Editorial policy
Every entry on this site is made the same way. This page is that process, written down, so you can judge the output against a stated standard rather than a vibe.
The standard every entry has to meet
- A reader has their answer within five seconds, without scrolling.
- The definition is one sentence and avoids jargon that is not itself defined here.
- Every statistic, percentage or date is attributed to a named source.
- Where a metric belongs to a company rather than to Google, the entry says whose it is.
- Where something is contested or unknown, the entry says so instead of picking a side and sounding confident.
- No entry sells anything, recommends a purchase, or ends in a call to action.
Where the research comes from
Primary documentation first: Google Search Central, Google's own help centres, and specification documents such as the HTML and Schema.org specs. Where a practice is established but undocumented — much of link building, most of what happens after an algorithm update — the entry uses well-known industry references and names them, rather than presenting practitioner consensus as if Google had confirmed it.
Citations are restricted to a fixed allowlist of 269 publications, listed on the sources page. An agent writing an entry cannot cite outside that list; if a claim cannot be supported from it, the claim is cut rather than softened.
Where automated tooling is used
This site is large, and I am one person. Entries are drafted with language models against a fixed brief that is built before any drafting begins: the term, the section it belongs to, the exact word budget, the research extracted from primary sources, the real questions people search for, and the exact internal links the entry must carry.
The important part is what the tooling is not allowed to decide:
- It does not decide which entries exist. That comes from search demand analysis and a deduplication pass, so two entries never cover the same concept.
- It does not decide what is true. Facts come from the research pack; claims that cannot be supported from the allowed sources are removed.
- It does not decide the links. Every internal link and its anchor text is assigned before writing, so no entry can invent a URL.
- It does not decide length. The word budget is enforced, because padding is what makes reference sites unreadable.
Every draft then goes through automated checks before it can publish: word budget, one-sentence definition, all required links present exactly once, sources restricted to the allowlist, a named list of common false claims (for example, that Google uses Domain Authority), a filler-phrase scan, a reading-level floor on beginner entries, and a cross-corpus duplication check that rejects an entry sharing too much text with any other. A draft that fails a check is rewritten, not published with a warning.
The claims that are actively blocked
Some things are repeated so often in SEO writing that a model trained on the web reproduces them as fact. These are checked for specifically:
- Domain Authority and Page Authority are Moz metrics. Google does not use them.
- Google has repeatedly denied using Google Analytics data, bounce rate or dwell time as a ranking signal.
- There is no keyword density target that Google rewards.
- The meta keywords tag has been ignored since 2009.
- E-E-A-T is a concept from the Search Quality Rater Guidelines, not a score the algorithm computes.
- Third-party authority scores are vendor estimates, not Google's numbers.
Updating and decay
SEO writing rots. Thresholds change, tools are renamed, features are retired. Every entry carries the date it was last edited, and the recently-updated page lists all of them in edit order so you can see for yourself whether this site is maintained. Entries about things that change often say so in the text.
Corrections
Errors are fixed and logged publicly with a date on the corrections page. Nothing is quietly edited: if a factual claim changes, the change is recorded. That is the deal — I get to write a reference site alone, and in exchange you get to see when I was wrong.
What this site will never do
- Accept payment for an entry, a link, or a favourable description of a tool.
- Add advertising, affiliate links or sponsored content.
- Publish a statistic without a source.
- Use a fake byline, a fake reviewer, or an invented editorial board.
- Track you. There are no cookies and no analytics scripts — see privacy.