Research methodology

About this project · Using this research · Privacy information

What this database establishes

Records describe public provider information, not first-hand course reviews. Confirmed means the checked source supports the field. It does not establish teaching quality or ranking outcomes.

Member-reported coverage

Some topic tags are based on a course member’s confirmation of private materials. They are labelled member-reported in field evidence and exports, and are not independently verified. Topic filters and documented-coverage counts include these reports; missing information remains unknown.

Evidence and unknowns

Current course pages take precedence, followed by official syllabi and provider-controlled platform listings. Archived sources describe historical facts. Independent claims and inference do not silently become confirmed attributes. Conflicts remain unresolved until reviewed.

Equivalent concepts require a recorded editor, evidence summary and mapping reason. Absence remains unknown rather than no. Subject and product types are editorial classifications with a rationale.

Calculated measures

• AI coverage: documented yes fields (including labelled member reports) divided by five measured topics—GEO, AEO, LLM citations, entity SEO and AI measurement. Unknowns remain in the denominator.
• Evidence completeness: documented non-unknown values (including labelled member reports) divided by 13 core fields: price type, amount, format, level, duration, certificate, community, live coaching and the five AI topics.
• Verification age: elapsed whole days since the recorded check date.
• Programme count excludes guides, webinars and child products; resources are counted separately.
• List statistics use the stated qualifying-record denominator, including unknowns. Data changes propagate on the next build.

Selection and freshness

Price/access checks are due after 7 days, curriculum/features after 14 days and provider information after 30 days during the experiment. Changed sources trigger review sooner. Rebuilding does not advance verification dates.

Collections need at least three useful records before release review. Similar result sets are flagged for editorial examination. Overlap percentages are project heuristics, not Google thresholds.

Inspect sources · Download records