Agentic AI is rapidly moving from theory to practice. Instead of merely answering questions, these systems can now act on a user’s behalf by booking flights, comparing prices, filling out forms, or making purchases, which will significantly impact how travel products are booked.
The ease of use will allow users to bypass both travel agents and OTA sites to purchase non-ATOL bonded packages curated by Google, with Agentic AI guiding them through the complex booking process in minutes using pre-stored booking preferences and credit card details.
Google’s recent explanation of how Chrome will secure its upcoming agentic AI features provides a glimpse into how this future might operate at scale.
At first glance, the consumer benefits appear convincing and provide initial steps toward creating the “Digital Shopping Twin” that I have long forecast will alleviate much of the shopping hassle for users.
Travellers who are accustomed to navigating an average of 72 sites before booking their holidays to compare prices and read reviews often experience tab overload, decision fatigue, and repetitive online tasks. Agentic AI offers a form of digital delegation. However, as these agents become integrated into browsers, arguably the most commercially significant layer of the internet, because they control product access, crucial questions around trust, control, and commercial neutrality arise, as let’s be honest, Google don’t have a good track record here.
Google’s approach to agentic security is sophisticated, layered, and thoughtfully designed. Yet it also highlights a deeper tension: when the same company that controls the browser, the AI model, the ad ecosystem, and the marketplace is acting “on your behalf,” how confident can users be that convenience won’t quietly blur into influence based on who pays Google the most?
Why Google Agentic AI is a major threat to the UK Travel Trade.
The promise of agentic AI lies in time compression. Tasks that currently require dozens of micro-decisions: searching, filtering, logging in, and comparing options, can be reduced to a single voice instruction. “Book me the best-value Flight ticket from Manchester in May 2026 to Majorca for a 7-night duration, combined with the best-reviewed 4-star hotel costing less than £200 per night on a bed and breakfast basis in Alcudia with a private taxi transfer from the airport”.
In terms of productivity, this is transformative. For less tech-confident users, it could also be empowering, reducing cognitive and technical barriers to booking holidays online and posing a significant threat to the traditional role of travel agents and OTAs.
Forget hallucinations and silly errors: Google’s Architecture prevents these.
Google’s outlined safeguards demonstrate a strong understanding of the risk of AI hallucinations and mistakes. Central to this is a multi-model oversight system in which one AI plans actions, and another, called the User Alignment Critic and powered by Gemini, assesses whether those actions genuinely align with the user’s stated goal.
This separation is important. It mirrors patterns found in safety-critical systems, where execution and oversight are purposely separated, reducing the likelihood that Google Travel tools will mistakenly book incorrect items or fail even if the rest of the travel industry hopes they will and jump on any mistake to harangue them or disparage the reputation of AI Tools.
Equally important are Agent Origin Sets, which clearly define where an agent can read data and where it can act. Read-only zones may enable price comparisons on a travel site, whereas read-write zones limit interactions to authorised buttons or form fields and prevent items from being booked without explicit user approval.
Layered on top are URL monitoring systems to prevent unsafe navigation, the explicit requirement of user consent for sensitive actions such as payments and strict separation of password data. In short, Chrome’s agentic AI is not given free rein; it operates within a tightly constrained sandbox when processing customer travel requests.
The Big Risk: Commercial Bias Driven by Convenience
However, security is only part of the story. The more complex issue is alignment, not just with user intent regarding which holidays they prefer, but also with user interests in finding the best value provider, not merely the one suggested by Google because of the highest commission.
Most travellers still do not realise that the top of the Google search page is not dominated by the best providers but by those willing to pay the highest click costs to attract customers to their sites.
Google is in a unique and dominant position. Chrome manages access to the web. Gemini interprets intent. Google Shopping, Ads, Flights, and Search already facilitate commercial discovery. When a Google AI Travel Agent chooses which product to buy, which option is “best,” or which site is trustworthy, those choices are clearly influenced by Google’s commercial interests rather than the customers.
Even if no explicit manipulation occurs, defaults matter. Ranking logic matters. Training data matters. Over time, an agent that consistently chooses Google-affiliated services, preferred partners, or ad-optimised outcomes could quietly reshape user behaviour without users’ knowledge.
Unlike search results, which users can scan and override, agentic decisions are made on the user’s behalf. That efficiency and time saving are the key drivers, but it also reduces visibility.
A Powerful Tool That Demands Ongoing Scrutiny
Agentic AI in Chrome could save users significant time and mental effort when booking a holiday, and, bluntly, it’s tough luck for the travel industry if it replaces some of our jobs. However, although Google has discussed at length its technical safeguards, it remains silent on the clear ethical conflict of potentially recommending not the best travel options for customers, but those that generate the most revenue for Google.
It also triggers the age-old ATOL debate. Why should dynamically packaging retailers bear the substantial burden of “Principal” status when Google can offer precisely the same services without any bonding?
Both the UK Government as a whole and the Civil Aviation Authority must thoroughly examine Google’s plans for Agentic AI in travel and determine whether these constitute responsible activity or an abuse of a dominant market position.
To be entirely clear, I support customers’ right to choose, and using AI in the travel booking process will make holiday arrangements much simpler, but transparency of recommendations needs to be a legal requirement imposed on Google ASAP.