Autonomous agents stand out as the leading trend in artificial intelligence. OpenAI, Meta, Google, Instinct, and others assert that their agents can browse the internet for users, managing tasks such as reserving dinner tables, sorting email, or buying shoes. Yet difficulties emerge when websites refuse cooperation.

Users of Meta’s Muse, xAI’s Grok, OpenAI’s Dots, and Instinct have recently faced restrictions that stop their AI agents from completing actions on popular sites. Yelp stated to TechCrunch that it accepts only AI agents in its data licensing program, while Delta Air Lines bars them from booking tickets. JPMorgan Chase and Charles Schwab confirmed they limit AI agent account access over security worries. Some users report that sites like Cars.com and Amtrak, once workable, now fail, hinting at quiet blocks by more firms.

Amazon has opposed AI agents since last year by banning Perplexity’s agentic shopping tool and suing the startup, which raises questions about agents’ online future. Amazon noted it had repeatedly asked Perplexity to halt agent shopping there, adding that third parties buying for customers should act openly and respect provider choices on participation.

Perplexity replied that Amazon resists agents because they cut human ad views, a genuine issue for the retailer. The startup likened its agent to a user extension, arguing Amazon rejects the right to hire help. A judge first stopped the agents, but an appeals court reversed the decision in August. Amazon then filed an updated complaint.

Even without outright bans, agents may still clash with anti-bot systems

The legal fight has not halted OpenAI, Meta, and Google from releasing their agents, but they encounter the same issues as Perplexity. Muse met Amazon opposition for policy violations. Reports show Grok, Instinct’s agent, and Dots also fail on Amazon. Elon Musk suggested launching an Amazon rival. More access problems increase doubts about agents’ usefulness.

“Does this affect the perceived value of Anthropic and OpenAI? Absolutely,” says Chris Canal, CEO of EquiStamp. Claims that agents will handle every online task weaken if companies block them, reducing what agents can achieve.

Meta addressed blocks by noting that rejecting agents rejects customers, preferring collaboration to avoid lost business. Meta partnered with Walmart, Gap, Sephora, Best Buy, and others to let Muse shop there. Shopify has also welcomed agents to its storefronts.

The Verge contacted OpenAI and Google for comments on these roadblocks but received no immediate replies. Instinct declined to comment.

Even without bans, anti-bot systems create problems. Reddit reports indicate Google blocks some Muse accounts for suspicious activity, and checkout issues persist on Walmart despite the Meta partnership. Canal notes these systems target bots generally, and agents trigger the same safeguards.

Cloudflare helps sites block unwanted AI visitors. After a one-click AI crawler block in 2025, it began stopping crawlers that scrape for both search indexing and AI training. This effort aims to protect web infrastructure from AI bot strain that burdens servers and email systems, and could complicate buying tickets or products.

While blocks may limit AI agents, Canal warns they will make the web more frustrating as sites add more checks to confirm users are human. “We could see them on a simple grocery shopping website soon, where you’re having to jump through a bunch of hoops to prove that you’re a human using the computer or your phone.”

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

Emma Roth