AI agents hit website blocks: what operators need to change

2026-11-02 · NoobClaw Newsroom · Social Media Tools & Comparisons

Meta's Muse, Perplexity and other AI agents are hitting CAPTCHAs, bans and security controls. Website owners need to separate crawler blocks from legitimate buyer agents—and adjust checkout rules accordingly.

What happened

AI agents that shop or book on a user's behalf are running into access problems even when a site has not deliberately banned them. Meta's Muse hit human-verification checks at Walmart, one of its retail partners; Walmart says the checks were not intended to stop the assistant. Amazon has been more direct, deliberately blocking the tool from shopping on its site, according to TechCrunch.

Some restrictions are commercial or security-driven. Amazon is battling Perplexity over its Comet AI browser, arguing the agent violates its access rules and puts customer data at risk, while Perplexity says users have authorized it. A federal appeals court vacated a preliminary injunction against Perplexity in August 2026, but the underlying dispute remains unresolved. Amazon also has its own Buy for Me feature for purchases from other brands' websites.

Security is a bigger concern when money is involved. During JPMorgan Chase's October earnings call, CFO Jeremy Barnum said some agents use shared login credentials and screen scraping to transfer funds, methods he described as "not secure." The bank blocks those methods where possible, and Barnum warned customers who hand over credentials could be responsible if an agent takes an unwanted action.

Key facts

Our analysis

The friction is not just about a single blocked bot. It splits automated traffic into three groups: search crawlers, AI training bots, and agents acting for users. Cloudflare now lets website owners distinguish those groups, and its default since September 15 blocks AI training bots and agents on ad pages for new domains. That suggests a site can accidentally stop a legitimate buying assistant while still welcoming search engines, especially if the owner does not adjust the rule.

This matters because the appeal of agents like Meta's Muse is to navigate websites, fill shopping carts and book travel instead of handing users recommendations. If a CAPTCHA or anti-scraping defense interrupts that flow, the user returns to manual work. Zillow has acknowledged its anti-scraping defenses sometimes catch agents it otherwise permits, so the user-facing symptom may look like an assistant that simply cannot finish the job.

For platform operators, the control point is shifting from blanket blocking to permissioned agent access. Shopify's checkout tools show a middle path: compatible browser agents can update checkout details and submit orders after buyer confirmation, with control handed back to shoppers for required authentication steps. That suggests sites can allow useful agents without giving up security or final buyer control.

What it means for operators

Source:Why some websites are blocking AI agents from accessing the web — Engadget(2026-11-02)

Editor's note: prepared by the NoobClaw newsroom with AI assistance from the public report above. Facts are as reported by the source; the analysis is our view. Spotted an error? Contact us and we will correct it.