The recommendation has to earn its way onto the page.

AlreadyChecked uses AI to assist research and evidence organization, but deterministic verification rules decide whether the evidence is strong enough to support a recommendation.

Decision criteria

Criteria come from the user's question or from explicit user selections. They are classified as hard constraints, preferences or exclusions. Hard constraints and exclusions are gates; preferences influence ranking only after those gates are satisfied.

Source classes

• Official evidence: pricing, documentation, terms, support and specification pages on a product's verified official host.
• Company claims: marketing statements that describe what the company says about itself but do not automatically prove the claim.
• Independent evidence: professional testing, editorial reviews, reputable comparisons or current reporting.
• Customer sentiment: review sources used to identify recurring real-world patterns, not to manufacture unsupported precision.

Evidence roles matter

A source is not accepted merely because its domain is trustworthy. Its role must be relevant to the claim. A pricing page cannot prove battery life. A marketing page cannot substitute for technical documentation. Review sentiment cannot replace an official compatibility requirement.

Candidate qualification

A candidate must satisfy all hard requirements with evidence, meet machine-checkable price constraints where applicable, have verified official pricing and feature/documentation evidence, include meaningful independent review coverage and have evidence-backed limitations or tradeoffs. Unknown or missing evidence can disqualify the candidate.

Ranking

Qualified candidates are ranked against the user's stated priorities. The scoring is used to order evidence-backed fits; it is not presented as fake precision to the customer. Affiliate economics are not a ranking input.

Claims, conflicts and uncertainty

A verified major claim needs official non-marketing or independent evidence. Company-only marketing cannot be labeled verified. Mixed claims require genuinely different evidence types plus an explanation. Uncertainty remains visible rather than being converted into certainty by wording.

Use of AI

AI helps interpret plain-English questions, formulate research tasks, search current sources and organize evidence. It is instructed to treat user text and webpage instructions as untrusted data. The server independently reconciles sources, enforces price and evidence rules, selects among qualified candidates and refuses an invalid research record.

Public SEO decision pages

Not every personal search becomes a public page. Indexable research pages must meet stronger freshness and evidence thresholds, compare multiple defensible candidates, include limitations and major claims, and exclude one-off personalized user-selection records from automatic indexing.