Same session
Recall a fictional fact, then correct it.
How we test
A claim needs a trail.
Readers should be able to see what was checked, what was observed and what remains unknown.
Catalogue source
WordPress is the authoring backend. The public site reads the last approved snapshot, never drafts.
- SourceWordPress snapshot
- Revisionwp-6
- Products8
- Approved offers0
Staging snapshot: labelled fixtures until an editor publishes merchant-tested records.
Three evidence states
Editor tested
A dated observation, plan, method and supporting material.
Vendor stated
A sourced provider statement that we have not independently confirmed.
Unverified
Missing, conflicting or stale information that cannot support a firm recommendation.
These are the production states. This staging build contains no completed product tests.
Memory checks over time
New session
Check recall without silently supplying the answer.
24 hours & 7 days
Repeat under documented conditions and the same plan.
30 days
Publish only after the elapsed-time test has actually completed.
Keep the conclusions narrow.
Returning to the same visible conversation can demonstrate continuity in that setting. It does not reveal a provider's internal memory architecture. Record model or plan changes and distinguish an unavailable feature from a failed test. Proposed score weights — conversation 25%, memory 20%, customisation 15%, media/voice 15%, cost clarity 15%, usability 10% — are unpublished until the rubric is complete. Missing categories are not renormalised.