OpenAI has announced a significant upgrade to ChatGPT that they claim will allow it to autonomously complete complex tasks across the web. “ChatGPT agent”, launched on July 17, 2025, can navigate websites, analyze data, and execute multi-step workflows independently – raising questions about its potential impact on hotel discovery and booking.
NB: This is an article from Triptease
Subscribe to our weekly newsletter and stay up to date
What is ChatGPT agent?
While ChatGPT’s Operator offering (which was previously only available on the platform’s expensive “Pro” tier) was intended to be able to navigate the web like a human, this new iteration claims to combine the research and composition capabilities of the classic ChatGPT language learning model (LLM) with this agentic approach. This type of model was previously pioneered by Manus, an agentic tool that’s known for its ability to logically break down tasks in order to complete larger projects.
Users can make requests like “analyze three competitor hotels and create a presentation” or “research and plan a complete travel itinerary,” and in theory ChatGPT might be able handle the entire process autonomously.
The system operates through its own virtual computer environment, equipped with direct API connections to various platforms, allowing it to navigate through different steps in order to complete the tasks it’s asked to do. OpenAI reports that ChatGPT agent outperformed humans on realistic data science tasks and achieved state-of-the-art results on complex problem-solving benchmarks. But how this translates to practical travel booking remains unclear. And as we often see in AI, the actual applications are likely more limited, while the potential for error is quite high.
The Reality Check
While the technology sounds compelling, several factors suggest its short-term impact on hotel bookings may be limited.
- Usage Caps: Pro users get 400 agent messages per month, and other paid tiers receive just 40, with the option to add on additional usage by purchasing credits. For complex travel research involving multiple destinations, hotel comparisons, and itinerary planning, these limits would prove restrictive.
- Liability limitations: In order to protect themselves and avoid the agent making mistakes, there are a lot of limiters put in place by OpenAI like requiring explicit user confirmation in any situation with real consequences (ie booking a trip), and denying any higher-risk requests like making a bank transfer. And while these are totally sensible precautions, they also limit the usefulness of these agents–what’s the real benefit of employing an agent to do this if you still need to enter payment details at the end?
