Major websites are reportedly beginning to restrict autonomous AI agents from completing tasks on their platforms, raising new questions about how reliable agent-based automation will be for businesses. According to reporting cited in The Verge, some services from major AI vendors have recently faced blocks when trying to interact with popular consumer websites.

What happened

The Verge reports that users of AI agent services from companies including Meta, xAI, OpenAI and Instinct have encountered limits when those agents attempt to perform actions on third-party websites. The examples cited include Yelp, which reportedly said it only allows AI agents that participate in its paid data licensing programme, and Delta Air Lines, which is said to be preventing agents from booking tickets directly.

These reported restrictions point to a broader issue in the AI market: while AI providers are promoting agents that can browse, compare, book and transact on behalf of users, website owners may not want automated systems using their content, infrastructure or checkout flows without permission, commercial agreements or technical controls.

Why it matters for European businesses

For European companies, this matters because many expected business uses of AI agents depend on predictable access to third-party websites. If site owners increasingly block, rate-limit or contractually restrict these systems, some agent-based workflows may become less reliable than product demos suggest.

This has practical implications in several areas:

• Customer service and concierge use cases: Businesses hoping to let AI agents make bookings, gather quotations or complete support-related tasks across external websites may face inconsistent results.
• E-commerce and retail: Online retailers may need to decide whether to allow external AI shopping agents, restrict them, or create structured partner access instead.
• Marketing and SEO: If AI agents become a meaningful discovery and transaction layer, websites may need to think beyond traditional browser traffic and consider how automated visitors are identified, authenticated and governed.
• Automation projects: SMEs using browser automation or agent tools for repetitive web tasks should not assume ongoing access to supplier, travel, marketplace or comparison sites.

The development also highlights a strategic tension: AI vendors want agents to act across the web, while publishers, platforms and service providers want control over access, brand experience, pricing data, fraud prevention and infrastructure costs.

Who may be affected

The issue is likely to be most relevant for:

• SMEs adopting AI assistants for research, procurement, travel, lead generation or admin tasks.
• E-commerce businesses that may see new types of automated traffic from shopping agents or may want to expose products to them in a controlled way.
• Digital teams and IT departments responsible for automation reliability, website security, bot management and third-party integrations.
• Founders and operations teams evaluating whether AI agents can replace manual workflows or outsourced support tasks.
• Platforms and regulated sectors where identity checks, consent, pricing rules or fraud controls make autonomous actions more sensitive.

What companies should consider

European businesses do not need to react to every AI agent announcement in the same way, but they should treat web-based agents as a channel that may require governance, not just experimentation.

• Review dependencies: If an internal automation project depends on external websites, check whether access is officially supported, tolerated or likely to be blocked.
• Assess commercial terms: Some platforms may prefer licensing, API access or formal partnerships instead of uncontrolled agent traffic.
• Plan for fallback processes: Where agents are used for bookings, purchasing or data collection, keep manual or API-based alternatives available.
• Define your own website policy: Decide whether your business wants to allow AI agents, restrict them, or offer structured access through feeds, APIs or partner programmes.
• Coordinate AI, legal and security teams: Agent access can raise questions about terms of service, authentication, consent, scraping, fraud prevention and infrastructure load.

For now, the key takeaway is that AI agents may become an important business interface, but access to the web is not guaranteed. For European companies investing in AI automation, the practical question is no longer only what agents can do in theory, but where they are actually permitted to operate reliably.