Verified public
A current first-party product, help-center, legal, or app-store source supports availability.
DEARHIM RESEARCH · METHODS
How the AI Companion Index
separates evidence from claims.
The evidence hierarchy, status definitions, correction rules, and testing boundaries behind DearHim’s living AI companion dataset.
CURRENT STATUS
Public-source verification is live. Behavioral memory results remain pending until delayed tests produce preserved evidence.
DIRECT ANSWER
Evidence status is not a product-quality score.
The Index records what current first-party evidence supports. It does not convert a marketing claim, help-center statement, or app-store listing into a finding about quality, reliability, safety, or relationship value.
STATUS DEFINITIONS
Four labels. No hidden grade.
Public claim
The product makes the claim, but DearHim has not independently validated the behavior.
Not verified
No adequate first-party evidence was found. This is not proof that a feature is absent.
Pending test
A behavioral protocol exists, but no valid completed evidence supports a result yet.
EVIDENCE HIERARCHY
Trace material facts to the source closest to the product.
- Prefer current official product pages, help centers, legal policies, and app stores.
- Use independent reporting to identify questions—not as a substitute for current first-party product facts.
- Date each review because plans, prices, models, settings, and regional access change.
- Keep checkout-required or in-app-only fields visibly unresolved until they are observed.
CONFLICTS & CORRECTIONS
Disagreement stays visible.
When official sources conflict, the Index preserves the conflict, dates the evidence, and marks the field for checkout or in-app verification. Material corrections record the previous value, new value, source, date, and reason.
TESTING BOUNDARY
A source can verify access. It cannot prove experience quality.
- Public evidence cannot establish memory accuracy, latency, emotional continuity, call quality, safety, or clinical benefit.
- Unsupported channels are marked not available, not scored as memory failures.
- DearHim is evaluated under the same behavioral rules and receives no home-field scoring advantage.
- The external research below informs the dimensions; it does not academically validate DearHim’s consumer protocol.
SOURCES
Trace the framework.
LongMemEvalLong-term memory abilities, timestamped histories, and public benchmark construction.LoCoMo · ACL 2024Very long-term conversational memory evaluation across multiple tasks.NIST AI RMF CoreContextual, documented, repeatable measurement and ongoing review.Google Dataset guidanceDataset and downloadable distribution structured-data guidance.
DEARHIM IS THE AI COMPANION
The method explains the evidence. Your AI boyfriend is the relationship.
DearHim is a full AI companion for adults who date men—including straight women and gay men—built around continuity, real conversation, and a companion that keeps the context.