Ottom

Benchmarks / All stores

Agentic commerce benchmark across all stores

New: the monthly benchmark report. How these figures move, by email once a month. Read the issues or sign up below.

An AI shopping agent reached the payment step on 81% of the 468 stores Ottom scanned in the 90 days to 13 November 2026, in every category and including stores whose category could not be determined, counting stores that blocked the agent or stopped it early as not reached. Of the 436 measured at every stage, 87% reached it, and the median agentic readiness score was 97 out of 100. Cite

This benchmark pools every store Ottom has scanned, across all categories, including stores whose category could not be determined. Each store is counted once, whichever category it is in. It is counted under exactly the rules each category is, set out in the methodology.

n = 436 measured of 468 scanned · 15 August 2026 to 13 November 2026 · v4.0.0 · bench-v7 · methodology

Scores come from the 436 stores Ottom measured at every stage. A further 32 were scanned but stopped before checkout, usually at bot protection or a robots.txt rule, so they have no comparable score. All 468 are still counted in the stalls below, because what was found on a store does not depend on how it was scored.

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 methodology sets out what that excludes.

Between 1 and 10 October 2026 the median score across all stores rose from 55 to 97. The stores did not change. Our agent did. On 4 October 2026 we replaced the scripted agent that walked checkout, agent-v1, with a new one, agent-v2, and from 8 October 2026 we measured every surveyed store again with it. On 250 stores measured by both, the old agent reached the payment step on 5 and the new one on 212. Figures before 10 October 2026 were measured wholly or partly by the old agent and are not compared with figures after it. The methodology sets out which agent measured these figures.

Reaching the payment step

An AI shopping agent reached the payment step on 81% of the 468 stores Ottom scanned in the 90 days to 13 November 2026, counting stores that blocked it or stopped it early as not reached, and on 87% of the 436 measured at every stage. Ottom stops there and never submits payment. Cite

81%Of all 468 stores scanned
87%Of the 436 measured at every stage
Each square is 1% of the 436 stores measured. Green: the agent reached the payment step.

Where agents stall

Of the 436 stores Ottom measured at every stage, 85% scored 80 or better at the discover stage, against 89% at the compare stage. Cite

Each stage of the journey, and the share of stores whose score at that stage was 80 or better.

Green: stores that scored 80 or better at the stage. The rest of each bar is stores that did not.
All stores agentic readiness by journey stage, 15 August 2026 to 13 November 2026
StageMedian scoreStores passingStores measured
Discover10085%436
Evaluate8888%436
Compare10089%436
Buy10087%436

What stops them

The most common stall across all stores is products have no GTIN or MPN to match against, found on 63% of the 468 stores Ottom scanned in the 90 days to 13 November 2026. Cite

The most common stalls across all stores, as a share of all 468 stores scanned. A stall found on fewer than 5 stores is withheld, because at that count it points at a merchant rather than at a pattern.

Each stall sits under the stage where it stops an agent. The bar is the share of stores it was found on.
Most common stalls for stores, 15 August 2026 to 13 November 2026
StallStageSeverityShare of stores
Products have no GTIN or MPN to match againstEvaluateEvaluateMedium63%
Variants are not described in structured dataEvaluateEvaluateHigh11%
Shipping cost cannot be determined before checkoutCompareCompareHigh9%
Product pages have no schema.org markupDiscoverDiscoverHigh8%
Add to cart needs JavaScript the agent cannot runBuyBuyMedium7%
No sitemap for agents to crawlDiscoverDiscoverHigh6%
No llms.txt to orient agentsDiscoverDiscoverLow4%
Agent could not reach the payment stepBuyBuyCritical3%
Feed price and storefront price disagreeCompareCompareHigh2%
No machine-readable product feedDiscoverDiscoverHigh2%

What to change first

12 further stalls were found but withheld: too few stores showed them to report a share without pointing at an individual store.

Readiness score

The median of the 436 stores Ottom measured at every stage in the 90 days to 13 November 2026 scored 97 out of 100, and a quarter of them scored 93 or below. Cite

The spread of overall readiness scores across the category. We publish quartiles rather than a best and worst store, because the best and the worst are each one identifiable merchant.

93Lower quartile
97Median
97Upper quartile
Shaded: the middle half of stores. Dark line: the median store.

Most stores score near the top: three quarters score 93 or more, and the middle store scores 97. Reaching the payment step is worth 40 of the 100 points, and a store where the agent did not reach it can earn at most 18 of those 40. So the score separates the stores an agent can buy from from the stores it cannot, and does not separate stores within each group by much.

How it has moved

A trend is drawn once this benchmark has a week of nightly figures under methodology bench-v7. Until then, the figures above are the only ones, and there is no line to read.

How to cite this benchmark

These figures are published under CC BY 4.0. Quote them, chart them and reuse them, commercial use included. Credit Ottom and link to this page, as below.

Suggested citation

Ottom, Agentic commerce benchmark across all stores v4.0.0, 15 August 2026 to 13 November 2026, methodology bench-v7, agent-v2. https://ottom.io/benchmarks/all. Licensed CC BY 4.0.

As HTML, for a web page

Source: Ottom, <a href="https://ottom.io/benchmarks/all">Agentic commerce benchmark across all stores v4.0.0</a>, 15 August 2026 to 13 November 2026, methodology bench-v7, agent-v2. Licensed under <a href="https://creativecommons.org/licenses/by/4.0/" rel="license">CC BY 4.0</a>.

Cite one figure

Each figure has its own link, so a quote can point at the sentence it came from.

  • Summaryhttps://ottom.io/benchmarks/all#summary
  • Reaching paymenthttps://ottom.io/benchmarks/all#checkout
  • Stageshttps://ottom.io/benchmarks/all#stages
  • Most common stallhttps://ottom.io/benchmarks/all#stalls
  • Median scorehttps://ottom.io/benchmarks/all#score

The citation names version v4.0.0, so the exact figures quoted can be checked later.

The same figures as JSON and as Markdown.

Previous versions

A new version is issued when a nightly refresh changes a figure on this page. Earlier versions stay here for 90 days after they were issued.

Can AI agents read these stores?

Before an AI agent can compare or buy anything, it has to be let in and find something it can read. These figures come from the first step of every scan, across the 465 online stores Ottom scanned in the 90 days to 13 November 2026, in every category.

  • Under 1%

    Block AI agents in robots.txtEvery store scanned: 465

    Fewer than 1% of the 465 online stores Ottom scanned in the 90 days to 13 November 2026 block AI agents in robots.txt.

  • Under 1%

    Stop agents with bot protectionStores robots.txt lets in: 465

    Of the 465 stores Ottom scanned in the 90 days to 13 November 2026 whose robots.txt lets AI agents in, fewer than 1% stop an AI agent with bot protection before the storefront loads.

  • 2%

    No machine-readable product feedStores an agent could open: 463

    Of the 463 stores Ottom scanned in the 90 days to 13 November 2026 whose storefront an AI agent could open, 2% publish no machine-readable product feed.

  • 4%

    No llms.txtStores an agent could open: 463

    Of the 463 stores Ottom scanned in the 90 days to 13 November 2026 whose storefront an AI agent could open, 4% publish no llms.txt.

  • 8%

    No schema.org Product markupStores whose product pages an agent read: 456

    Of the 456 stores Ottom scanned in the 90 days to 13 November 2026 whose product pages an AI agent could read, 8% carry no schema.org Product markup on those pages.

  • 9%

    No shipping costs before checkoutStores whose product pages an agent read: 456

    Of the 456 stores Ottom scanned in the 90 days to 13 November 2026 whose product pages an AI agent could read, 9% give no shipping costs an agent can read before checkout.

Which stores publish a UCP manifest?

A Universal Commerce Protocol manifest is a file at /.well-known/ucp that tells an AI agent platform which UCP services a store offers and where to reach them. Ottom reads it once per scan and counts it here as an observation, not a score: a store without one is not marked down. The figures are split by platform because a hosted platform that supports UCP usually publishes the file for every store on it, so one pooled share would mostly measure the platform mix. See what UCP is.

  • 99%

    ShopifyStores checked: 162

    Of the 162 Shopify stores Ottom scanned in the 90 days to 13 November 2026 whose UCP manifest it checked, 99% publish one at /.well-known/ucp.

Each share is counted over the stores that reached the point where it is checked, which the sentence names. When fewer than 5 stores show something, the exact share is not given: the panel gives a bound it is under, so no store can be singled out, or withholds the figure when the stores counted are too few for that bound to be small. A store enters Ottom’s survey only if its catalogue can be read, so these are mostly stores an agent can reach, and blocking may well be more common across online retail as a whole. The methodology sets out each denominator. The same figures as JSON.

Read next

See where your store sits