If you have ever tried to book a hotel online and found yourself unsettled by the AI chatbot trying to help you, science has your back. A new study from Texas A&M College of Agriculture and Life Sciences confirms that hotel booking chatbots are genuinely creeping people out, and it is actually hurting bookings.
What is giving hotel chatbots their creep factor?
Researchers surveyed 340 adults in the UK who had used chatbots to book hotels and found three main culprits behind the ick factor: inaccuracy, deceptive behavior, and intrusiveness. Inaccuracy was by far the biggest offender, triggering a negative response more than four times stronger than the other flaws.
It results in quoting incorrect rates for the rooms, bungled cancellation policies, or questions that just get dodged entirely. That discomfort is not just a vibe. It cut users’ willingness to keep chatting with the bot by nearly 38% and nearly doubled the chances they would delay or ditch the booking altogether.
Researchers also flagged something called the “uncanny valley” effect, where a chatbot’s failures feel even creepier the harder it tries to sound human. Lead researcher Babak Taheri summed it up perfectly, saying that when a human-like system fails to actually behave like one, it triggers something deeper than disappointment in users.
There is a simple fix that hotels mostly ignore
The good news is that the researchers found a simple solution that most hotels are not using. When a chatbot declares it’s an AI, users are far more forgiving of its mistakes. A simple opener like “Hi, I am your AI assistant” goes a long way.
Researchers also recommend making it easier to reach a real human for complex queries and invest in upgrading the AI itself so it can actually handle the basics without fumbling.
This research lands at a fascinating moment, because AI travel booking is the hottest thing in tech right now. Google recently added AI trip planning to Search, and Uber just launched hotel booking through Expedia inside its app.
Background: The rise of AI in hospitality
The adoption of AI chatbots in the hotel industry has accelerated rapidly over the past five years. Major hotel chains like Marriott, Hilton, and IHG have deployed conversational agents on their websites and mobile apps to handle booking inquiries, modify reservations, answer FAQs, and upsell amenities. The promise is 24/7 availability, reduced labor costs, and faster response times. However, the execution often falls short. Many chatbots are built on narrow, rule-based systems that fail when faced with unexpected user inputs. This leads to the inaccuracies identified in the Texas A&M study.
The hospitality sector spends billions on technology to enhance guest experiences, but a poorly performing chatbot can alienate customers before they even step foot in a hotel. The study’s findings highlight a disconnect between the technology’s potential and its current reality. Inaccuracies in quoting rates or policies are particularly damaging because they erode trust. A guest who receives a wrong cancellation policy quote might later be charged unfairly, leading to negative reviews and lost revenue.
The uncanny valley explained
The concept of the uncanny valley, first proposed by roboticist Masahiro Mori in 1970, describes the discomfort people feel when a robot or artificial agent looks or acts almost human but not quite. In the context of chatbots, the effect is triggered when the AI attempts to use natural language, emojis, friendly greetings, and conversational fillers but then fails to maintain coherent, accurate dialogue. The mismatch between the human-like presentation and the machine-like performance creates a sense of eeriness. This is exactly what the Texas A&M researchers observed. Participants reported feeling uneasy when a chatbot answered with human-like phrases but then failed to understand simple follow-up questions or provided contradictory information.
From a psychological standpoint, humans have evolved to expect consistent behavior from entities that appear human. When that expectation is violated, a cognitive dissonance arises, often manifesting as distrust, irritation, or anxiety. In a booking scenario, where money and personal plans are at stake, this discomfort translates directly into abandonment of the booking process. The study quantified this: a mere 10% increase in perceived creepiness led to a double-digit drop in conversion.
Practical implications for hoteliers
For hotel operators, the findings offer a clear roadmap. First, ensure chatbot transparency. A simple disclosure that the user is talking to an AI lowers expectations and reduces the uncanny valley effect. Second, invest in robust natural language understanding and backend integrations so that the chatbot can access real-time pricing and policy data. Third, implement a seamless handoff to human agents for complex or sensitive queries. The study notes that 67% of users who encountered a problem with a chatbot still wanted to complete the booking, but only if they could quickly speak to a person.
These recommendations are not expensive to implement. Many chatbot platforms offer APIs for live agent escalation, and disclosure messages can be added with minimal code changes. Yet the researchers found that a majority of hotel websites do not clearly label their chatbots as AI. Some even use human names and profile pictures, which backfires when the chatbot fails to perform. The study’s senior author, Dr. Taheri, suggests that hotels should “stop pretending the bot is human and start focusing on making it accurate.”
Broader industry trends
The findings come as the travel industry experiments with AI in other areas. Booking.com, Expedia, and Hopper have rolled out AI travel planners that suggest destinations and itineraries. Google’s Search Generative Experience now includes travel itineraries in search results. These tools often use large language models (LLMs) like GPT-4, which are more sophisticated than the typical hotel booking chatbot. However, even advanced LLMs can hallucinate hotel rates or policies, especially if they lack real-time data access. The study’s insights about transparency and fallback to human agents are equally applicable to these new generations of AI.
Uber’s partnership with Expedia to offer hotel booking within its ride-hailing app is another example of AI-driven convenience. But if the underlying booking chatbot suffers from the same flaws, it could undermine the entire user experience. The Texas A&M researchers advise that any company deploying AI for travel transactions should conduct user testing for creepiness metrics, not just accuracy or speed. Emotional factors like trust and comfort are critical for conversion in high-stakes purchases like hotels.
Methodology of the study
To arrive at these conclusions, the research team designed a survey with 340 participants who had used a hotel booking chatbot in the past 12 months. The sample was balanced for age, gender, and income to represent typical UK travelers. Participants rated their experience on scales measuring perceived inaccuracy, deceptiveness, intrusiveness, creepiness, and booking intention. Statistical analysis used structural equation modeling to identify causal paths. The results showed that inaccuracy had a direct effect on creepiness (β = 0.42, p < 0.001) and an indirect negative effect on booking completion through creepiness. The effect of deceptiveness and intrusiveness was smaller but still significant. The study is published in the journal Computers in Human Behavior and has been peer-reviewed.
Dr. Taheri, who leads the hospitality research group at Texas A&M, noted that the UK market was chosen because of its high rate of online hotel bookings and chatbot usage. The findings are likely generalizable to other Western markets, though cultural differences in trust of AI may affect the strength of the effects. Future research will explore whether similar patterns hold in Asia and the Middle East, where chatbots are also widely used in hospitality.
In the meantime, hoteliers have a clear incentive to overhaul their chatbot implementations. A 38% drop in willingness to continue chatting and a near-doubling of abandoned bookings represent significant revenue leakage. With the simple fixes identified by the study—clear AI labeling, accurate data integration, and easy human escalation—hotels can turn a creepy experience into a helpful one, boosting both guest satisfaction and bottom line.
Source: Digital Trends News