Ottom

Why it matters

Why it matters

When an AI assistant cannot complete a purchase on your store, nothing happens. It does not bounce, it does not abandon a cart, it does not email support and it does not appear as a lost sale. It goes to the next shop, and your analytics record a visit that looked fine.

The failure is silent, which is the whole problem

Every other kind of website problem announces itself eventually. A broken payment page produces complaints. A slow site shows up in your numbers. A customer who cannot find a size emails you about it, or at least leaves a trail you can see afterwards.

An agent does none of that. It reads what it can, tries what it can, and stops. There is no version of your reporting where this appears, because from the outside it is indistinguishable from somebody who looked and was not interested.

An agent working down the 4 stages of a purchase and stopping at BuyDiscover: passed. Evaluate: passed. Compare: passed. Buy: stopped here.DiscoverEvaluateCompareBuyit stops hereAn agent completing three stages of a purchase and stopping at the fourthA stall is the point where the agent stops making progress. It is not an error message on your store. It is the agent giving up and going somewhere else.DiscoverEvaluateCompareBuyit stops here
Illustrative. The useful part of a scan is not the score, it is which of these four it stopped at, and what it was looking at when it did.

This is not a bet on one company

You do not have to believe any particular assistant will win, or that this will be most of retail. The relevant fact is narrower and it is a matter of record rather than opinion: the large assistant makers and the large commerce platforms have each published ways for software to read a catalogue and complete a purchase. Our agentic commerce page sets out what each of them shipped, with dates and sources. Nothing on it is our estimate.

The fixes are ordinary, and they are not wasted

What stops an agent is rarely exotic. A price that exists only inside an image. A size picker that needs a mouse. A returns policy published as a document rather than a page. A checkout that insists on an account before it will show a total.

These are the same changes that help somebody using a screen reader, somebody on a patchy connection, and a search engine trying to understand what you sell. That is the part that makes this decision easy: if the whole premise turns out to be smaller than expected, you have still improved your store for people. The stall library lists every one we detect and what to change.

What we will not tell you

We are not going to quote you a percentage of revenue at risk. We do not know your revenue or how much of it arrives through an agent, and a number we cannot show the working for would undermine the one thing this product sells, which is a result you can check.

What we can show the working for is how far agents get. Across 317 stores Ottom measured at every stage in the 90 days to 19 October 2026, in every category and including stores whose category could not be determined, the median agentic readiness score was 97 out of 100. 87% of them let an AI shopping agent reach the payment step.

A store enters Ottom’s survey only if a catalogue of at least ten products can be read from it in one request, which in practice means Shopify and WooCommerce storefronts. These figures therefore describe the storefronts Ottom can read in bulk rather than online retail as a whole. The stores are a convenience sample, not a random or representative sample of online retail: a store is counted because someone scanned it or because its catalogue could be read in bulk, so these figures describe what Ottom has seen and nothing wider.

The all stores benchmark has the full figures, aggregated and anonymised, with the method published beside them.

What it costs to find out

One scan of one store is free and needs nothing installed. If the answer is that your store is fine, that is a useful thing to know and you have spent nothing. The paid plans exist for the part that is genuinely ongoing: a store that passed last month is not a store that passes today, because themes change, apps get added and the assistants themselves change.

Frequently asked questions

Is anyone actually buying through AI assistants yet?

Not in the numbers that would make this urgent, and we are not going to pretend otherwise. What is documented is that the large assistant makers and the large commerce platforms have all shipped ways for an agent to read a catalogue and complete a purchase. Our agentic commerce page sets out what each of them has published, with dates and sources, and nothing on it is our own estimate.

Why not just wait until it is obviously worth it?

Because the work does not compress. Most of what stops an agent is ordinary: a price only shown in an image, a size picker that needs a mouse, a checkout that demands an account. Those take an afternoon each when you have a list, and they are the same changes that help a screen reader and a search engine. Waiting does not make the list shorter, it just means you find out later.

Can I not just test this myself?

You can try, and it is worth doing once. The difficulty is that you know your own store: you know where the size picker is and what the returns policy says, so you supply the missing information without noticing. The value of a test is that it does not know any of that.

What would make this argument wrong?

If assistants converge on buying only from a handful of large marketplaces, an independent store’s own checkout matters much less and this is a smaller problem than we think. We would rather say that out loud than have it discovered later. The fixes still stand on their own: everything a scan asks you to change also helps a person on a phone and a search engine reading your page.

Scan your store, or read how it works first.