The race to build AI assistants capable of navigating the internet independently is running into growing resistance from the very websites these tools need to access. Several major online services are restricting automated agents, complicating efforts by technology companies to turn AI into a tool that can handle everyday digital tasks on users' behalf.
Agents developed by OpenAI, Meta, Google, xAI and other companies are being marketed as personal assistants that can search websites, manage email, make purchases and arrange travel. However, reports from users in recent weeks suggest that access restrictions are preventing some of these tools from completing tasks on popular platforms.
The restrictions reflect a widening divide between AI developers seeking to automate online activities and businesses concerned about security, data licensing, advertising revenue and control over how their services are used.
Financial, Travel and Consumer Platforms Tighten Access
Users of Meta's Muse, xAI's Grok, OpenAI's Dots and Instinct have reported difficulties getting their agents to operate on certain websites.
Some companies have confirmed that they are deliberately restricting access. Yelp told TechCrunch that AI agents can use its data only through its paid licensing programme. Delta Air Lines, meanwhile, is blocking agents from booking flights on behalf of customers.
Financial institutions are also taking a cautious approach. JPMorgan Chase and Charles Schwab have confirmed that they prevent AI agents from accessing customer accounts, citing security concerns.
Other users have reported losing access to services that previously worked with their agents, including Cars.com and Amtrak. These reports suggest that some restrictions may be introduced without prominent public announcements, although the available information does not establish whether every reported access problem resulted from an intentional ban.
For AI developers, such barriers create a practical problem: an agent may be capable of carrying out a task but unable to complete it if the relevant website refuses automated access.
Amazon's Dispute With Perplexity Foreshadowed the Conflict
The pushback against autonomous browsing is not entirely new. Amazon challenged Perplexity's AI-powered shopping feature last year, banning the agentic tool and suing the AI startup.
Amazon argued that it had repeatedly asked Perplexity to stop allowing its agents to shop on the platform. The ecommerce company maintained that third parties purchasing products on behalf of customers should act transparently and respect a service provider's decision about whether to permit that activity.
Perplexity disputed Amazon's position, arguing that the retailer had an incentive to prevent AI agents from navigating its website because automated shopping could reduce the number of people viewing its advertisements.
That disagreement exposed a commercial tension likely to become more significant as AI-driven shopping grows. If an agent can find products, compare offers and make purchases without a customer browsing a retailer's website, the retailer could lose opportunities to display ads and promote additional products.
At the same time, businesses may be concerned about automated systems accessing information or performing transactions under conditions they have not approved.
Security and Commercial Interests Shape the Restrictions
The reasons for blocking AI agents differ between industries, but security and control over digital services are emerging as central concerns.
Banks and brokerage firms handle sensitive financial information and must protect customer accounts against unauthorized access. Allowing an autonomous agent to operate within those environments raises questions about authentication, permissions and responsibility for mistakes or unintended transactions.
Travel companies face a different set of considerations. When an agent attempts to book a ticket, it may need to navigate availability, pricing, passenger details, and payment procedures. Delta's restrictions illustrate how a company can limit automated transactions even as AI developers promote tools designed to complete bookings independently.
Data licensing presents another challenge. Yelp's position indicates that access to its information is not automatically granted simply because an AI system can retrieve it. Its licensing programme provides a framework through which agents can access data under commercial arrangements.
Retailers also have reasons to protect the existing economics of their websites. Advertising, sponsored placements and customer engagement remain important parts of online commerce, potentially creating friction with agents designed to minimize the time users spend browsing.
AI Assistants Face a Web They Do Not Control
The restrictions expose a limitation in the vision of fully autonomous digital assistants. While AI companies can develop agents capable of interpreting instructions and interacting with websites, they cannot independently guarantee that every service will permit those interactions.
A user might ask an agent to compare products, book transport or complete an online transaction, only for the process to stop when the website blocks automated access or requires a form of verification the agent cannot complete.
This creates a gap between what AI companies promise and what their tools can reliably deliver across the wider internet.
The outcome may depend on whether AI developers and website operators establish acceptable ways for agents to access information and conduct transactions. Paid licensing arrangements, approved integrations and clearer rules governing automated activity could provide routes forward, although the reported developments do not establish a single solution accepted across the industry.
For now, the growing restrictions show that the future of agentic AI will depend not only on how capable the technology becomes, but also on whether websites, financial institutions, retailers and other service providers are willing to let it operate on their platforms.
Source: The Verge