AI browsers aren’t ready yet

A wave of half-baked agentic tools is splitting the web into human and machine lanes, pushing developers to rethink how sites are built and who they really serve

AI-powered browsers from companies like OpenAI and Perplexity are beginning to reshape how users interact with the internet by putting chat-driven assistants at the center of browsing. Despite their promise, most current AI browsers remain technically immature, struggle with complex tasks, and rely on expensive agent capabilities that only work smoothly under specific conditions.

Traditional browsers like Chrome, Firefox and Safari still provide more reliable search experiences — especially in areas such as shopping, travel research and form-based interactions — where accuracy, trust and verification remain critical.

Developers are now rethinking how websites should evolve, potentially creating separate layouts optimized for human users and automated AI agents.

Key takeaways

• Two-lane internet emerges: AI browsers and legacy browsers increasingly drive different user behaviors, forcing website developers to consider both human visitors and automated AI agents.
• Agentic browsing is still underdeveloped: AI browsers excel at summarizing content or completing simple online actions but struggle with multi-step tasks that involve visuals, pop-ups or structured decision-making.
• User trust remains limited: Most consumers feel comfortable using generative AI only for low-stakes or easily verifiable tasks, often combining AI answers with manual search to ensure accuracy.
• Website structure becomes a major barrier: AI agents perform best when sites contain structured, machine-readable elements rather than visually rich or narrative formats designed primarily for humans.
• Reliability and resource demands are problematic: Advanced agent features can stall, loop, overthink instructions or overload computing resources, limiting widespread adoption.
• Developer adoption is slow: Major online platforms are hesitant to build AI-optimized versions of their services, limiting the usefulness of AI browsers and slowing ecosystem development.
• Trust and transparency will be central to progress: For AI agents to autonomously carry out meaningful tasks such as purchasing or form completion, users will require strong guardrails, transparency and confidence that agent activity is not influenced by commercial incentives.

Get the full story at Bloomberg (subscription required)

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →