How we work
What our guides are checked for, and what they are not
Our library is produced with substantial automation against a structured brief per topic, then put through the checks below. Where a piece has no expert reviewer, it says so on the page instead of borrowing a name.
The number
How much of the library carries a named human review
0 of 1,000
No published piece currently carries a named expert review. Every piece is pipeline-approved and every page states that plainly. This number is on the page because a review claim you cannot check is worse than no claim, and because we would rather publish a low figure than a false one.
Provenance
The two states a page can be in
Every guide is one of these, and the byline tells you which.
Pipeline-approved
The piece passed every automated check above. No named person with subject-matter standing read it and put their name to it. This is the majority of our library.
How to tell: The byline says the piece was not reviewed by a named expert, and links here. No reviewer is named, and no reviewedBy structured data is emitted for the page.
Reviewed by a named expert
A named reviewer from our register read the piece, and the specific things they checked are recorded and shown. Their name, role, and the date sit on the page.
How to tell: The byline names the reviewer, links to their profile, gives the review date, and lists what they checked. The page emits reviewedBy structured data pointing at that person.
The gate
What every published piece has to pass
Each check is paired with what passing it does not tell you.
Evidence is present, first-party, and dated
Every piece must cite at least two sources it depends on, and at least two of those must be Schoolyi product documentation pages describing behaviour that exists. Each source carries the date it was last checked. A piece whose evidence is undated or unreachable does not release.
What it does not tell you: It confirms the source exists and was checked on a date. It does not confirm the source still says what the piece claims it says, which is what a human review would catch.
Internal links resolve to indexable pages
Every internal link on a piece is checked against the live sitemap. Links to sign-in, sign-up, or pages we deliberately keep out of the index are rejected outright rather than warned about.
What it does not tell you: A link can resolve and still be the wrong link for the sentence it sits in. Relevance is not machine-checked.
Claims stay inside what the product does
Each piece declares the claim scope it was written against — its cluster and its primary query. Copy that drifts outside that scope, or that contains an unresolved placeholder, fails the gate. We separately maintain a list of capabilities we do not have, and marketing copy is audited against it.
What it does not tell you: The scope check is structural. It catches a piece talking about payroll in an admissions guide; it does not catch a subtly overstated sentence inside the right subject.
The piece answers its question directly, early
A piece must open with a direct answer rather than a preamble, carry at least four sections, reach a minimum length appropriate to its type, and end with a concrete next step rather than a summary of itself.
What it does not tell you: Structure is not substance. A well-structured piece can still be shallow, and the near-duplicate check below is how we find the ones that are.
No two pieces are near-duplicates of each other
Titles, descriptions, and normalised body text are compared across the whole corpus. Exact duplicates block the release. Near-duplicates above a similarity threshold also block it, because at scale the failure mode is not one plagiarised page, it is two hundred pages saying the same thing in slightly different words.
What it does not tell you: Similarity is measured on wording. Two pieces can be lexically distinct and still make the same argument.
Limits
Where our content is weakest
Stated here so you do not have to work it out from the pages.
Most of the library has no expert reviewer
The register of named reviewers is small, and the number of pieces carrying a named review is stated on this page as a real figure. We would rather show a low number than a false one.
The automated checks are structural, not editorial
They confirm that evidence exists and is dated, that links work, that claims sit in the declared scope, and that no two pieces are near-copies. They cannot confirm that an argument is good or that a nuance is right.
We publish no customer outcome figures
No piece in our library attributes a percentage improvement to a named school, because we hold no consented, baselined evidence for one. Our rollout walkthroughs are labelled as composites, not case studies.
Depth varies across the library
Some subjects are covered with genuine operational detail; others are covered adequately and no more. Where a page is thin we keep it out of the index rather than pad it, which is why our sitemap is smaller than our page count.
Editorial standards FAQ
The questions a reader should be asking about content produced at this scale.
Is your content written by AI?+
Our library was produced with substantial automation against a structured brief for each topic, then put through the checks described above. We are not going to characterise it as hand-written by a team of specialists, because that is not what happened. What we will stand behind is that the product behaviour it describes is drawn from our own documentation, that the evidence is dated, and that where we do not have a capability we say so.
Why did you remove reviewer names from your articles?+
Because they were not true. Every piece previously carried "reviewed by Schoolyi Editorial Team" and emitted reviewedBy structured data to match, on content no individual had reviewed. A review claim a reader cannot check is worth nothing at best, and is a deliberate misrepresentation at worst. We now name a reviewer only where there is one, and say plainly when there is not.
How do I know a claim about the product is accurate?+
Each piece lists the documentation it was written from, with the date that source was checked. If a guide and the documentation disagree, the documentation is correct and the guide is stale — tell us and we will fix it. Our changelog records when each capability shipped and what its limits are, and our pricing page states what we will not put a number on.
What is your correction policy?+
If you find an error, contact us with the URL and what is wrong. Factual corrections to product behaviour are made against the current documentation. When we correct a piece we update its modified date, which is what the page and the structured data both report.
How often is content updated?+
Each piece declares an update cadence — quarterly, annual, or event-driven — and event-driven means it is revisited when the underlying capability changes rather than on a timer. A piece that has not been touched in a long time shows its last modified date openly rather than being re-dated to look fresh.
How do I become a reviewer, or dispute something a reviewer signed off?+
Reviewers are named people scoped to specific subject clusters, with their standing stated on their profile. If you have relevant standing and want to review in a subject area, contact us. If you think a signed-off piece is wrong, say so with the URL: the reviewer's name is on it precisely so that the claim has someone accountable for it.
Found something wrong?
Send us the URL and what is inaccurate. Corrections to product behaviour are made against the current documentation.
Already using Schoolyi? Sign in