Editorial policy
How we decide what gets published, and what does not
Our editorial rules, written as the process this site actually runs on: a required brief before drafting, a source behind every figure, a named person signing off, and a machine that is never allowed to publish.
What this covers
What we publish, and the test a page has to pass first
This site publishes three kinds of writing: pages explaining what we do, pages explaining how something in search or on the web actually behaves, and pages answering a question a business owner has asked us more than once. Nothing is published because a keyword tool found a gap.
The test is applied before drafting rather than after. Can we say something specific and true here that the reader could not already get from the results ranking above us? If the answer is no, the page is not commissioned. A great deal of what a marketing agency could publish fails that single question.
What follows is not a set of aspirations. Most of it is enforced by the content model this website is built from, and the enforcement takes the form of a failed build rather than a note in a document nobody opens.
The rule everything else hangs off
No brief, no page. Nothing is drafted here until someone has written down who it is for, what they want, what it has to prove and what it is not allowed to claim.
The brief is a required, validated structure rather than a document. It names the search intent, the buyer in their own words, their pain points and objections, the proof the page needs, the insights it will offer that a competitor page would not, its sources, and the claims that are forbidden on that page specifically.
It is enforced rather than encouraged. A brief with a missing field does not compile, and a brief padded with generic filler is obvious immediately, because the fields ask for things you can only write down if you have done the research.
Sourcing
How figures, claims and dates are handled
Most marketing statistics in circulation have been laundered. A figure appears in a blog post, is quoted without a link by a second, becomes "studies show" in a third, and by the fourth nobody alive can find the original. We do not take part in that.
- Every published figure carries a named source and a link to it. If we cannot link it, we delete it rather than soften it into "research suggests".
- Primary sources first: official platform documentation, standards bodies, regulators, government publications and public datasets. A vendor blog summarising a study is not the study.
- A figure is published with the period it describes, because a number about search behaviour is a snapshot rather than a fact.
- We do not publish a figure we produced ourselves unless the method behind it is published in the same place.
- Where something is our judgement rather than an established fact, the page says so in those words instead of borrowing the grammar of certainty.
- Every page records when it was created and when it was substantively updated, and the second date moves only when the substance moved.
Machine assistance
Where AI is used in our production process, and where it is not
| Dimension | What a machine may do | What a person must do |
|---|---|---|
| Research and clustering | Group queries, summarise a page of documentation, surface angles we had not considered. | Decide what the page is for, and read every summarised source directly before relying on it. |
| The brief | Propose a first pass at intent, structure and likely objections. | Write the buyer, the pain points and the prohibited claims. Those come out of conversations, not out of a model. |
| Drafting | Produce draft prose and alternative outlines from an approved brief. | Rewrite it. Nothing reaches a reader in the shape a model produced it. |
| Facts and figures | Nothing at all. A model is not a source and is never cited as one. | Verify every checkable claim against the primary source, or remove the claim. |
| Review and publication | Flag broken links, duplicate headings, invalid schema and near-duplicate copy. | Read the finished page and record approval. No automated check can grant it. |
The lines
What we commit to, and what we refuse
Independence is the part most editorial policies leave vague, so it is stated here in the same plain terms as everything else.
We commit to
- Publishing only pages a named person has read and is willing to defend in public
- Linking the source of every figure, including when the source is inconvenient for us
- Saying plainly when something is an opinion, an estimate or an early read
- Correcting errors visibly, with a date, rather than editing them away
- Keeping the difference obvious between a page selling something and a page explaining something
We will never
- Invent a statistic, a study, a benchmark or a survey result
- Publish machine-generated text that no person has rewritten and checked
- Recommend a platform, tool or partner because someone paid us to
- Accept a sponsored post, a paid mention or a paid link on this site
- Produce near-identical pages by swapping an industry or a city name
- Move an update date to make a page look fresher than it is
Corrections
What happens when we get something wrong
We will get something wrong. What matters is what happens next, so it is written down before it is needed.
You tell us
Email us the page and the sentence. You do not have to prove anything — if you think a claim is wrong, that is enough to make us go and check it. Anonymous reports are fine and are treated the same way.
You get: An acknowledgement, usually within two working days
We check it against the source
We go back to the primary source rather than to whatever quoted it. If that source has changed, moved or been withdrawn, the claim counts as unsupported until we find another one, not until we find a replacement we like better.
You get: A decision: correct, clarify, or leave it with an explanation
We fix it in the open
A factual error is corrected and a dated note on the page says what changed. We do not quietly delete a sentence and carry on. Where the error changed the meaning of the page, the note also says what the page previously claimed.
You get: A corrected page carrying a visible correction note
We withdraw it if it cannot be supported
If a claim cannot be substantiated and the page depends on it, the page comes down rather than being rewritten around the hole. Withdrawal is a one-line change to the approval register and it takes effect on the next deploy, without deleting anything.
You get: The page withheld from search, or retired outright
We tell you what we did
If you left us a way to reply, you get one — including when we concluded the original was right. That reply names the source we checked so you can disagree with it properly.
You get: A direct answer to whoever raised it
Questions
Questions about how this site is written
How do I report a mistake on this site?
Use the contact form, name the page and quote the sentence you think is wrong. You do not need to explain why or prove it — a page and a sentence is enough for us to go back to the source and check.
We acknowledge reports within a couple of working days. If the claim was wrong, the page is corrected with a dated note saying what changed. If it was right, you get a reply telling you which source we checked and what it says.
Do you use AI in producing this content?
Yes — for research summarisation, clustering, alternative outlines, editing suggestions and automated quality checks. No page on this site reaches a reader in the shape a model produced it.
The boundary is set out on this page rather than implied. A model is never treated as a source and never cited as one, and it cannot approve anything. The approval that makes a page eligible for search is recorded by a named person, in a file no automated process writes to.
What does "substantively updated" mean on your pages?
It means the substance changed: a claim was corrected, a section was rewritten, guidance moved because a platform moved, or an argument was revised. The update date on a page moves only for that.
Fixing a typo, swapping a link or restyling a component does not move it. Refreshing dates to make old content look current is a common tactic and we treat it as a mild form of misrepresentation.
Can a company pay to appear in your writing?
No. We do not accept sponsored posts, paid mentions, guest articles, affiliate placements or paid links on this site, and we do not sell recommendations inside client work either.
When we recommend a platform or a tool it is because we would use it ourselves. If a commercial relationship ever exists with something we write about, the page will say so in the body text, not in a footer disclosure nobody reads.
Why do so few of your pages include statistics?
Because most of the statistics in circulation in this industry cannot be traced back to a source we would be willing to stand behind. When a figure fails that test the sentence is deleted rather than hedged.
The pages that do carry figures link to primary sources — platform documentation, standards bodies, regulators and public datasets — and state the period the figure describes. It makes for fewer confident-sounding sentences and better ones.
Found something wrong on this page?
Tell us which sentence and we will take it back to the source. Corrections here are published with a date rather than edited away quietly, which is the only version of a corrections policy that means very much.
Last updated · Reviewed by Zubair Afzal