AI companies have spent years promising assistants that can do more than answer questions. They want AI agents to book flights, shop online, manage accounts and handle everyday digital tasks without requiring users to navigate dozens of websites.
But there’s a problem. Some of the internet’s biggest platforms don’t want those agents operating on their websites.
Companies including Amazon, Delta Air Lines, JPMorgan Chase, Charles Schwab and Yelp have introduced restrictions affecting automated AI assistants. Their concerns range from customer security and unauthorized access to commercial control over online services.
The growing resistance creates a serious obstacle for technology companies developing autonomous AI systems, including OpenAI, Meta, Google and xAI.
As reported by The Verge, the conflict is exposing a fundamental weakness in the emerging AI agent industry. An assistant might be capable of completing a complicated task, but that ability means little when the website it needs refuses access.
Amazon Challenges AI Agents Trying to Shop on Its Platform
Amazon has emerged as one of the most visible opponents of unauthorized AI shopping agents. The company previously challenged Perplexity’s automated shopping technology, arguing that outside services must respect its platform rules when interacting with Amazon’s marketplace.
Perplexity has defended the idea that an AI assistant acting on behalf of a customer should be able to perform the same shopping activities as that customer. The disagreement has developed into a legal dispute over whether automated agents can access commercial websites without the platform owner’s approval.
For Amazon, the issue extends beyond automated traffic. Third-party AI agents could change how customers discover products, compare prices and complete purchases. Instead of browsing Amazon’s website and viewing its recommendations, shoppers might increasingly rely on an outside assistant to make decisions.
That shift could weaken the influence of traditional online storefronts and create uncertainty around advertising revenue, customer relationships and marketplace control.
Banks and Airlines Restrict AI Agents Over Security Concerns
Financial institutions have particularly strong reasons to question autonomous AI access. JPMorgan Chase and Charles Schwab have imposed restrictions affecting AI agents, reflecting concerns about sensitive customer information and automated account interactions.
An AI assistant that can access financial accounts introduces questions about authorization, identity verification and responsibility when something goes wrong. Even when customers want an assistant to perform a task, financial institutions must determine whether the system is trustworthy enough to handle account information securely.
Delta Air Lines has also restricted autonomous agents from booking flights through its platform. Airline reservations involve payments, personal details and changing availability, making automated transactions more complicated than ordinary website browsing.
Yelp has approached the issue from another direction by requiring certain AI services to use its commercial data licensing arrangements.
These decisions show that websites are not blocking AI agents for one universal reason. Some are protecting sensitive systems, while others are defending business models built around controlling access to their services.
Meta and OpenAI Encounter Barriers to Autonomous Web Access
The restrictions are creating problems for companies that want AI assistants to become everyday digital tools. Meta’s Muse, OpenAI’s Dots and other emerging autonomous systems are designed to interact with websites and perform tasks with limited user involvement.
However, reports of access problems show that advanced AI capabilities do not automatically translate into unrestricted internet access.
Amazon has challenged automated interactions involving outside AI assistants, while users of several agent-based services have encountered difficulties completing shopping activities.
The situation highlights a gap between what AI companies advertise and what their products can consistently deliver. An assistant might successfully identify a product, compare available options and prepare a purchase, only to encounter a restriction during checkout.
Meta has pursued partnerships with major retailers as an alternative to relying entirely on independent website access. Working directly with participating businesses could provide a more dependable way for AI agents to complete shopping tasks.
For developers, these relationships may become just as important as improvements in AI reasoning and automation.
Anti-Bot Security Systems Make AI Agent Access Even More Difficult
Website restrictions are not always the result of a company deliberately targeting AI assistants. Existing cybersecurity systems can also prevent autonomous agents from performing routine activities.
Many websites rely on automated defenses to identify suspicious traffic, prevent fraud and block malicious bots. AI agents can trigger those protections because they interact with pages through automated processes rather than conventional human browsing.
That creates a difficult distinction for website operators. A legitimate assistant attempting to complete an authorized customer request may resemble an automated bot trying to collect information or abuse a service.
Cloudflare has introduced controls that allow website owners to restrict AI crawlers and other unwanted automated traffic. Although AI crawlers and task-performing agents serve different purposes, both demonstrate how website operators are asserting greater control over machine-driven access.
OpenAI has also acknowledged that website security measures can prevent its agents from accessing certain pages or completing requested activities.
These limitations create frustrating experiences for users who expect autonomous assistants to work across different online services without interruption.
AI Companies May Need Partnerships Instead of Unrestricted Website Access
The growing resistance raises an uncomfortable question for the AI industry: can autonomous agents become widely useful without the cooperation of major websites?
AI developers have made substantial progress in building systems that understand instructions, plan multiple steps and interact with digital interfaces. Yet none of those improvements guarantees access to a retailer, bank or travel platform.
Website operators still control their own services. They can require authentication, impose restrictions on automated traffic and establish commercial agreements governing how third-party systems interact with their platforms.
This could push AI companies toward a partnership-driven model in which approved agents receive access through official integrations rather than attempting to navigate websites independently.
Retail partnerships, verified agent identities and secure application programming interfaces could make automated transactions more reliable. They could also help businesses establish clearer rules for customer consent and accountability.
The trade-off is that smaller AI developers may struggle to secure the same access as larger technology companies with established commercial relationships.
The Future of AI Agents Depends on More Than Smarter Models
The next stage of artificial intelligence is increasingly focused on action rather than conversation. Developers want assistants that can make reservations, purchase products, organize information and complete complicated online workflows.
But the internet was not originally designed around autonomous software acting independently on behalf of millions of users.
Businesses have legitimate concerns about security, fraud and unauthorized data access. AI companies, meanwhile, want their systems to operate across websites without negotiating individual agreements for every service.
Neither side can easily ignore the other.
Amazon’s conflict with AI shopping services, restrictions from financial institutions and the expansion of automated website defenses all point toward a changing relationship between AI developers and online platforms.
The companies that succeed in autonomous AI may not simply be those building the most capable models. They may be the ones that can establish trusted access to the services people actually use.
For now, AI agents are learning an important lesson about the internet: knowing how to complete a task does not necessarily mean being allowed to do it.
Sources
- The Verge – AI Agents Are Being Blocked Across the Web
- OpenAI Help Center – ChatGPT Agent
- Cloudflare – AI Crawl Control
