Agentic readiness: Particula

An audit of particula-tech.com against the five things AI shopping agents (ChatGPT, Gemini, Perplexity, Shopify's own agent) rely on to understand a product and recommend it.

Checked November 5, 2026. Read-only: nothing in the store was changed. Every fix needs its own approval.

The plumbing is ready; the product data is not. The store already exposes an llms.txt file and a full agent commerce endpoint (UCP/MCP), so an agent can search, build a cart and reach checkout. But it reads the same product data, and most of the fields it leans on are empty.
0 / 310images with alt text
58 / 60products with no product type
45 / 60descriptions empty or under 80 words
9internal or duplicate products visible to agents

The five criteria

CriterionStatusWhat we found
Titles and descriptionsWeak40 products have under 80 words and 5 are empty. Even the flagships are thin (about 55 words on Wizard Air and Wizard Pro), and some descriptions belong to a different product or still use the old names (Mini, Lite).
ImagesMissingNone of the 310 images has alt text. The theme shows the product title in its place, but the field agents receive is empty. 26 products have fewer than 3 images and 5 have none. The GoChess "What's in the box" images still say Mini and Lite.
Structured attributesMissingProduct type is empty on 58 of 60 products. The tags that exist are mostly internal (chessupsell, swag, B2B) and do not describe the product. 8 products have no tags at all.
Barcodes (GTIN)PartialSome products have a barcode, but Wizard Alive (the $699 flagship) does not. More important: the theme does not output gtin in the structured data (JSON-LD) on any product page, so even existing barcodes never reach Google or agents.
PoliciesGoodRefund, shipping, terms and privacy pages all exist. Missing: return and shipping policy are not in each product's structured data, and the contact page is 9 words.

Recommendations, in order

  1. Hide the 7 internal products, and decide on 2 more

    An agent cannot tell an internal listing from a real one, and may recommend "Rubik's Connected Pro (internal)", a $199 retail display stand, or a duplicate GoBalance Sport at a different price. This is the fastest fix: remove them from the Online Store channel, nothing gets deleted.

    In the file: Action column = hide (7 products), review (2 products)

  2. Alt text for every image

    291 alt texts are ready. The 129 images of the 11 flagship products were written by looking at each photo. The rest are drafts built from the file name and are marked for review before they go in.

    In the file: "Image alt text" sheet

  3. Product type, category and descriptive tags

    Every product has a proposed product type, Shopify standard category and tags to add (brand, kind, audience, use). Keep the existing tags: some of them drive the upsell logic.

    In the file: "Products" sheet, proposed type / category / tags to add

  4. gtin in the structured data, and a barcode for Wizard Alive

    One change to the product JSON-LD in the theme: add gtin from the variant barcode. Worth adding hasMerchantReturnPolicy and shippingDetails in the same place. In parallel, fill in missing barcodes on physical products.

    This is a live-theme change: one file, its own approval

  5. Fix the wrong descriptions, then expand the flagships

    First fix copy that describes another product: Wizard Air opens with "GoChess Lite... tournament-size", and Wizard Pro opens with the membership perks. Then give every flagship its dimensions, age range, what's in the box, app and platform compatibility, and battery life.

    In the file: the Content notes column says what is wrong per product

  6. Replace the old names in titles and images

    Titles still carry "(MINI Model)" and "(LITE Model)", and the GoChess "What's in the box" images say Mini and Lite. An agent that reads Mini will not connect the product to Air.

Download the fix list (Excel)

All products

ProductPriorityProposed type / reasonBarcodeIssuesNotes

Not covered

Three things that need a further check

Metafields and the official product category are not exposed in the public data. Seeing them needs read access to the admin.

We tried to see what an agent actually receives through Shopify's agent endpoint, but it requires a registered agent profile and returned nothing.

Barcodes were read from the product pages. Products not yet scanned say "not checked": the store blocks fast crawling.