Websites push back on AI agents as booking and shopping hit blocks

AI companies are racing to make agents that can browse, buy, book and manage accounts for users, but parts of the web are pushing back. According to a new report from The Verge, users of several agent products are running into blocks on popular services, raising a basic question for everyday automation: what can an assistant do if websites refuse to let it in?

For creators, freelancers and small businesses, this matters because many promised agent workflows are mundane but valuable: comparing purchases, reserving travel, managing inboxes or handling repetitive account tasks. The current friction suggests those jobs may still need a human fallback.

The web is setting boundaries around agents

The Verge reports that users of Meta’s Muse, xAI’s Grok, OpenAI’s Dots and Instinct have seen AI agents blocked while trying to complete tasks on major websites. The examples span several categories: Yelp told TechCrunch it only permits AI agents that pay for its data licensing program, while Delta Air Lines is stopping agents from booking tickets for users.

Financial services are drawing a harder line. JPMorgan Chase and Charles Schwab confirmed that they are blocking AI agents from accessing accounts, citing security concerns, according to The Verge. Users have also said that Cars.com and Amtrak, which previously worked with their agents, are no longer accessible, though The Verge frames that as an indication that other companies may be quietly restricting access.

That creates an awkward reality for people trying to use agents as personal operators. If the task touches a marketplace, a bank, a travel provider or a site with strict bot controls, success may depend less on the AI model and more on whether the destination website allows that kind of automated use.

Amazon and Perplexity show the legal fault line

The clearest fight is still around Amazon. The company previously banned Perplexity from using its agentic shopping feature on Amazon and sued the AI startup. Amazon said at the time that it had repeatedly asked Perplexity to stop letting agents shop on its site, and argued that third parties buying goods for customers should be transparent and respect whether a service wants to participate.

Perplexity pushed back, saying Amazon’s real concern was that AI agents would reduce human exposure to ads. The startup argued that its agent acts on behalf of the user. A judge later ordered Perplexity to stop its agents from shopping on Amazon, but The Verge reports that an appeals court overturned that ruling in August. Amazon has since filed an amended complaint.

The dispute has not stopped other companies from entering the same territory. OpenAI, Meta and Google have continued rolling out agent products, according to The Verge. For users familiar with tools like ChatGPT and Google Gemini, the broader direction is clear: assistants are moving from answering questions toward taking actions. But action requires cooperation from the sites where those actions happen.

Partnerships may matter more than model quality

Meta has taken a partnership route for Muse rather than only fighting blocks. The company said in a post that rejecting a personal agent means rejecting the customer behind it, and said it would rather work with businesses on the concerns involved. The Verge reports that Meta has partnered with retailers including Walmart, Gap, Sephora and Best Buy so Muse can shop on their sites.

Shopify has also opened its network of storefronts to AI agents, according to The Verge. That fits with other commerce-agent moves we have covered, including Shopify opening its catalog to more AI shopping agents and Amex introducing an intent framework for AI shopping agents.

For businesses, the strategic choice is becoming clearer: allow agents under rules, charge for data access, or block them for security, policy or infrastructure reasons. For users, the practical advice is simpler:

• Treat agent workflows as conditional, not guaranteed.
• Keep a manual backup for checkout, booking and account access.
• Be especially cautious with financial accounts and other sensitive services.
• Expect some sites to work through partnerships while others reject agents outright.

Anti-bot systems could make normal browsing harder

Not every failed agent task is necessarily a deliberate AI ban. The Verge notes that agents may also trip existing anti-bot systems built to stop scams, hacking and automated abuse. Reports on Reddit cited by The Verge say Google has blocked some accounts using Muse because of suspicious activity, and others have reported Walmart checkout problems despite Meta’s partnership with the retailer.

Cloudflare is also part of the wider response. After launching a one-click block for AI crawlers in 2025, the web infrastructure company began blocking crawlers that scrape sites for both search indexing and AI training, according to The Verge. The broader concern is that AI bots can strain internet infrastructure, disrupt inboxes and make high-demand online tasks harder for humans.

Chris Canal, CEO of AI safety evaluator EquiStamp, told The Verge that these limits affect the perceived value of major AI companies because there are simply “fewer useful things” agents can do when websites resist them. He also warned that more defenses could make the web more annoying for people, with more prompts to prove they are human even on routine shopping sites.

That is the tension now facing agentic AI: users want delegation, websites want control, and bot defenses do not always know the difference between an assistant and an attacker.