How ready are independent stores for AI shoppers?
Only 3.9% of the 414 independent stores we tested are agent-ready. The median store scores 79 out of 100, a point short of a B, held back by the same few gaps.
The Agentic Commerce Grader tested 414 independent online stores in 20 niches, once each. Here are seven findings: what an AI shopping agent can and can't do on these stores, where it gets stuck, and what a store owner can do about each.
The seven findings
- 01Only 3.9% of the stores tested are agent-ready.Half the stores score 79 or more, a point short of a B.
- 02Agents can get in. Trust is where stores fall short.Delivery and returns terms are the weakest group of checks.
- 0392% hide shipping and returns from agents.The terms aren't in the structured data agents read.
- 04Agents can read the price, not compare you.GTINs, specific descriptions and ratings are missing.
- 05robots.txt lets agents in.No tested store blocks AI shopping agents in robots.txt; 20 turned the agent away with a bot check, an access-denied error or a redirect.
- 06The platform matters.Shopify stores average 78.3; no other platform reaches 60.
- 07Niches differ, the gaps don't.Activewear brands lead, Tech accessory brands trail.
Only 3.9% of the stores we tested are agent-ready.
16 of the 414 stores earn an A (Agent-ready). Another 156 earn a B (Mostly ready). The median score is 79 and the average 73: half the stores score 79 or more, a point short of a B, and the findings below show the few gaps that hold nearly all of them back.
- A · Agent-ready16
- B · Mostly ready156
- C · Partly ready140
- D · Hard for agents65
- F · Closed to agents37
What you can do Find out where your store stands. The free Agentic Commerce Grader tests it in about a minute and gives you its grade, a replay of the agent's visit and the fixes that matter most. Test my store →
Source: Scayl, Agentic Commerce Grader, 414 independent stores tested once each, 26–27 Oct 2026. Grades as the Grader gives them.
Agents can get in. Trust is where stores fall short.
The Grader's 21 checks fall into four groups. Of the checks that could be tested, the stores passed 84% of Seen and 84% of Buyable, but only 54% of Understood and 35% of Trusted.
What you can do Work on your weakest group first. On average that is Trusted, then Understood. The Grader shows your four group scores side by side. See my four scores →
Share of passed checks among those that could be tested on each store. Buyable here covers express wallets, cart links (inferred from the store's platform) and agent checkout (a published UCP manifest): checkout itself (3 checks) is tested only when a store's owner opts in, so it is left out.
92% keep shipping and returns out of the data agents read.
An agent comparing stores needs the delivery cost, the delivery time and the returns terms. Most stores have a readable returns page (85%). But only 16% state both delivery cost and time before checkout, 44% show them only at checkout, and 92% put neither shipping nor returns in their structured data, the markup agents read first.
- Shipping and returns are in structured data92%failed, of 414
- Shipping costs and times are stated44%failed, of 413
- The returns policy is easy to find15%failed, of 404
PassedPartlyFailedShare of the stores each check could be tested on
What you can do Add shipping and return policy markup. Repeat the terms in structured data so agents can read them without parsing prose.
Then Show shipping cost and delivery window on the product page. Agents compare delivery before they buy. Show cost and a delivery window next to the price. Check my store →
Shares of the stores each check could be tested on. Source: Scayl, Agentic Commerce Grader, 26–27 Oct 2026.
Agents can read your price. They can't compare you.
80% of stores put the price in their structured data, and 73% the stock. But only 19% give brand, SKU and barcode (GTIN): most (65%) have brand and SKU but no GTIN. Only 22% describe products specifically enough to choose by, and 17% put ratings in their markup. On 65% the agent found no ratings at all.
- Ratings can be read17%passed, of 413
- Products have identifiers (GTIN, SKU)19%passed, of 414
- Descriptions are specific enough to choose by22%passed, of 413
- Stock is in the structured data73%passed, of 414
- The price is in the structured data80%passed, of 414
- Products carry structured data82%passed, of 414
- Variants can be read and picked89%passed, of 366
PassedPartlyFailedShare of the stores each check could be tested on
What you can do Add brand, SKU and GTIN. Agents match products by code. Add the barcode (GTIN) and your SKU to each product.
Then Rewrite descriptions to answer buyer questions. State materials, dimensions, fit, compatibility and care. These are the questions agents are asked. Check my products →
The figure is the share that passed, of the stores each check could be tested on; checks sorted from least to most often passed.
No store in the sample blocks AI shopping agents in robots.txt.
What you can do Let verified shopping agents past the bot filter. Add an allow rule in your CDN or bot manager for verified AI agents. Keep challenges on login and checkout.
And Switch on agent checkout where your platform offers it. On Shopify this is the Agentic Storefronts channel, currently for stores selling to US buyers. Other platforms are catching up. Check my store →
robots.txt is checked for AI shopping agents (such as ChatGPT-User, Claude-User and Perplexity-User); a store without one counts as open. Cloudflare's bot protection was seen on 17 of the stores. A store that fails the bot check can't grade above D, however well the rest of it scores. 2 more stores couldn't be tested at all and are left out: they show no product pages to a bot, or close everyone out in robots.txt.
Shopify stores average 78.3. No other platform reaches 60.
336 of the 414 stores run on Shopify. WooCommerce stores average 57.3, headless Shopify builds 46 and other platforms 47.3. On WooCommerce, 87% of stores have no agent checkout protocol.
What you can do Off Shopify, or on a headless build, these fixes are yours to make. The most common gaps there are shipping and returns markup (finding 3) and ratings: Add aggregate rating markup. Repeat the rating and review count in the product markup. Test my store →
Platforms with fewer than 5 stores are counted under Other platforms.
Activewear brands lead at 83.1. Tech accessory brands trail at 60.5.
Each niche has 10 to 33 stores, so read small differences with care. The gap is the same almost everywhere: shipping and returns in structured data is the most-failed check in 19 of the 20 niches.
What you can do Find your niche, download its card below, and test your store to see where you sit against it. Test my store →
Average score out of 100 per niche. Select a niche for its card.
What this is, and what it isn't.
- 414 independent online stores in 20 niches, each tested once by the Agentic Commerce Grader on 26–27 Oct 2026. An AI shopper visits the store and reads its product, policy and robots pages, as a shopping agent would.
- How the stores were chosen: a hand-checked shortlist of independent online stores in 20 niches, mostly in the US, UK, Canada and Australia; marketplaces and big retail brands were left out. It is not a random sample: 81% of the stores run on Shopify, and the figures describe these 414 stores, not every online store.
- Read-only: the Grader respects robots.txt and reads public pages only. It adds to the cart and walks to payment only on stores whose owners opt in, and never places an order.
- Not tested here: checkout itself (guest checkout, the number of steps, a captcha). It runs only on stores whose owners opt in, so those 3 checks are unknown for every store in this sample and left out of the scores.
- Stores that turn the AI shopper away (a bot check, an access-denied error, a redirect to another site) are tested and graded: they fail the bot-wall check. 2 more stores couldn't be tested at all and are left out of every figure: they show no product pages to a bot, or close everyone out in robots.txt.
- Scoring: each of the 21 checks has a weight. A pass counts in full, a partial pass half, and a check that couldn't be tested is left out. Grades: A 90+, B 80–89, C 65–79, D 50–64, F 0–49. Scores are shown rounded, and grades come from the unrounded score, so a store shown as 80 can still be a C. A store that fails the bot-wall check is capped at D, whatever else it scores: an AI shopper that is turned away can't shop there.
- One scan per store is a snapshot: stores change, and a second visit can find something different.
- Aggregates only. No store is named; groups under 5 stores are folded into Other.
- This is a different sample from the Grader benchmarks page, which follows 182 stores through one purchase, step by step.
Where does your store stand?
Run the same test on your store. It's free, takes about a minute, and you get a grade, a replay and the fixes that matter most.



















