Mixed signal. Agents can reach the page, but guess from there.

Glossier scores 61 out of 100, a mixed signal that masks an unusual split personality. The discoverability and transactability foundations are essentially modern: every tracked AI crawler is welcomed in robots.txt, all nine sampled product pages carry both Product and Offer schema, Shop Pay sits alongside Apple Pay and PayPal, and the Shopify cart endpoint at /cart/add is reachable. Then the site turns around and fires reCAPTCHA and hCaptcha at agents on the homepage and cart, ships zero structured product attributes, and offers no price history or validity windows. The infrastructure is agent-ready. The posture is not. The most expensive failure, at six points lost, is product attribute structure (check 2.2). Across nine product pages, exactly one returned a single structured attribute (brand, on the field greens blanket), and not one page hit the two-attribute threshold. This is striking given that Glossier already emits rich Product JSON-LD: the schema scaffolding is there, but the fields agents need to filter and compare (color, size, material, GTIN, ingredients) are not populated. A hoodie sold in seven sizes and a fragrance sold in three volumes both return empty structuredHits arrays. For a beauty brand whose differentiation lives in shades, scents, and skin types, this is the single biggest leak in the report. The second wound is self-inflicted: CAPTCHA prevalence (check 4.2, zero of four). The scanner detected both reCAPTCHA and hCaptcha on the homepage and cart, with transaction_blocking flagged true. Pair this with the 403 on https://glossier.com/checkout in check 4.1 and you have a site that says "come on in" via robots.txt and then slams a challenge gate at the exact moment an agent tries to transact. The block rate is technically only 3 percent, but it is concentrated precisely where revenue happens. Checkout flow structure (check 3.1) loses four more points for the same reason. The classifier could only observe two steps (Cart Page, Checkout Page) before checkout returned unavailable, and guest checkout (check 3.2) inherited the blind spot: the classifier could not confirm guest availability because the checkout page would not render without an initialized cart. Shopify's platform default is guest-friendly, but agents do not get to take that on faith. Three smaller but cumulative failures round out the picture. Sitemap quality (check 1.3) loses two points because https://glossier.com/sitemap.xml carries 757 URLs with no lastmod values and no product URLs surfaced, making freshness invisible. Price history (check 5.3) scores zero despite seven of nine sampled products showing sale markup: compare_at_price values are present in the raw schema (2100 down to 1400 on Pro Tip, 4800 down to 4200 on the Balm Dotcom Trio), but no schema validity dates accompany them, so an agent cannot verify the discount is real. And stated agent policy (check 4.4) is a flat zero because the scanner could not retrieve terms or privacy content to classify, leaving Glossier silent on whether agent-mediated purchases are permitted. The biggest opportunity: populate structured product attributes (color, size, ingredients, GTIN) inside the existing Product schema and remove CAPTCHA from the cart path, and Glossier converts a mixed-signal grade into a clear green light for agentic commerce.