CITAQ

Methodology

How Citation Readiness is scored.

The score isn't a vibe and it isn't a black box. This page explains what the 0–100 Citation Readiness Score measures, the seven checks behind it, and the one thing we deliberately don't publish — in the same plain language your reports use.

Inside the engineproprietary
check_weightingswithheld
safeguard_thresholdswithheld
archetype_field_mapswithheld
translated for merchants
78 AI-Strongper product

What's limiting the score, and what to fix first — in plain language, with the reason attached.

01

What the score measures

The Citation Readiness Score measures one thing: how ready a single product is to be found and cited by AI systems — ChatGPT, Perplexity, Gemini, Google AI Overviews, and marketplace assistants. It is not a domain-authority score, not an SEO grade, and not an average of your store.

That granularity is the point. AI assistants recommend products, not domains — and a store-level number hides exactly the products that are silently invisible. A catalog can look healthy on average while its best sellers never appear in an AI answer.

The unit of scoring

One product page in. One 0–100 score out. Every SKU on its own.

Each score comes with the reasons behind it: what was found, what's missing, what's limiting it, and what to fix first — never a bare number.

02

Two dimensions: found, then quoted

Citation is a two-step problem, so the score reads out along two dimensions. A product can fail at either step — and the fixes are different.

R

Retrieval Readiness

Can AI systems find, access, and index this product at all? Crawler access, structured data, and identity clarity live here. A product that can't be retrieved can't be cited — no matter how good its content is.

C

Citation Readiness

Once found, is the content concrete, consistent, and clear enough to be worth quoting in an answer? Vague copy, thin descriptions, and conflicting facts cost you here — an AI won't repeat what it can't trust.

03

The seven checks

Every product runs through seven checks. Each answers one question a buying assistant implicitly asks before citing a product. What follows is what each check looks at — how they combine into the final number is part of the engine (see what we publish).

01

Attribute completeness

Are the product's concrete attributes present — and structured?

Material, dimensions, size, capacity, color — the specifics a buyer would ask about. AI systems can only cite details they can actually find on the page, in a form they can parse.

02

Entity definition

Can an AI system tell exactly what — and whose — this product is?

Brand identity, product identifiers, and clear price signals. A product an AI can't confidently identify is a product it won't confidently recommend.

03

Ambiguity

Do the details actually say something?

“Premium quality” and “long-lasting” aren't facts an AI can repeat. This check flags vague, binary, or bare values that read like copy but carry no citable information.

04

Content clarity

Is the description written so a machine can quote it?

Description depth and structure. Thin or muddled content gives AI systems nothing to lift into an answer — clarity is what makes a page quotable.

05

Consistency

Does the product say the same thing everywhere?

Price, availability, and key facts are compared across your page, your structured data, and your catalog. Conflicting sources make AI systems hedge — or skip you.

06

Intent clarity

Does the page say who — and what — this product is for?

Use-case and fit signals. Buyers ask AI assistants “what's best for X” — a page that never states its X can't win that question.

07

Archetype fit

Does the product carry the fields a buyer of this product type expects?

A sunscreen is judged as a sunscreen, a camping tent as a camping tent. Each product is measured against the expectations of its own product type — see the archetype system below.

In your report

Reports translate these checks into five named pillars, each with a status and the top gap holding it back: Product Information Completeness, Brand & Price Signals, Content Clarity, Information Consistency, Category Fit & Purpose.

04

The archetype system

You can't score a sunscreen and a camping tent against the same checklist and call it fair. CITAQ maintains 155 product archetypes — each one a template of what a buyer of that product type actually expects to see — and classifies every product into the archetype that fits it.

When a product is confidently matched to a specific archetype, it's scored in canonical mode: completeness is judged against that archetype's own expectations, not a generic list. An SPF value matters for a sunscreen; a hydration capacity matters for a water bottle; neither is held against the other.

This is the mechanism that makes honest per-product scoring possible at all — and it's why a store-level tool structurally can't tell you which specific fields your specific product is missing.

Same check, different expectations

Chemical exfoliantactive %, skin fit, routine stage
Camping tentcapacity, season rating, packed weight
Mechanical keyboardswitch type, layout, connectivity

Illustrative examples of the kinds of expectations archetypes carry — the full per-archetype field maps are part of the engine.

05

Safeguards and score tiers

Some conditions are serious enough that a high score would be misleading no matter what else is right — a page AI crawlers can't access, a severe price mismatch, a description too thin to assess. In those cases a scoring safeguard limits the score, and the report says so explicitly: which safeguard fired, why, and what removes it. A limited score is always labeled — never silent.

The tier ladder — how every score reads out

  • AI-Strong

    AI systems can reliably find and cite your product.

  • AI-Eligible

    AI systems can find your product but citation reliability varies.

  • Needs Attention

    Gaps are limiting AI discovery of your product.

  • Not AI-Ready

    Significant issues are preventing AI systems from seeing your product.

  • AI-Blocked

    This product is effectively invisible to AI systems.

06

A worked example

A chemical-exfoliant serum from our public Beauty test corpus (anonymized). Same product, same page — before and after three specific fixes. Illustrative of how tiers move; not a formula you can reverse-engineer, by design.

BeforeNeeds Attention

58

  • No structured product data — price and availability invisible to AI systems.
  • Vague attribute values — “gentle formula,” no concentration, no volume.
  • No stated use case — nothing says which skin types or routines it's for.
AfterAI-Strong

78

  • Added structured product markup with price, availability, and identifiers.
  • Replaced vague copy with concrete values — active percentage, volume, pH range.
  • Stated skin-type fit and routine stage in plain language on the page.
07

What we publish — and what stays inside the engine

This page and your reports use the same translation, drawn from the same boundary. Everything you need to act on is published. The machinery that makes the score hard to game is not.

Published — here and in every report

  • The seven checks and what each one looks at
  • Both readiness dimensions and what they mean
  • The full tier ladder and every tier's meaning
  • Every safeguard that limits a score, named and explained
  • Every recommendation, with the reason and expected effect in plain language

Proprietary — inside the engine

exact check weightingswithheld
safeguard thresholdswithheld
per-archetype field mapswithheld
extraction & routing mechanicswithheld

Withheld deliberately: publishing the exact formula would make the score easy to game and worthless as a signal — for you and for everyone AI systems serve.

A–Z

Glossary

Every CITAQ term you'll meet across the site, your dashboard, and your reports — defined once, linkable from anywhere.

Citation Readiness Score (CRS)
The 0–100 score CITAQ assigns to a single product, measuring how ready that product is to be found and cited by AI systems. Every product in a catalog gets its own score — there is no store-level average standing in for individual products.
Retrieval Readiness (R)
One of the score's two dimensions: can AI retrieval systems find, access, and index this product at all? Crawler access, structured data, and identity clarity live here. A product that can't be retrieved can't be cited, no matter how good its content is.
Citation Readiness (C)
The score's second dimension: once an AI system has found the product, is the content concrete, consistent, and clear enough to be worth quoting in an answer? This is where vague copy, thin descriptions, and conflicting facts cost you.
Archetype
A product-type template — sunscreen, camping tent, chemical exfoliant, mechanical keyboard — that defines what information a buyer of that product type expects to see. CITAQ maintains 155 archetypes, and every product is scored against its own.
Canonical mode
The scoring mode in which a product has been confidently matched to a specific archetype, so its completeness is judged against that archetype's expectations rather than generic ones. This is what makes the score category-fair: a serum is never penalized for lacking tent poles.
Score tier
The plain-language band a score falls into — AI-Strong, AI-Eligible, Needs Attention, Not AI-Ready, or AI-Blocked. Tiers are how reports and dashboards communicate standing without asking you to memorize numbers.
Live Verification
Real queries run against live AI platforms — ChatGPT, Perplexity, Gemini, and Google AI Overviews — to observe what they actually say about your product right now, rather than inferring it from your page alone.
Identity Fact Defense
The Live Verification capability that checks whether AI platforms state your product's identity facts — brand, name, price, availability, key claims — correctly, and alerts you when an answer gets them wrong.
Contradiction
A verified conflict between what an AI platform says about your product and what your own catalog says is true. Contradictions are classified by severity, because an AI confidently stating the wrong price is a different problem than one omitting a detail.
Citation Leaderboard
A per-category ranking of which products AI platforms actually cite for buyer questions in that category — your position, who outranks you, and per-platform differences.
Indexability
Whether crawlers and AI systems can technically reach and index your pages in the first place: robots directives, sitemaps, and AI-crawler access. Indexability problems cap everything downstream — you can't be cited from a page that can't be read.
Marketplace / Rufus Readiness
How ready a product listing is for marketplace AI assistants such as Amazon's Rufus — whether the listing carries the signals those assistants use to surface products inside the marketplace itself.

See the methodology applied to your own products.

Every scan runs all seven checks and returns the score, the tier, and the plain-language reasons — per product.