Aethos AI

[feature] · 05 of 05

Catalog intelligence & product-kind scoring

Every product is graded as what it actually is — the kind is read from your own Shopify signals, carries its provenance, and decides which rules apply.

In the console: Product kind & applicability

Aethos Product detail: a product graded as a garment — the kind read from the Shopify product type, with the one-click override beside it

Captured from the running app on a demo store.

[method] How this is measured — the methodology →

What ships today

  • A kind with provenance

    A small vocabulary of kinds — only the distinctions that change how a product is graded — resolved from your own signals: native flags, Shopify taxonomy, product type. Every kind names its source on the product page.

  • Applicability, not a universal checklist

    Rules that do not apply to a product's kind are excluded from its grading and say why — a gift card is never asked for its fabric, and the exclusion is written on the receipt.

  • Fail-closed, never guessed

    When no source is sure, the product grades as unknown, against the universal basics only. The classifier that resolves harder cases is allowed to decide only because its precision was measured on a labeled corpus and cleared a published bar.

  • One-click override

    You know your catalog better than any signal. Change the kind on the product page and the override outranks every automatic source.

  • Basis-named re-scores

    When the grading rules themselves change, your next score is marked as a new basis and the cross-basis delta is suppressed — a rule change is never shown as catalog movement.

What this does not do
  • A kind is not a scoring bonus. It decides which rules apply — what the fair questions are — not how well a product answers them.
  • Automatic classification never overwrites your override, and when the automatic sources disagree or go quiet, the product falls back to the universal floor rather than to a guess.
  • The measured precision bar is the classifier's licence to decide. Below it, the model may only propose — and a proposal waits for you.

Also in the console

AI answers & prompt trackingAI-written fixes & Shopify pushAgent & crawler readinessCompetitor tracking & share of voiceAll features →

Graded as what it actually is

Connect Shopify and every product is graded on rules that apply.

Run the free brand checkSee pricing

[✓] Describes shipped behavior only