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

Nike scores 54 out of 100, a mixed signal that captures the strange shape of this property: a site engineered for human discovery and SEO, retrofitted unevenly for the agent era. The fundamentals an agent first reaches for, robots posture and product schema, are immaculate. The fundamentals an agent needs to actually transact, guest checkout and price transparency, are absent or hostile. Nike has built a storefront that AI crawlers can read beautifully and that AI agents cannot buy from. The largest single hole is structured product attributes (check 2.2, zero of six). Out of nine sampled product pages, only three exposed brand, color, size, gtin, or mpn in machine-parseable form. The Air Max 90 women's page (2ZsM2w), the Free Metcon 7, the Raiders tee, and the Calm Mules all returned attrCount of zero. Nike already publishes Product plus Offer JSON-LD on every sample (check 1.2, full credit), so the infrastructure exists. The attribute payload is just empty on most SKUs. This is the cheapest fix on the report and the highest leverage. The transactability dimension is where Nike actively repels agents. Guest checkout (3.2) scored zero of four with high classifier confidence, and price transparency (3.3) scored zero of four because the product page literally says "You'll see our shipping options at checkout" with no tax disclosure preauthentication. Combined with checkout flow structure (3.1, two of six) flagged as requiring an initialized cart, an agent attempting an end to end purchase faces a login wall, opaque totals, and a multi step flow with no machine native payment rails detected (3.4, one of three). The cart endpoint pattern at /cart is detectable (3.5, full), but that is a runway leading to a closed door. Trust signals are a near total miss for a brand that should dominate this dimension. Returns policy (5.1) scored zero because the classifier hit only navigation chrome at nike.com/returns with no window, conditions, or process exposed. Third party trust anchors (5.4) returned zero hits across Trustpilot, BBB, Yotpo, Bazaarvoice, and eight other platforms. Merchant verifiability (5.2) confirms "Nike, Inc." but no physical address or registration ID. An agent vetting whether to trust a transaction here is left with brand recognition alone, which is fine for Nike and useless as a generalizable signal. The discoverability layer earns its grade. Robots.txt allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot, anthropic-ai, Applebot-Extended, Bytespider, and Amazonbot, with zero blocked (check 1.1, full). But the sitemap audited (1.3) was sitemap-us-help.xml with a most recent lastmod of February 2019, stale by any standard, and there is no llms.txt or ai-plugin.json (1.4). Agent posture is excellent: zero blocks across 29 fetches, no CAPTCHAs anywhere, no rate limiting at five requests per second. Content quality is the quiet bright spot. Specification completeness (6.1) hit high or medium on eight of nine pages, conversational descriptions (6.2) on eight, and the absence of manipulation tactics (6.6) earned full marks with legitimate event signals like the Moon Shoe 5.7 launch flagged correctly rather than as fake urgency. Price stability (6.5) scored zero because no priceValidUntil or schema validity windows are published, a structural omission given Nike's frequent promotional cadence. Nike's biggest opportunity is to populate the Offer schema with attributes and validity windows it already knows internally, turning a readable catalog into a transactable one.