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

Everlane lands at 53 out of 100, a mixed signal that captures a brand fluent in the mechanics of selling but illiterate in the language agents now use to choose. The plumbing works. Shop Pay, PayPal, and Apple Pay sit on every product page and the cart, JSON-LD Product plus Offer markup hits nine of nine sampled URLs, and robots.txt waves through GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and six other named agents without a single Disallow. The problem is that when an agent actually arrives, the pages give it almost nothing to reason with, and the cart it tries to interact with throws a CAPTCHA. The single largest hole is agent-native content, where Everlane scores zero across specification completeness (6.1), conversational descriptions (6.2), decision-support metadata (6.3), delivery date transparency (6.4), and price stability (6.5). Every one of the nine sampled product pages, from the Way-High Sailor Jean to the Transit Chino, classifier-tagged as low signal with no quotable specification text. For a vertically integrated apparel brand whose marketing rests on radical transparency about materials and factories, shipping product pages that an agent reads as empty is a strategic miss. The fabric story exists somewhere; it is not on the PDP in a form a model can extract. The second wound is in agent posture. CAPTCHA prevalence (4.2) scored zero because both reCAPTCHA and hCaptcha fire on the cart and homepage, and the check flagged transaction_blocking as true. Combine that with the 403 returned on https://everlane.com/checkout (4.1) and the silent stance on agent policy in the terms and privacy pages (4.4), and the picture is a merchant that has technically allowed crawlers in robots.txt while operationally blocking the agentic transaction at the moment of value. Guest checkout (3.2) is also unavailable, with the cart routing through a SIGN UP / LOG IN prompt, costing two more points. Product attributes (2.2) is the third drag, scoring zero of six. Across nine PDPs, only two surfaced any labeled attribute (color, on the Slim Oxford navy and white), and none reached the two-structured-attributes threshold. The Shopify-shaped JSON in the raw HTML carries SKU and price beautifully but stops short of size, material, fit, and care in any agent-readable form. Variant handling (2.5) tells the same story: six of nine products use swatch pickers, zero enumerate variants in markup. Sitemap quality (1.3) deserves a specific callout. https://everlane.com/sitemap.xml returns 200 with only 8 locs, no product URLs, and no lastmod timestamps anywhere. For a catalog this size, that is not a sitemap, it is a placeholder. Agent-specific endpoints (1.4) compound the issue: no llms.txt, no ai-plugin.json, nothing at the well-known paths. What works is worth naming. Price in raw HTML (2.1), payment rails (3.4), cart API accessibility (3.5), rate limiting (4.3), price history (5.3), and third-party trust via Yotpo (5.4) all scored full marks. The transactional spine is real. The biggest opportunity is to push structured product attributes, materials, fit, and delivery dates onto the PDP and remove CAPTCHA from the cart, converting a site that agents can find into one they can actually buy from.