Documentation.

Citely works on your Shopify store's machine layer: the structured data that AI engines and crawlers read, without touching your visible theme. The principle is non-negotiable: amplify what's real, never invent. Citely exposes facts that already exist in your catalog; it fabricates none.

What Citely changes — and doesn't

• Changes: the schema.org/Product markup, structured attributes, and the feed meant for engines.
• Doesn't touch: your theme, your design, your prices or your visible content.

The connection uses standard Shopify OAuth. You authorize the app on your store; Citely gets read access to your catalog and write access only to the metafields it manages.

Steps

1. Start the install from the authorization link (/api/shopify/install).
2. Approve the requested permissions in the Shopify screen.
3. At the end of the flow (/api/shopify/callback), your store is connected and your catalog read.
4. Run a first scan to get your citability score.

The scan evaluates each product on its citability: how complete, structured and consistent its page is for an AI. The result is a per-product citability score, and the list of gaps to fix.

What the audit checks

• Structure: presence of clean schema.org/Product.
• Completeness: your industry's decisive attributes, identifiers (GTIN).
• Consistency: agreement between page, feed and markup.

The audit grades neither design nor classic SEO: it measures only what an AI can extract and cite.

From the detected gaps, Citely generates an enriched data layer — from your real data — then writes it to the metafields. The theme extension then injects the matching schema.org/Product on your pages, without modifying the visible theme.

The cycle

1. Preview: Citely proposes the improvements, you stay in control.
2. Apply: writing to the dedicated namespace's metafields.
3. Injection: the theme extension exposes the JSON-LD on the storefront.

The target format is schema.org/Product, exposed as JSON-LD. It's the grammar engines expect to extract facts unambiguously.

{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Sérum Vitamine C — Peau sensible",
  "brand": { "@type": "Brand", "name": "Veridian" },
  "gtin13": "3401590012345",
  "additionalProperty": [
    { "@type": "PropertyValue", "name": "Type de peau", "value": "Sensible" },
    { "@type": "PropertyValue", "name": "Sans parfum", "value": "Oui" }
  ],
  "offers": {
    "@type": "Offer",
    "price": "34.90",
    "priceCurrency": "EUR",
    "availability": "https://schema.org/InStock"
  }
}

The fields that matter

• Identity: name, brand, gtin — to be matchable against comparisons.
• Decisive attributes: additionalProperty — your industry's buying criteria.
• Offer: price, priceCurrency, availability — consistent with the feed.

Engines cross-check several sources: your page, your feed, your markup. A contradiction (mismatched price, wrong stock) sends a distrust signal and lowers your chances of being cited. Citely detects these gaps so you can fix them.

• Price gaps between feed and page.
• Inconsistent stock or availability.
• Missing or divergent GTIN.

Citely exposes your catalog to AI agents via a per-store NLWeb / MCP endpoint. An agent can query this access point to get structured answers about your products, kept up to date from your catalog.

GET   /api/nlweb/{shop}
POST  /api/nlweb/{shop}/ask

On the crawler side, you stay in control of access via robots.txt. The llms.txt file can complement it, but remains an unguaranteed proposal (see the blog).

Citely measures your citations in the answers of 5 engines: Claude, ChatGPT, Gemini, Perplexity and Google AI. Across a set of queries, it computes your share of answers, engine by engine.

Honest signaling

Below a data threshold (too few queries, runs or engines), Citely shows "insufficient" rather than a misleading percentage. A bad number is worse than no number.

Citely can generate factual, multilingual content from your real data, and structured FAQ (FAQPage) from real buyer questions — always without inventing anything.

• Factual descriptions, not hollow marketing.
• Marked-up FAQ, directly citable by engines.
• Multilingual, aligned with your markets.

Today, the connection and operations go through the authenticated app (Shopify OAuth). The per-store NLWeb / MCP agent endpoint, by contrast, is designed to be public for agents.

Data freshness

Shopify product webhooks keep the machine layer up to date when your catalog changes. Enabling them in production requires stable-URL hosting.