Rating Methodology — Why This Site Doesn't Publish Numeric Scores

This page exists mainly to say plainly what this site does not do: BestAIAgent.in does not compute, store, or display a numeric score — not an overall quality score, not a per-dimension score, not a composite ranking — for any agent, model, framework, or provider. There is no scoring function anywhere in this codebase.

Why no numeric scores

A numeric score creates an appearance of precision that the underlying evidence usually can't support. Whether an entity is a good fit for a specific task depends on requirements a static score can't encode — deployment constraints, pricing at your actual usage volume, compliance needs, language support. Presenting that as one number would trade honesty for a false sense of comparability. Earlier versions of this site did publish composite scores; that system was retired, and this page's job is to make sure that decision doesn't quietly drift back in through new content.

What's shown instead

• Verification status — whether an entity's identity has cleared the evidence gate (verified), is only source-linked, or remains pending/unknown. This is a binary identity check, not a quality judgment.
• Evidence receipts — the specific primary source, retrieval date, and content hash behind each verified fact, so you can check it yourself rather than trust a summary.
• Dated facts — specific fields (pricing, categories) that carry their own source and date, shown as facts, not folded into a composite number.
• Selection frameworks — pages like /best-ai-agents describe what to check for a given use case (e.g. latency for voice agents, license for self-hosting) without collapsing those criteria into one ranking.

What we refuse to publish

Consistent with not scoring, this site does not publish:

• AggregateRating, Review, or review-count structured data unless a page has genuine, checkable review data behind it (currently: none does).
• Adoption or popularity figures without an audited source.
• "Best in class" superlatives without the specific criteria and evidence spelled out on the same page.
• Latency, uptime, or cost figures without our own reproducible measurement or a cited primary source.

Related methodology pages

• Evidence methodology
• Comparison methodology
• Source classification
• Editorial methodology

Source receipts

• Internal — lib/catalog-types.ts and lib/catalog.ts. The real Entity/EvidenceRecord types and isEntityIndexable gate — neither includes a score field.
• Internal — README.md, "What's deliberately excluded" section. States the no-synthetic-rankings principle this page documents.

Last reviewed

• Page last reviewed: 2026-09-02