How Clione delivers your product data
Generating a signal, publishing it to your storefront and an assistant receiving it are different things. Check the live verification for your own store; a saved artifact or a valid schema does not prove that your theme serves it.
One set of facts, different placements
The prepared entity bundle uses the same commercial facts for structured data and an accessible specification table. Matching FAQs belong in the served page beside their FAQPage markup. Image-alt proposals preserve supplied descriptions; they do not invent materials, camera angles or certifications. Review before publishing. Inferred scores and keywords are not displayed by default.
Supported server transports use Shopify theme app extensions, BigCommerce Stencil widgets, WooCommerce/PrestaShop hooks, or the headless SDK's server helpers. Availability still depends on platform installation, approved publication and the current theme. New visible-table integrations require rollout and live checks; this page is not a claim that every existing installation has been upgraded. Browser-only embed scripts are labelled Google fallbacks, not direct-AI delivery.
What the controlled study measured
Arrival means detection in saved answers or exposed tool content. An absent canary in an answer-only interface does not independently prove bytes were stripped. The vendors expose different observation boundaries.
The 29 October 2026 m-40 study used unique canaries in page placements and measured their detection in saved answers or exposed tool content through APIs and Claude Code. It kept dated models, autonomous and forced prompts, denominators and confidence intervals separate.
| Placement | Observed boundary |
|---|---|
| Served body and title | Canaries arrived across the measured direct surfaces, at different rates |
| Served JSON-LD | Arrived in the three valid Gemini API arms; not in the other measured direct surfaces |
| Browser/GTM JSON-LD | No canary arrival in any measured arm |
| Image alt | Arrived in several measured surfaces, not Perplexity Agent API |
| Meta description | Arrived in Claude API, not the other measured surfaces |
| llms.txt | No request in complete measurement intervals |
Hidden-div controls also delivered canaries. Visible-first is our shopper-facing delivery policy, not a claim that every reader ignores all hidden text. The new specification table has not been separately measured for recommendation effects. Perplexity's observed generic Chrome user agent is not proof of JavaScript execution or a reliable vendor identifier.
The study does not measure consumer chat apps, search rankings, Google/Bing indexing, the indirect search-to-AI path, recommendations or sales lift. No kill decision follows from it. Search-engine features are a separate integration purpose; direct-AI results do not establish search eligibility.
Scores and verification
The published score remains formula version 1: a structural completeness measure. An inactive per-reader formula version 2 is pending approval; missing raw evidence is unknown, not zero. Do not compare the proposal with an old benchmark.
After publishing, use Verify to check your storefront. A delivery check confirms observed HTML, not that an assistant will recommend a product. For account-specific rollout questions, contact support through the app.
Optional legacy files
Existing llms.txt, Markdown and agents.json endpoints remain optional and experimental. They receive no new investment or promotion, are not required for onboarding, and contribute no readiness points. The MD domain is available only in Store Settings. Keep JSON-LD for SEO; review and publish useful visible product facts.