ChatGPT Shopping reaches 250M+ weekly users through conversational queries like "recommend a Dutch oven for an induction stove." It doesn't read XML feeds — it reads Schema.org JSON-LD on your product pages, and it cites the sources behind each claim. Central writes the verified facts, source list, and confidence scores ChatGPT needs to pick you over a guess.
The same buyer asks the same question in ChatGPT. Without verified product data on your page, ChatGPT either refuses or invents. With Central's grounded JSON-LD, it answers — and it cites the source behind every number.
I can see a product page for the Le Creuset Signature Round, but the page doesn't expose structured product data — no @type Product, no Offer, no source citations.
I can guess that cast-iron Dutch ovens usually handle around 250 °C? and that most are induction-compatible? — but I can't verify either for this product. I'd rather not recommend it than risk an unsupported answer.
The Le Creuset Signature Round Dutch Oven (26 cm, 5.3 L, Cerise) is a strong match. From the structured product data on the page:
It is oven-safe to 260 °C1, induction-compatible across all heat sources2, made of enameled cast iron3, and rated 4.8 / 5 across 1,287 reviews4. Currently € 379.00, in stock5.
Same buyer. Same product. Same query. ChatGPT moves from refusal to a confident, cited recommendation — because Central writes the keys AI shopping agents need to verify a claim before they speak.
The Schema.org Product, Offer, AggregateRating, and PropertyValue fields ChatGPT and other agents parse from your page — plus Central's x-central provenance extension so the model can show its sources, not its guesses.
Each property carries the source that backed it and a readiness state. The model has everything it needs to verify a claim before it speaks.
ChatGPT, Perplexity, and other AI shopping agents don't take a feed — they crawl what you publish. Central makes sure what you publish is something they can verify, cite, and confidently recommend.
Paste a shop URL, upload a CSV, or connect a catalog. Central maps your fields to Schema.org properties and shows which ones are filled and which are missing.
Each claim is checked across 10–20 sources — manufacturer specs, retailers, reviews, certifications. Two agreeing sources reach 0.82; five reach 0.97. Anything below the floor is held.
Central emits a full Schema.org <script type="application/ld+json"> block, including x-central provenance with sources, confidence, and timestamp.
Inject into your PDP via the Central widget or your platform. ChatGPT, Perplexity, Apple Intelligence, and Claude pick up the same JSON-LD — one record, every AI surface.
The same Le Creuset SKU. A thin barebones Schema.org block on the left — the kind that gets ignored. The Central-emitted JSON-LD on the right: full Product, Offer, AggregateRating, FAQPage, and x-central provenance with 15 sources, confidence avg, ISO timestamp, and one marketing claim held.
<script type="application/ld+json"> { "@type": "Product", "name": "Signature Round Dutch Oven", "sku": "SIG-RND-26-CER", "price": "379.00" } </script>
<script type="application/ld+json"> { "@context": "https://schema.org", "@type": "Product", "name": "Le Creuset Signature Round Dutch Oven, 26 cm, 5.3 L, Cerise", "sku": "SIG-RND-26-CER", "gtin13": "0024147210504", "brand": { "@type": "Brand", "name": "Le Creuset" }, "image": "https://shop.le-creuset.eu/products/signature-round-26.jpg", "description": "Enameled cast iron, oven-safe to 260 °C, induction-compatible.", "material": "Enameled cast iron", "color": "Cerise", "weight": { "@type": "QuantitativeValue", "value": 5.4, "unitCode": "KGM" }, "offers": { "@type": "Offer", "price": "379.00", "priceCurrency": "EUR", "availability": "https://schema.org/InStock", "priceValidUntil": "2026-12-31", "url": "https://shop.le-creuset.eu/p/sig-rnd-26-cer" }, "aggregateRating": { "@type": "AggregateRating", "ratingValue": 4.8, "reviewCount": 1287 }, "additionalProperty": [ { "@type": "PropertyValue", "name": "maxOvenTemp", "value": "260 °C" }, { "@type": "PropertyValue", "name": "capacity", "value": "5.3 L" }, { "@type": "PropertyValue", "name": "heatSources", "value": "induction, gas, electric, ceramic, halogen" }, { "@type": "PropertyValue", "name": "warranty", "value": "lifetime" } ], // Central extension — provenance ChatGPT can show its work with "x-central": { "sources": [ "le-creuset.com/specifications", "williams-sonoma.com/products/le-creuset", "saveur.com/best-dutch-ovens", "cooks-illustrated.com/equipment/le-creuset", "trustpilot.com/review/le-creuset.com", /* 10 more sources — total: 15 */ ], "confidence": { "avg": 0.94, "floor": 0.82 }, "lastVerified": "2026-05-25T08:14:32Z", "held": [ { "claim": "World's best Dutch oven", "reason": "no source" } ] }, "@graph": [ { "@type": "FAQPage", "mainEntity": [ { "@type": "Question", "name": "Is it induction-compatible?", "acceptedAnswer": { "@type": "Answer", "text": "Yes — all heat sources." } }, { "@type": "Question", "name": "What is the max oven temperature?", "acceptedAnswer": { "@type": "Answer", "text": "260 °C / 500 °F." } }, /* + 6 more verified Q&As */ ] } ] } </script>
ChatGPT Shopping, Perplexity, Apple Intelligence, Google AI Overviews, Claude — all read the same Schema.org payload Central writes to your PDP. One record adapts to every conversational shopping surface.
Same JSON-LD payload, same provenance. Perplexity surfaces source-cited recommendations from the same verified record — including the x-central source list.
See Perplexity setupApple Intelligence consumes structured product data the same way. The Schema.org Product + Offer + AggregateRating Central emits is what Siri and Spotlight will read.
Redirects hereAnthropic's Model Context Protocol reads the same Schema.org payload. When MCP commerce arrives, Claude pulls the verified record without a separate integration.
Redirects hereAI Overviews and AI Mode read Merchant Center + Schema.org. One Central record powers the Google feed and the JSON-LD AI Overviews cite on the PDP.
See Google setupLooking for Amazon Rufus, Apple Intelligence, or Claude MCP commerce? /channels/amazon-rufus, /channels/apple-intelligence, and /channels/claude-mcp all redirect here — the same JSON-LD covers every AI shopping agent.
Central prepares the data ChatGPT Shopping needs to recommend you. ChatGPT Shopping isn't a feed-submission product like Google Merchant Center — OpenAI's commerce surface reads Schema.org JSON-LD that lives on your product pages, plus structured catalogs where its commerce integration is enabled. Central writes the JSON-LD, embeds the x-central provenance, and refreshes it as facts change. Platform inclusion, ranking, and account-level commerce settings stay with OpenAI.
Both, depending on the surface. Conversational shopping queries are answered primarily from structured data ChatGPT can read on your product page — Schema.org Product, Offer, AggregateRating, FAQPage in a <script type="application/ld+json"> block. Where OpenAI's commerce integration is enabled, structured catalogs can also be submitted directly. Central produces the same canonical record either way, so you don't have to choose.
x-central provenance metadata, and why does it matter?x-central is Central's Schema.org extension that attaches the sources, confidence avg, ISO 8601 last-verified timestamp, and held-claims list to a product. AI agents can cite the sources behind a claim instead of speaking with no grounding. When ChatGPT recommends your product, it can show its work — and a model that can show its work is a model that's safer to recommend you.
Yes — and increasingly so. Conversational shopping agents are biased toward sources they can verify because every recommendation carries reputational risk for the model. A page with no Schema.org loses to a page with full Schema.org; a page with Schema.org but no sources loses to a page with Schema.org plus x-central sources. Central writes the version that wins the citation.
They all read structured product data the same way. Perplexity is live and cites x-central.sources directly. Apple Intelligence and Anthropic's MCP commerce are emerging and consume the same Schema.org payload — when they arrive, no separate integration is needed. Central's record is the abstraction; every surface is a render of it.
Yes — that's the entire point. One canonical product record produces the XML feed Google MC consumes, the JSON-LD ChatGPT and Perplexity crawl, the structured catalogs OpenAI's commerce integration ingests, and the JSON your in-store widget renders. Confidence, sources, and held claims travel with every output — so what's blocked in one place is blocked everywhere.