We analyzed 3,813 online stores — fewer than 1% are readable to AI

Before our July study of 148 Finnish stores, we ran a much larger sweep: in late March 2026 we scanned 3,813 international e-commerce sites — sourced from top-traffic domain lists and search, most of them US-based — with an earlier version of our AI-readability scanner.

The question was simple: if an AI assistant tried to read these stores' products, what would it find?

The headline numbers

• Average AI Readiness score: 61/100
• Only 30 stores out of 3,813 earned an A — that's under 0.8%
• Grade distribution: 30 × A, 955 × B, 2,042 × C, 749 × D, 37 × F
• Only 10% of stores exposed schema.org Product markup on the product page we analyzed
• On less than a third of the sites could the scanner find a product page with usable product data at all

In other words: e-commerce at large is not ready for the AI shopping channel. Not in Finland, not internationally.

Platform breakdown — no platform saves you

Two things stand out. First, Shopify stores were 2.6× more likely to expose Product schema than WooCommerce stores — modern Shopify themes emit JSON-LD by default. Second, and more interesting: it barely moved the average score. Shopify's 61.9 vs WooCommerce's 63.0 is a rounding error. A schema wrapper doesn't help when the data inside it is thin: missing identifiers, missing availability, descriptions written for humans-in-a-hurry rather than machines-that-compare.

That's the core lesson of this dataset: the platform gives you plumbing, not readability. What AI actually ranks on — complete prices, identifiers, availability, factual specs — is store-specific work no theme does for you.

Method and honest caveats

Scans ran March 28 – April 1, 2026 with an earlier version of our scanner (site-level checks + one product page per store; same grade bands as today). The current scanner adds stricter JSON-LD completeness checks, so these scores are not directly comparable with our July 2026 Finnish study — if anything, today's version would grade this sample lower. The sample skews toward high-traffic US domains; one product page per store is a sample, not a full-catalog audit.

What to do with this

If under 1% of stores pass, being readable to AI is not table stakes yet — it's a head start. Run the free scan on your own store to see where you stand, or read what the fix involves.