Vercel and Ora launched a free public scanner on August 20, 2026 — is-agentic.com — that scores a URL 0–100 “based on what agents can discover, access, and use”. If you’ve seen it score a site and want to know whether it checks that an agent can buy something there, the short answer: the pages we read don’t describe a scored purchase test, and the tool itself says its report is not “a guarantee that every model or browser automation system will complete every task”.

Two scores, one engine

Ora’s own methodology page describes the full Ora ranker as four layers a real agent moves through: Discovery (15 checks, 20 points), Access (42 checks, 30 points), Usability (56 checks, 40 points) and Payments (6 checks, 10 points) — “can it pay you?” — out of a possible 100, as read on 2026-10-09 (the same page listed 16 Discovery and 41 Access checks on 2026-08-31). Ora’s August 20 launch post gave 127 checks across those four layers; the four stage counts on the methodology page add up to 119.

is-agentic.com is not that ranker as such. Ora’s launch post calls it “an extension of the Ora ranker we built together, tuned for what most websites need”, on which “a small set of essentials carries the score, and capability checks like MCP servers or payments sit in a recommended tier and in bonus signals that only add credit.” is-agentic.com’s own methodology page gives the split: “Essential checks share an 80-point pool and recommended checks share a 20-point pool. Emerging signals can add a small bonus, capped at five points, without becoming requirements.” Its homepage says Recommended checks activate when scan evidence indicates an API, OAuth flow, GraphQL endpoint, MCP server, developer portal or commerce surface, and that checks which don’t apply “are excluded rather than counted as failures” — a marketing page isn’t marked down for lacking a checkout. The methodology page also describes site-type views, commerce among them, that change “which checks are required, bonus, gated, or emerging” without changing the score.

Why that matters for a store

Two things follow from those numbers. First, on the full Ora ranker (Ora’s launch post speaks of “the 16,000+ domains we scan”), payments is worth 10 of the 100 points, the smallest of the four layers; Usability alone weighs four times as much. Second, on is-agentic.com, the pages we read place commerce among the Recommended checks, in the 20-point pool, not among the Essential fundamentals that share 80 points, and those checks only switch on if the scan detects a commerce surface in the first place. The pages don’t list what the commerce checks test.

Neither number is wrong for what the tool sets out to do — is-agentic.com is explicit that a site isn’t penalized for an interface it doesn’t provide, and Ora’s post says most websites “do not need half of that catalog”. But it means a store can post a high is-agentic.com or Ora score while its actual purchase flow — cart state, login walls, payment timeouts — may never have been exercised by the scored checks. Each is-agentic.com report does include one observed agent journey, kept out of the score; the pages we read don’t say which task it attempts or whether it reaches a checkout. is-agentic.com’s own page says it directly: “Use the report as a prioritized technical review, not as a guarantee that every model or browser automation system will complete every task.” Cloudflare’s Agent Readiness Score has a similar gap: per Search Engine Journal’s walkthrough, its commerce checks include a UCP profile at /.well-known/ucp and an ACP discovery document at /.well-known/acp.json, and nothing we read describes it attempting a purchase against them.

What to do with a score like this

If your domain scores well on is-agentic.com or ora.ai, treat it as a read on how well agents can discover, access and use your public pages — worth having. It isn’t evidence that an agent can add your product to a cart, authenticate, and pay. An AgentReady scan doesn’t run that full test either: it covers the store’s side of it up to the checkout boundary (whether shopping agents get past your door, whether your offer data and variants are machine-readable, whether the add-to-cart action is in the product page) and never places an order. Our Is Agentic comparison page sets what each of the two checks side by side.

FAQ

Does Vercel’s Is Agentic score check if AI agents can complete a purchase?

The pages we read don’t describe a scored purchase test. They place commerce among the Recommended checks (a 20-point pool, against 80 for the Essential fundamentals), which switch on when the scan finds a commerce surface, without listing what those checks test; Ora’s launch post adds that payments also sit in bonus signals that only add credit. Each report also shows one observed agent journey, kept out of the score, and the pages don’t say which task it attempts.

How much does payments count on the full Ora ranker?

Ora’s methodology page splits 100 points across four layers: Discovery (20), Access (30), Usability (40) and Payments (10). Payments is the smallest layer, worth a quarter of Usability, with 6 checks, and the page frames it as the last step “for products agents can buy or meter”.

What’s the difference between is-agentic.com and ora.ai’s own scanner?

Both read the same Ora scan: is-agentic.com’s methodology says “every check is run and scored by Ora”, and Ora’s launch post calls the two “one scan, two readings”. is-agentic.com groups the checks into an 80-point Essential pool, a 20-point Recommended pool and a bonus capped at five points, and leaves out checks that don’t apply. The ranker at ora.ai scores each of its four layers, Payments included, on how many checks passed and combines them into one number from 0 to 100; Ora’s post points products “agents transact with” to it.

Sources

A score that reads a checkout file is not the same as a scan that completes a checkout, and AgentReady doesn’t complete one either: our scan checks the store’s side of the path up to the checkout boundary.