Ottom

Synthetic monitoring for agentic commerce

Can AI agents actually buy from your store?

Our agents try to discover, evaluate, compare and buy from your store, the way an AI shopper would. Ottom reports exactly where they stall, in language you can forward to your developer.

By scanning you agree to the Terms of Service

Free, and no card. We email you a link, and the scan starts the moment you click it.

One free scan per store, per email address, every 30 days. Our agents identify themselves. See how.

Agent journey

example-store.com58/100
DiscoverPASS
EvaluatePASS
CompareSLOW
BuySTALLED
Stopped atBuy

Illustrative journey. Not a real store.

How it works

One journey. Four stages. Zero guessing.

Static scanners check your code. We check the outcome. An agent runs the whole trip, and every stage either passes or stops with a reason you can act on.

The same product page as a customer sees it and as software receives itA customer sees a photo, a price, sizes and a buy button. Software receives the title and a description, and nothing where the price, the sizes and the policy should be.WHAT A CUSTOMER SEESTrail Duffel 40L£129.00In stockAdd to basketWHAT SOFTWARE RECEIVESTrail Duffel 40LA roomy weekend bag.no priceno sizesno returns policyThe same product page as a customer sees it and as software receives itA customer sees a photo, a price, sizes and a buy button. Software receives the title and a description, and nothing where the price, the sizes and the policy should be.WHAT A CUSTOMER SEESTrail Duffel 40L£129.00In stockAdd to basketWHAT SOFTWARE RECEIVESTrail Duffel 40LA roomy weekend bag.no priceno sizesno returns policy
Illustrative, and ordinary. The price was in the photo, the sizes needed a hover, and the returns policy was a PDF. Nothing here is broken, and nothing here can be bought.
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 AI agent walking the four stages of a purchase and stalling at the buy stepThe agent passes discover, evaluate and compare, then stops at buy, which is marked as a stall rather than a pass.DiscoverEvaluateCompareBuySTALLED
Illustrative. A real scan reports which stage stopped, on your store, with the page the agent was looking at when it gave up.

  1. Stage 1

    Discover

    Can agents find your products from your data? Or are they blocked before they start?

  2. Stage 2

    Evaluate

    Can they read your prices, stock, sizes and returns policy without guessing?

  3. Stage 3

    Compare

    Do they have enough to confidently pick you over the next tab?

  4. Stage 4

    Buy

    Can they get through your checkout? Or does a popup, a variant or a bot wall end it?

See this on your own store, free, once.

Where it stops, and what it saw

Every stage gets a verdict and every stall gets its evidence. In the buy attempt that is the page the agent was looking at when it gave up. In the data stages it is what your pages and feeds returned when it read them.

Discover92
Evaluate78
Compare54
Buy0

Illustrative journey. Not a real store.

How Ottom turns a stalled agent into a finding you can act onThe agent stops at a real step, the page it was looking at is kept as evidence, the cause is matched to a named stall, and the report gives the change to make.The agent stopsat a step on your storeWe keep the evidencethe page it was looking atWe name the causeone of 36 known stallsYou get the fixand where to make itHow Ottom turns a stalled agent into a finding you can act onThe agent stops at a real step, the page it was looking at is kept as evidence, the cause is matched to a named stall, and the report gives the change to make.The agent stopsat a step on your storeWe keep the evidencethe page it was looking atWe name the causeone of 36 known stallsYou get the fixand where to make it
Every finding carries the evidence it was made from. That is the difference between a score and a result you can check, and it is why a Critical needs two agreeing signals before we will print the word.

What is actually measured

The numbers that exist, and who published them

Three figures, none of them ours. Each links to where it is sourced in full, and each carries what it does not mean, because two of the three are a platform reporting on itself.

We are not going to tell you what share of your revenue is at risk. We measure how far an agent gets on a storefront, not what the attempt was worth, so any revenue figure from us would be a guess wearing a measurement’s clothes. What we can tell you is what happens on your store, and what that would be worth at numbers you supply.

Work it out with your own numbers, or scan your store and use the failure rate we measure on it.

Then we keep watching

Stores break quietly. We check every day.

A store that passed last month is not a store that passes today. Themes change, apps update, and the agents themselves change. That is why the score is worth watching rather than reading once.

Up to twice a day

Full agent journeys

An agent attempts a complete purchase, to the payment step, on the products most likely to stop one: every month on Coverage, every week on Monitor, every day on Defend, twice a day on Patrol.

Every day

Data checks

Price, stock, structured data and policy readability, on every paid plan.

On deploy

Re-checks when you publish

Publish a theme on Shopify and we re-check the store, up to once a day. Defend, on the Shopify rail.

Alerts

In plain language

"Agents cannot pick a size on your duffel", with the stage, the severity and the fix, linking to the evidence in your dashboard.

The stakes

If an agent can’t buy from you, it buys from someone else

Traffic from AI assistants is small today and growing. The awkward part is that it fails quietly: an agent that stalls at your size picker does not bounce in your analytics, email support, or show up as a lost sale. It just goes to the next tab.

You do not have to believe this will be most of retail to want to know where you stand. You only have to believe it will not be zero.

Scan your store

New to this? Start with agentic commerce, work out what a stall costs you, or see what a scan costs.

Frequently asked questions

Is this a scanner?

Not in the usual sense. A scanner reads your code and checks it against a list. We send an agent through the whole purchase and report where it stopped. That is the difference between "your markup is missing a field" and "nobody could choose a size".

Will you place real orders?

No. The agent goes as far as the payment step and stops. No card details are entered and no order is placed.

Will this look like an attack to my bot protection?

It should not. Our agents identify themselves with a published user agent, arrive from a fixed set of static addresses, and visit at low volume. If your protection challenges them anyway, the report tells you that rather than pretending the store is broken.

Do I have to install anything?

No. You give us your shop address and we work from the outside, exactly as an assistant would. There is no app to add, no script to paste and no change to your theme.

What does it cost to find out?

The first scan is free and does not ask for a card. A single report is $9, which includes re-scanning that store while you fix things, for 7 days. Watching a store continuously starts at $19 a month.

Which platforms does this work with?

Any storefront we can reach. We recognise 10 platforms by name, including Shopify, WooCommerce, Magento and Salesforce Commerce Cloud, and the report tells you which one we found. If yours is not one of them the scan still runs.

How often does it run?

The free scan runs once, when you ask for it. We cap how often any one storefront is scanned in a day, counted across everyone who asks rather than per person. The default is 12 retrieval checks and 3 full agent journeys, and only a contract with that merchant raises it. On a plan the cheap checks run every day and the full purchase attempt runs every month on Coverage, every week on Monitor, every day on Defend, twice a day on Patrol.

Who can see my results?

You, and anyone you send the link to. Reports are never public and never indexed. The benchmark pages publish aggregates only, and only once a category holds enough stores for no single one to be identifiable.