Trust & standards · 03
Editorial standards
Who writes what, what we refuse to claim, and how corrections are handled. These rules bind every study, report, comparison, guide and board commentary on the site. Machine-drafted material is labelled as such on the page, and the per-app summaries are additionally kept out of the index.
03
The rules
- 01
Every claim ties to a figure we computed
If a sentence in a study cannot be traced to a number in the index, it is cut. Where a claim rests on judgement rather than measurement, it is written as judgement.
- 02
What is machine-drafted, what is human-reviewed, and how to tell
Studies and field guides are drafted with the assistance of a language model against figures computed from our index, and reviewed by our named author.
Commentary on our ranking boards and category reports is machine-drafted from the computed figures on the page it appears on, and is labelled as such on every page that carries it. It is not published unedited either: each draft passes an editorial gate that our named author defines and maintains, and that rejects the draft outright rather than publishing it if any of the following is true — it states a figure our data does not support, it reuses another page’s argument, it reads as drafting notes rather than finished prose, it stops mid-sentence, or it falls below our length floor. A draft that fails is withheld and the page carries no commentary at all until a later attempt passes. These pages are also excluded from advertising.
The per-app summaries are machine-drafted, labelled, and excluded from the index. Every figure in any of these, however drafted, is computed from our index and never estimated. We state which is which rather than implying a level of review we do not perform.
- 03
Nothing on this site is a review
We do not test apps in a lab. Comparisons state what was measured and what was not, and never imply hands-on testing we did not do.
- 04
Rankings are not for sale
No publisher can pay for a position, a mention or a recommendation. We use no affiliate links. Where a paid placement exists anywhere on the site it is visually separated and labelled.
- 05
Dates on everything
Every figure shows when it was read; every article shows when it was published and last revised. A page with no date is a page you cannot check.
- 06
Errors are logged in public
Corrections are recorded below with the date, what was wrong, and what changed. We do not silently edit published figures.
03 · continued
Sourcing & accuracy
How topics are chosen
Guides answer timeless, practical questions — the kind a real person asks before downloading, paying for, or deleting an app. We deliberately avoid topics that go stale quickly, so a guide you read today should still be accurate a year from now, even as the apps on our record boards change.
How we research and write
Each guide starts from a single clear question. We use AI tools to help draft and structure articles, and every guide is then reviewed against these rules before publication: plain language, no filler, no fear-mongering, durable accuracy. Guides carry the byline of Dor Bass, our founder, who is accountable for these standards and the record boards behind them. That byline reflects editorial ownership, not a claim of personal, hands-on experience with every app we mention.
How we handle sources
We list our sources at the bottom of every guide, strongly preferring official and primary sources such as Apple’s own support and developer documentation, alongside well-known, reputable publications. We prefer not to lean on forums, video pages, Q&A threads, or content farms, and we check that every link is live before it ships. Sources are further reading, not a guarantee — verify anything important for yourself.
Accuracy
App details, prices, ratings and policies change often, and our boards and studies recompute automatically so the figures stay current; each page shows when it was last updated. When we get something wrong we fix it — email [email protected] or use the contact page, and where a correction is material we note what changed. Always do your own research before acting on what you read here.
04
Corrections log
-
2 Sep 2026
The 31 August entry described a fix that had not shipped
That entry stated that board commentary “is now also regenerated when the leader’s value drifts.” An external audit checked it against production and found it was not true: on /record/most-prolific-developer the commentary still said the leader held 300 apps while the table beneath it showed 438, and named a different company as first three sentences later. Announcing a remediation we had not made is a worse failure than the original defect, because the corrections log is the one page a reader uses to judge whether the rest can be trusted.
The cause was narrower than the 31 August entry assumed. Regeneration was triggered by comparing a stored copy of the leader’s value against the live one; those two agreed to within two per cent, so nothing fired — while the sentence a reader actually saw quoted a third number that matched neither. The check was reading a stored figure, not the published prose.
Two things ship today. Commentary is now re-checked against the board it describes at the moment the page is built, and any commentary that awards the top position to an entity that is not first on that board is withheld entirely rather than shown with a caveat — a dated banner is an honest way to present figures that have moved, not a sentence naming the wrong company. Where commentary is withheld the page falls back to figures computed directly from the rows, or carries no commentary at all. Applied across every board, this removed the commentary from two of thirty-nine and left the rest unchanged. Separately, our field guides now carry a dated notice when the board they describe has since changed leader, or when a qualifying-pool figure they quote no longer matches the board.
We have also corrected Standard 02. It previously said studies and guides are “read and edited by our named author before publication.” That overstated the process: there was no per-page sign-off behind it. The standard now says they are reviewed by our named author, which is what actually happens — the interpretive claims across every indexed study and guide were read and accepted on 2 September. It no longer claims an editing pass we do not perform.
-
31 Aug 2026
The 25 August correction below was itself inaccurate
An internal audit found that the reinstated review described in the 25 August entry was not reliably happening. On /record/most-prolific-developer the commentary named three different leaders in four sentences and disagreed with the table printed beneath it, while carrying a human-review byline. The cause was not a missing review but a missing invalidation: board commentary was only regenerated when the leading app changed, so an essay quoting the leader’s figures went stale whenever those figures moved and the leader did not. Commentary is now also regenerated when the leader’s value drifts. Separately, our category reports were found to contain drafting notes from the language model that assembles them; the generator now rejects that output and those reports have been rebuilt. We have narrowed Standard 02 to describe what the process actually guarantees. Both earlier entries are left below unedited.
-
25 Aug 2026
Author review of indexed board commentary reinstated
After the April change, commentary on our indexed ranking boards is now read and edited by our named author before publication, so the human-review disclosure in Standard 02 is accurate for the indexed set. The purely computed per-app summaries remain machine-drafted and are excluded from the index.
-
7 Apr 2026
Board commentary claimed editorial review that had not happened
Board analyses described themselves as “editor-reviewed.” They are computed from the figures on the page and presented as analyst summaries; the claim of human review was removed sitewide and this standards page rewritten to match reality.
Report an error
Send the page, the figure, and what you believe it should be. We reply to every report, including the ones where we turn out to be right.