Learn
How to audit your store for agent readiness in an afternoon
A first agent readiness audit fits in an afternoon if you work in the order an agent meets your store and resist fixing things as you find them. Check permission, then enumeration, then legibility, then checkout, recording what you find rather than repairing it. Fixing as you go is what turns an afternoon into a fortnight and leaves the audit unfinished.
A first agent readiness audit fits in an afternoon if you work in the order an agent meets your store and resist the urge to fix anything while you are looking. Record findings. Fix afterwards. Fixing as you go is the single reason these audits do not get finished.
Stage one: permission
Read your robots.txt and write down every rule that mentions an agent or an assistant token. You are looking for disallows written for training crawlers that also catch shopping agents.
Then open your CDN bot settings and record what the categories are set to, including anything set by a vendor default. Then load your store from a network you have never used it from, and note whether you are challenged.
If you are blocked here, stop and record it. Nothing below this line matters until it is resolved.
Stage two: enumeration
Open your sitemap. Confirm it lists product URLs rather than stopping at collections, and spot check three products you know you added recently. If you publish a feed, fetch it, confirm it resolves, and check a price you changed in the last week is current.
Stage three: legibility
Disable JavaScript and open three product pages: your best seller, one with several variants, and one that is currently out of stock. For each, record whether the price, the availability, the variant options, the shipping cost and the returns terms are present as text.
The out of stock one matters more than it looks. Availability that is correct only when the answer is yes is a data defect that will be found by an agent and not by you.
Stage four: checkout
Open your store in a completely cold session, signed out, on a narrow mobile viewport. Add a product that requires choosing a variant. Reach the payment step. Stop there: there is no reason to submit a payment to test a checkout.
Record every point at which a human decision was required. A dismissed popup, an accepted consent banner, a chosen region, an account you created. Each of those is a place where an agent stops, and you will have done all of them reflexively.
Write it up so it survives
A finding should name the stage, the page or setting, and what you observed. That shape is forwardable to whoever fixes it and comparable against the next run, which is what makes the second audit take twenty minutes rather than another afternoon.
For the fuller reference version, use the agentic readiness checklist, and for when to do this again, how often should you test.
Questions
Why not fix problems as I find them?
Because the first real defect will absorb the rest of the afternoon and you will never reach the checkout, which is where the expensive failures are. Record everything, fix afterwards in the order an agent meets them.
What do I need before I start?
A browser where you can disable JavaScript, access to your robots.txt and sitemap, your CDN bot settings, and a way to open your store in a completely cold session. No tooling beyond that for a first pass.
What should the output look like?
A short list of findings, each naming the stage it belongs to and the page or setting it lives on. That format is what makes it forwardable to a developer and comparable against the next run.