Methodology
Research and review methodology
What “AI-assisted desk research” means here, step by step, and what it does not mean.
1. Choosing subjects
Many pages started as a question: other websites still link to an old page on this domain, so what was that page about, and can we responsibly publish something useful on the same subject? We also add new guides and records that help readers and book clubs.
2. Desk research
An AI research agent gathers evidence from publisher and author pages, library catalogues (such as Open Library), archived copies of web pages (the Internet Archive’s Wayback Machine), official help documentation and reputable reviews. Each claim is recorded with the URL that supports it.
3. Independent fact-check
A separate AI review pass, given the draft and its sources but not the researcher’s reasoning, re-checks bibliographic facts (title, author, year, ISBN, publisher and edition), the neutral premise, genre and audience labels, every link, and anything that might be outdated. Unsupported claims are removed or corrected. Items that cannot be verified are not published.
4. What we don’t claim
- No page claims to be human-reviewed, expert-approved, personally read or hands-on tested.
- Device steps in our ebook guides come from official documentation, not from testing every device ourselves.
- Reading notes describe themes and conventions that sources support; they are not personal reviews.
5. Records of review
Every published item shows when it was researched and checked, and lists its sources. Review findings and corrections are kept in an internal verification log that accompanies the site’s source code.
6. Keeping pages current
Each guide shows an “updated” date. When devices, services or editions change, we update the page and the date, or note what changed.