What an audit looks like, page by page.
The structure and delivery come from a real engagement, anonymized. The findings shown are the failures we see most across the 85 stores we’ve tested.
big brands had a hard blocker
Pyranthus benchmark, June 2026
Even big brands fail this. 5 of the 23 we tested had a hard blocker.
Companies with whole e-commerce teams still ship pages an AI assistant can’t use. In our mid-market cohort it’s worse: only 7 of 62 stores were fully readable. Because stores run on templates, most of it traces back to a short list of fixes.
The audit finds your blockers and ranks them.
The fix list, exactly as we deliver it.
The audit ends in work orders, not observations. This is the kind of list we deliver. Have a developer? They work from it. Don’t have one? We do the fixes ourselves.
- 01
Show your price and stock where an agent can read them, not just a shopper's browser.
- 02
Spell out sizes, options, and what's in stock in plain, readable data.
- 03
Make add-to-cart and checkout work without a popup an agent can't click.
- 04
Turn on Shopify's AI shopping channel so agents can see your catalog.
- 05
Run an agent through the store again to confirm it can check out.
Four gates an AI shopper walks through.
The audit tests all four. Across the stores we’ve tested, the gate that fails most is understanding: product data an agent can actually read.
Where an AI shopper meets your store
4 waysThe closed gate sets the order of work.
The work stays concrete.
Every flagged page has a finding, the consequence for the AI shopper, and the smallest useful fix.
The format is what clients receive. The finding shown is the one we see most.
Get this view of your store.
We’ll run the audit on your store and walk you through what we’d fix, in twenty minutes.