Being in AI shopping search takes two things: your product data has to be complete enough for an assistant to read, and current enough that what it reads matches what you're actually selling. AgentReady scores both on your catalog page, from your own data, and groups everything by how confident we can honestly be about it. This page explains the three parts.
Completeness: the eligibility you control
The completeness score is one number, 0 to 100, for how ready your product data is to be read by AI search. It is a weighted measure across the fields that decide eligibility:
- Taxonomy (category), the heaviest pillar, because the Shopify Standard Taxonomy category is the anchor the Global Catalog normalizes on. See the taxonomy ranking factor.
- GTIN / barcode, because identifiers are what cluster your product onto the shared cross-store listing so you compete on it instead of being invisible. See GTINs and barcodes.
- Attributes (tags), the faceting signals an assistant uses to match your product to a query.
- Brand (vendor), because agents weight brand when they match.
Each pillar shows its own coverage, and the weakest one is flagged as your biggest lever, so you always know the single change that moves the score most.
It is measured only over the products where a field was actually observable. A pillar we could not read is dropped and its weight redistributed, and the card says so, so the score never implies we checked something we didn't.
Freshness: is the data agents read up to date?
A complete catalog still goes stale. AI assistants read a copy of your catalog, not your live store, so when that copy lags, they quote the wrong price or recommend something that's sold out. The freshness score measures how current the AI-readable layer AgentReady maintains for you actually is, per surface (products, and content like collections and pages). It combines two signals:
- Lag, the time since the last successful sync.
- Drift, recent failed syncs, which mean changes that never made it into the layer, so it is actively diverging rather than just aging.
A fresh, fully-synced catalog scores high; one that hasn't synced in days, or is failing, scores lower, with the oldest surface called out. There's a deeper write-up in Stale catalog, wrong answer.
The three provenance tiers
The catalog page groups everything into three tiers, each labeled with what its numbers actually prove, so a score we verified is never read as a guarantee we can't make:
- Eligibility we verify is measured from your own data: completeness and freshness live here. These are the levers you control, and we can prove them.
- Presence we observe is point-in-time readings of Shopify's Global Catalog, like where you rank for a tracked query. Real, but what we saw when we last checked, not a promise of what an assistant returns this second.
- External truth we link is everything we can't verify from inside your store, where we point you to the source rather than claim it ourselves.
Keeping these separate is deliberate. Most tools blur them together and quietly overclaim; the tiers mean you always know whether a number is something we proved, something we observed, or something to go check at the source.
What it does not do
Both scores are read entirely from your own catalog. There is no cross-merchant benchmarking and no peer ranking in catalog quality: they measure your eligibility and your data's recency against what AI search needs, not against anyone else. And nothing here changes your catalog. Fixing a flagged product still runs through AgentReady's confirm-before-write flow, where you review and apply each change yourself.
Want a read on where you stand before you dig in? The free AI Readiness Checker scores your storefront the way an agent would, in about a minute.