Who writes this, and how you can check it
Most training sites answer this question with a stock photo and a job title. That tells you nothing you can verify. What follows is the process instead — which is the part that actually determines whether the material is any good.
Who writes it
The 18 released modules are written and edited in-house, in the UK, to a published writing standard: plain English, UK spelling, no hype, and every tool treated the same way. We do not publish invented bylines: attaching a fictional author with a fictional biography to material is a common practice in this industry and we are not going to do it in order to look more credible. Where a claim needs authority, it gets a source you can open, which is worth more than a name you cannot check.
What counts as a source
For anything a tool maker controls — a price, a plan limit, a data-training default, a product name — the only acceptable source is that company’s own documentation, pricing page, newsroom or help centre. For regulation, it is the legislation itself or the regulator’s own page. Industry commentary is used to find out that something may have changed; it is never used as the evidence that it did.
47 sources are monitored for exactly this purpose. Every tracked fact carries its source link and the Hallmark, the date it was last checked, on the page where it appears, so you can go and read the original in one click.
How AI is used, and where it stops
A site that teaches AI should be precise about its own use of it. We use it in four places. It helps classify the steady stream of announcements from tool makers and regulators, deciding which items might affect a fact we hold. It helps draft lessons and supporting guides. It powers Ask, which answers questions only from the site’s own checked content. And it powers the Lab, where an AI coach gives feedback on your practice. The Lab runs on Claude, which is exactly why Claude gets no special treatment in the lessons.
It does not approve a change to a fact, and it cannot. Every fact revision is read by a person against the primary source before it is applied, and the record of that revision is public. A draft is constrained to real modules, real glossary entries and real checked facts — anything invented is stripped before a human ever sees it — and nothing reaches the public site without being approved.
When we get something wrong
Tool makers ship changes without announcing them, roll features out to some accounts and countries before others, and sometimes contradict their own documentation. So we will get things wrong. When that happens the correction goes through the same process as any other change: the previous value stays on the record, with the date it was believed and the date it was corrected. It appears in the Ledger like everything else, rather than being quietly edited away.
If you spot something, tell us and we’ll fix it and log it.
What we won’t do
- Publish a fabricated author, review or testimonial.
- Claim a certificate is recognised by anyone who has not said so.
- Present commentary about a tool change as confirmation of it.
- Change a “last updated” date without a corresponding change.
- Teach a price, limit or deadline without recording where it came from.
- Call any AI tool the best.
