METHODOLOGY · V1.2
How we decide what is worth recommending.
AI Soul Map separates product facts, editorial judgment and user fit. A product must first pass a quality threshold; matching a need or category cannot make an unqualified product recommendable.
01 — TRUST LEVELS
Listing and recommendation are different decisions
Candidate
An internal research lead. It is not public and carries no endorsement.
Listed
A public bilingual profile with current core facts, usable media and at least medium evidence confidence.
Recommended
A Listed product with high overall evidence, complete fit boundaries and evidence for the capabilities behind the recommendation.
Editor’s Choice
A future, deliberately scarce distinction requiring structured first-hand experience and independent review. It is not currently inferred from a featured flag.
02 — EVIDENCE
Four evidence layers
Official sources
Product sites, pricing, documentation, policies, release notes and official app stores. Marketing claims prove only that the vendor made the claim.
Support evidence
Reproducible answers from official support or moderated official communities, with date and context.
Editorial tests
Recorded platform, version, plan, language, steps and observations. A test is not generalized beyond its conditions.
Independent sources
Regulators, research, audits and third-party testing with a stated method. Fast-changing facts still require a current primary source.
03 — EDITORIAL REVIEW
What every review must answer
What is verifiably available now?
Who is it best for, and under which constraints?
Who should avoid it or consider an alternative?
What are the meaningful strengths, caveats, privacy and content boundaries?
Which claims remain unknown, disputed or due for re-checking?
04 — WHO TESTS
Who does the testing
Research tools help us track changes and organize public evidence, but they do not write our conclusions. Before a product is listed, a member of our editorial team opens it directly, uses it across multiple sessions rather than a single visit, and records what they actually observed—platform, version, plan and steps included. "Best for" and "not ideal for" judgments are written by that person, not generated automatically.
05 — FRESHNESS
Freshness changes the conclusion
Links, availability, pricing, trials and age boundaries are checked most frequently. Capabilities, languages, platforms, privacy and safety facts have their own review periods. Expired evidence can remove a recommendation reason, downgrade a product or return it to review.
06 — INDEPENDENCE
Commercial independence
Affiliate links affect only the outbound destination and disclosure. They do not affect taxonomy, inclusion, ordering, recommendation eligibility or editorial language.
07 — SAFETY
Professional and crisis boundaries
Emotional, health and financial topics are framed as reflection or information support—not diagnosis, treatment, crisis intervention or professional advice. High-risk states should lead to local emergency or professional help, not a commercial recommendation funnel.