How We Research
AI Tools & Agents
This is the standard we hold our work to: claims checked against the vendor's own documentation, pricing pages and changelogs, and dated so you can see when we looked. Older articles are still being brought up to it. Here is how we research, what we verify, and what we will not assert without evidence.
From primary source to published guide
We start from primary sources
We read the vendor's own material first: official documentation, pricing pages, changelogs, release notes and status pages. We do not build articles by summarising other blogs, and we do not treat a marketing page as a specification.
We verify claim by claim
Material claims — pricing, usage limits, features, API availability, integrations — are checked individually against a primary source, and we record which source and the date we checked it. Where two sources disagree, we publish the disagreement rather than quietly choosing one.
We label the evidence level
Each article states how much evidence sits behind it, using the three levels below. We do not publish a numeric score for a product we have not used, and we say plainly when something could not be confirmed.
We re-check and date everything
AI pricing and features change quickly, and a fact that was right in January can be wrong by July. Pricing and limits are the fastest to rot, so they are re-checked most often. When we correct something we log it rather than editing quietly.
How much evidence sits behind an article
This is our internal editorial framework for describing the strength of the evidence behind a piece. It is not a rating, not a ranking signal, and not a score for any product. Where an article carries no level, we have not yet assessed it — and we would rather say nothing than imply a standard we cannot show.
About the existing archive: OneAppleFall has published for several years, and not every older article was written to this standard. We are working through the archive rather than claiming it is already done, and no article currently carries an L3 label because we do not presently perform hands-on testing. If an older page has not been re-checked, we would rather you knew that than assumed otherwise.
Desk Verified
Claims are checked against official documentation, pricing pages, changelogs and other credible public sources, with the source and the date recorded. Nothing is installed or run, and no first-hand use is implied. This is the level our current research is designed to meet.
Reproduced
A documented behaviour was reproduced through a free tier, public demo or other openly available path, and what we saw is shown. This confirms a specific behaviour exists; it is not a judgement on the product as a whole.
Hands-on Tested
Genuine first-hand use by OneAppleFall, documenting what was done, over what period, at what cost, and where it fell short. We do not currently publish at this level, and we will not label an article L3 until that changes.
Where the lines are
We do take money — through affiliate links and paid placements — so the honest question is not whether commercial relationships exist, but what they are allowed to change. These are the boundaries we hold, stated as policy rather than as a claim about the past.
We do run paid placements, and every one is labelled where it appears. What payment cannot buy is a favourable conclusion, a place in a ranked list, or the removal of a criticism.
We earn a commission on some links, and those links are disclosed. A tool we earn nothing from can still be the one we point you to, and the presence of a commission is never a reason to recommend something.
We do not currently test products hands-on, and we will not write as though we do. Where a claim comes from a vendor's documentation rather than from our own use, the article says so.
Vendors are welcome to flag factual errors after publication. They are never given pre-publication access to edit, moderate, or approve our findings.
Corrections & updates
We make mistakes, and this site has published claims it could not support. When we find or are told about an error, we verify it against a primary source and correct it with a dated note explaining what changed — rather than editing the page quietly and moving on.
Found an error? Email oneapplefall@gmail.com with "Correction:" in the subject line.
Methodology FAQ
Do you test the products you write about?
Do you give products a score out of ten?
What do you cover?
What happens when a product changes after publication?
How do affiliate links and paid placements work?
Can a company challenge something we have written?
Check the sources yourself.
This is the process we hold new work to. Older articles are still being brought up to it, and where one has not been re-checked we would rather you saw that than assumed otherwise.