40% of major retailers turn an AI shopping agent away at the front door.
BlynkAudit sends a real browser agent through the shopping funnels of 676 major retailers across 14 categories in Australia, the UK and the US, every week. In the latest complete measurement, 273 of them (40.4%) refused our automated shopper before it could read a single product: enterprise bot defence, working exactly as configured.
| MARKET | BLOCKED AT THE DOOR | RATE |
|---|---|---|
| United States | 108 / 204 | 52.9% |
| United Kingdom | 101 / 227 | 44.5% |
| Australia | 64 / 245 | 26.1% |
| All three markets | 273 / 676 | 40.4% |
The angles
- US retail is twice as walled as Australian retail (52.9% vs 26.1% blocked).
- The wall isn't where you think it is. Some brands' regional sites block agents while their own global storefront lets them straight in, and for the brands whose door does open, the funnel often dies later, at CAPTCHA and login walls in checkout.
- The wall doesn't show up in anyone's dashboard. Blocked brands keep ranking on industry boards via previously-measured data; nothing tells them the agents stopped getting in.
- The rate is stable, three consecutive weekly measurements read 41.1%, 42.3% and 40.4%, the last on a population nearly twice the size. This is a property of the industry, not noise.
"A blocked scraper rotates IPs and retries. A blocked shopping agent doesn't argue with the door, it buys from the competitor whose door opens."
What "blocked" means, exactly
A retailer counts as blocked only when both of our crawler passes, a plain HTTP client and a full headless Chromium browser, run from our servers in Google Cloud (Sydney), fail on the homepage: an error status, an empty document, or a bot-challenge page (Cloudflare "Just a moment", Akamai "Access Denied", CAPTCHA interstitials). We say "turned away our AI shopping agent": ours is one instrument, applied identically to every retailer, every week. Residential-connection measurement favours the retailer: data-centre agent traffic fares worse, not better.
Being blocked is a configuration choice with a cost, not incompetence, the same defence that stops scrapers stops buyers. Blocked sites are marked † PARTIAL on our boards and their scores are withheld, never faked.
About BlynkAudit
BlynkAudit scores retail websites on two horizons: NOW, the human sale you win today, and NEXT, whether an AI shopping agent can find, read, and buy from you. Four pillars (Technical, Agentic, Market, Sentiment), measured by live crawls, public data, and a browser agent that actually attempts the funnel. Founded by Sean, 20+ years in retail, most recently leading consumer insights and enterprise AI adoption at IKEA Australia.
For journalists
The full blocked list, per-category breakdowns, week-over-week movement, and methodology are available on request, every number traces to a logged scan. Contact: [email protected]. Boards: the NOW/NEXT Index.