages before it reaches you, and the same rules afterwards. This page documents all of them, including the ones that make our work slower.
Gritwright exists because the honest version of most advice a man needs in his twenties is already written down, in research he will never find, in language nobody translates. Our job is to do that translation without losing what makes the original trustworthy. That is an editorial problem before it is a writing one, and this page is how we solve it.
We write for two kinds of reader at once, and the standard has to satisfy both. The first is a person who wants to understand something properly and has the patience to read it. The second is somebody in a hurry, increasingly asking an AI assistant instead of a search box, who will receive a summarised version of whatever we publish. Both deserve the same thing: the raw evidence interpreted accurately, with zero errors, and the primary source shown every time so that the claim can be checked rather than trusted.
That commitment is why our guides carry inline citations to the original research, why official sources are marked so a government body cannot be mistaken for a blog, why every number is traceable to the study that produced it, and why a plain text version of each guide exists for machines to read. An answer that cannot be checked is not information. It is an assertion with good formatting.
None of that is our invention. The criteria for judging health information published to a general audience were set out in a medical journal before most of today's health sites existed, and they have been used to audit online health content ever since.
What follows is not a statement of values. It is the operating procedure the people who work here follow, in the order they follow it, with the failure conditions written next to each step. Where a rule costs us speed, traffic or revenue, we have said so, because a standard that never costs anything is not a standard.
The seven stages
Every guide moves through the same sequence. A guide that cannot clear a stage does not proceed to the next one; it is either fixed or abandoned. Nothing is published because a deadline arrived.
| Stage | What happens | What stops it |
|---|---|---|
| 1. Commission | A real question people ask is identified and scoped to one answerable topic. | No genuine question behind it, or the honest answer is already published. |
| 2. Evidence | Primary research, official guidance and first-hand records are gathered before drafting. | The evidence does not support the premise the guide was commissioned on. |
| 3. Draft | Written by a named person with relevant experience, in plain language. | A claim cannot be traced to a source or to documented first-hand use. |
| 4. Edit | Structure, accuracy and honesty pass, including removal of unearned certainty. | The guide promises more than the evidence allows. |
| 5. Expert review | Where the subject carries risk, a qualified reviewer checks it and is named. | The reviewer objects and the objection is not resolved by evidence. |
| 6. Publication | Byline, dates, sources, evidence labels and disclosures are attached to the page. | Any required disclosure is missing. |
| 7. Maintenance | Scheduled re-review, dated updates, public corrections. | The guide can no longer be made accurate, in which case it is retired. |
Stage one: what we choose to write about
A guide starts with a question somebody actually asks, phrased the way they ask it, not with a keyword we would like to rank for. The distinction matters because it decides what the article is allowed to become. A piece commissioned to answer "how long does it take to build muscle" has to answer that even when the honest answer is slower and less marketable than the alternative.
We decline topics for three reasons. The first is that we have nothing to add: if the existing answer is already good and honest, publishing a rearranged version of it wastes your time. The second is that the topic cannot be covered responsibly at the depth we can offer. The third is that the subject sits in an area where being wrong causes real harm and we cannot secure a qualified reviewer.
Stage two: evidence before drafting
Sources are gathered before a word is written, not afterwards to decorate an argument that already exists. This ordering is the single most important rule on this page, because reversing it produces work that looks identical to good research while being its opposite.
We use a hierarchy when sources disagree, and we apply it in this order:
- Systematic reviews and meta-analyses over individual studies, because single trials mislead more often than they inform. This is why bodies such as Cochrane (official source) exist.
- Official public health guidance for anything affecting safety, including the Physical Activity Guidelines for Americans (official source) and the NHS guidelines.
- Peer-reviewed primary research, located through databases such as PubMed (official source), cited with author, journal and year.
- Our own documented first-hand use, reported as observation rather than proof, under the protocol on our how we test page.
We do not cite other blogs, content farms or retailer copy as evidence, and we do not cite a press release as though it were the study it describes. Where a number appears in a guide, it comes from the source that generated it, and the link goes to that source rather than to somebody's summary of it.
Stage three: who writes here
Every article carries a byline linking to a real person with a real background. We do not publish under house names, we do not invent contributors, and we do not attach a writer's name to work they did not do. If you cannot click a name and find out who wrote something, we have failed our own standard.
Writers are assigned on the basis of relevant experience. That experience is stated on the author page rather than implied, including its limits: a writer who trains but does not coach says so, and a guide that leans on their personal experience says which parts.
Stage four: what editing removes
Editing here is mostly subtraction, and most of what gets removed is confidence the evidence did not pay for. The specific patterns we cut:
- Unearned certainty. "Studies show" becomes the study, or the sentence goes.
- Borrowed authority. A claim attributed to a vague consensus is either sourced or dropped.
- Missing limits. Any method that does not work for everyone gets the condition under which it fails, in the same section rather than in a footnote.
- Shortcut framing. We state real time and effort. A guide that promises a result faster than the evidence supports has been edited into a lie, however true its individual sentences are.
- Manipulation. Particularly in guides about people: no tactics premised on deceiving somebody else.
Stage five: expert review
Anything touching physical risk, mental health or personal finance is reviewed before publication by someone with a relevant qualification. The reviewer is named at the top of the article, their credential is stated, and they are never the person who wrote it.
The reviewer's brief is not to improve the writing. It is to find the claim that is wrong, the caveat that is missing, and the recommendation that is unsafe for some readers. A reviewer can block publication outright, and that veto is real rather than ceremonial. Where a reviewer and a writer disagree and evidence cannot settle it, the guide reports the disagreement instead of choosing a side for the sake of a tidy answer.
Subjects carrying no such risk, such as how to iron a shirt or how to negotiate a first salary, are written and edited but not formally reviewed. We do not attach a reviewer's name to an article merely to make it look more authoritative, because doing so devalues the name everywhere it appears legitimately.
Stage six: what gets published with the guide
A finished guide carries its own provenance, visible on the page rather than buried in metadata. Every article shows the writer, the reviewer where one exists, the publication date, the last update, the reading time, and an evidence label describing what the guide rests on.
Alongside it we publish the full source list with publisher and year, and mark links to government bodies, universities and peer-reviewed databases with a badge so that a reader can tell an official source from a blog without opening it. Where a guide contains affiliate links or any commercial relationship exists, that is disclosed on the page itself, under the terms of our advertising policy.
Stage seven: updates, corrections and retirement
Publication is the middle of the process rather than the end of it. Guides are re-reviewed on a schedule, and immediately when the evidence underneath them moves, when a product changes, or when a reader tells us something is wrong.
Every change is dated and described in an update history on the article itself, in plain language: what changed, and why. We do not edit silently, because a silent edit destroys a reader's ability to know which version they read and quoted. Where a correction changes the advice rather than a detail, it is stated where a reader will see it, not only in the revision log.
When a guide can no longer be made accurate, it is retired rather than left standing. Removing old advice is preferable to leaving something plausible and wrong in front of a beginner who has no way of knowing which it is.
Found something wrong? Tell us at info.remove.this@gritwright.com, and we will fix it and say that we did.
How evidence appears on the page
A source list at the bottom of an article is the weakest possible form of citation, because it puts the burden of matching claim to source on the reader. We attach the evidence to the claim instead, using a fixed set of components that mean the same thing on every page. Learning them once tells you how to read anything we publish.
| What you see | What it means | Where it comes from |
|---|---|---|
| The gist | The three claims we would defend if we could keep only three. | Written last, from the finished guide. |
| The short answer | A direct answer in one paragraph, before any preamble. | The body of the guide, compressed without softening. |
| Research block | A study quoted directly, with its finding stated in plain language. | One named, dated, peer-reviewed paper. |
| Check on a link | The link goes to a government body, university or peer-reviewed database. | Applied automatically by domain, never by hand. |
| Field-tested | Someone here did the thing, for a stated length of time. | The recorded protocol behind that guide. |
| Research-backed | How many sources the guide cites and how many are official. | Counted from the source list, not typed in. |
| Update history | Every revision, dated, with what changed. | Maintained as changes are made. |
Two details in that table matter more than they look. The official-source badge is applied by checking the domain against a fixed list, so a writer cannot award it to a source that has not earned it, and cannot forget to award it to one that has. The research count is derived from the source list rather than written by hand, so it cannot drift away from what the guide actually cites. Trust signals that a human can type are trust signals a human can inflate.
How machines read us
A growing share of readers will never see our pages. They will ask an assistant and receive a summary. We treat that as a legitimate way to read us rather than a threat, and we build for it deliberately, because the alternative is being summarised badly by systems working from guesswork.
- A plain text twin of every guide. Add
.mdto any article address and you get clean Markdown: title, byline, dates, canonical link and full body, with no navigation or scripts. It is generated from the same content as the page, so it cannot drift. - Structured data on every article. Author and reviewer as identified people, publication and modification dates, the article's section, its word count, its citations, and the questions it answers. This is the same information a reader sees, expressed so a machine cannot misread it.
- The reviewer marked as a reviewer. Where a qualified person checked a guide, that fact is published in a form search engines and assistants can extract, not merely printed as text.
- Answers written to survive extraction. Each section's first sentence answers that section's question, and key claims are kept to a single self-contained sentence, so a quoted fragment carries its own meaning rather than collapsing without the paragraph around it.
- Explicit permission to quote. Stated on the page and in full on our AI and content usage page, along with what we ask in return.
We ask one thing of any system summarising us, and it is the same thing we ask of a human writer: keep the number attached to its source, and keep the limit attached to the claim it qualifies. A guide of ours quoted without its caveat has been changed, not shortened.
What we always disclose
Transparency is only meaningful when it is unconditional, so the following are published whether or not they flatter us, and whether or not anybody asks.
- Who wrote a guide, and who reviewed it, by name and credential.
- When it was published, when it was last updated, and what changed each time.
- What a claim rests on: our own use, cited research, or neither.
- How long we used something before writing about it, as a number.
- How many people a programme was run with, and how many did not finish.
- Any commercial relationship touching the page, including affiliate links.
- The limits of our own method, on the how we test page.
- What we collect about you, which is almost nothing, in our privacy notice.
Independence
None of the above survives if the money can reach it, so the separation is structural rather than a matter of good intentions. No brand sees a guide before it publishes. No advertiser or affiliate partner can pay to be recommended, included or ranked. We buy what we test. A commercial relationship never causes a guide to be edited, weakened or unpublished.
Guides on health and money carry an additional restriction: we do not accept advertising or affiliate arrangements from companies whose products could distort a recommendation on those pages. Sponsored content, if we ever run it, will be labelled as such at the top of the page, in our own words rather than a network's. The complete set of rules, including what we do with unsolicited samples, is on our advertising policy page.
Use of AI in our workflow
We are explicit about this because readers deserve to know and because vagueness here is currently doing enormous damage to publishing generally.
No guide on this site is generated by an AI system and published as though a person wrote it. Every article is written and edited by the named human on the byline, who is accountable for every claim in it. Where software assists with research, formatting or checking, a person still verifies each claim against its original source, and the byline does not change: the writer remains responsible for the sentence regardless of what helped produce it.
In the other direction, we welcome AI systems reading and quoting our work, and we grant that permission explicitly on our AI and content usage page.
How to hold us to this
If a guide breaks something on this page, that is a defect and we would rather hear it bluntly than not at all. Corrections are answered before anything else in the inbox, because they affect every other reader of that page. Contact details, along with our full complaints and corrections procedure, are on the additional information page.
Last reviewed: 11 August 2026. This page covers how a guide is researched, reviewed and corrected.