Does ARC try to solve CAPTCHAs?

No. ARC never attempts a CAPTCHA and records it as the stop.

Are human checks common in ARC's data?

They are one of the smaller stops. Add-to-cart is the most common. The live counts are on this page, with the snapshot date.

Can I let good agents skip the check?

Many security vendors support verified bots or signed requests. ARC does not measure how many stores have turned that on.

ARC · Answers

Why do human checks stop AI shopping agents?

Human checks are designed to stop bots, so they often stop AI shopping agents too, even at stores that want the sale. In the October 2026 Index report, a human-check was the stop at 3 of 52 stores where a shopper failed. This page also shows the current snapshot's stop counts.

Short answer

A human-check is built to stop bots. It often stops a shopping agent as well, including at a store that wanted the sale. ARC does not try to solve the check. It records the stop.

Why the check cannot tell a shopper's agent from a scraper

The check looks for automation. A shopper's agent and a scraper can look alike from the outside. Press reports, including TechCrunch on 6 October 2026, said Walmart described its Muse failures as unintentional. That is the retailer's account, not an ARC measurement of Walmart.

What ARC's stop reasons show

Published report: State of agent buyability, October 2026. Sample: 52 scored Index stores where at least one shopper failed to add to cart.

Human-check was the published stop: 3 of 52MeasuredOctober 2026 Index report. A different sample from the current snapshot below.Snapshot 2026-10-12.r2, 12 October 2026.

ARC open dataset snapshot 2026-10-12.r2 (2026-10-12). Measured scores only. The Measured stops, most common first, are Couldn't add it to the cart (32), then Blocked at the size picker (8), then Stopped by a human-check (3), then Store blocks automated agents at the door (2).

Stopped by a human-check: 3 storesMeasuredSnapshot 2026-10-12. One published stop per store. Inferred scores are not in this count.Snapshot 2026-10-12.r2, 12 October 2026.

Add-to-cart is the most common stop in the current Measured Index. A human-check is one of the smaller counts. The store page shows which step stopped the browser shoppers.

Verified agents, signed requests, protocol checkout

Some security vendors let a signed or verified agent through while still challenging everyone else. ARC does not measure how many stores have turned that on. A UCP checkout link skips the storefront page. See Web Bot Auth.

Check your store

Run /agent-checkout. If the stop was a human-check, the store page says so. A missing check is not a pass.

Frequently asked questions

Why do human checks stop AI shopping agents?: Human checks are designed to stop bots, so they often stop AI shopping agents too, even at stores that want the sale. In the October 2026 Index report, a human-check was the stop at 3 of 52 stores where a shopper failed. This page also shows the current snapshot's stop counts.
Does ARC try to solve CAPTCHAs?: No. ARC never attempts a CAPTCHA and records it as the stop.
Are human checks common in ARC's data?: They are one of the smaller stops. Add-to-cart is the most common. The live counts are on this page, with the snapshot date.
Can I let good agents skip the check?: Many security vendors support verified bots or signed requests. ARC does not measure how many stores have turned that on.

On ARC

Cite this: ARC, arcreport.ai, accessed 2026-10-13, index snapshot 2026-10-12.r2, CC-BY-4.0. Protocol figures: Live saved checks, accessed 13 October 2026. Not a frozen snapshot. Index browser figures: Snapshot 2026-10-12.r2, 12 October 2026. Platform figures: Live saved checks, accessed 13 October 2026. Not a frozen snapshot.

Every figure is labelled Measured, Inferred or Not measured. Not measured is not zero. A missing result is not a mark against a store. ARC stops before payment.