ChatGPT is in the news almost daily. The possibilities for ChatGPT are almost limitless, and certainly, this powerful new tool has a place in the hospitality industry in some capacity. But is it a reliable source for creating guest review response content?
What is ChatGPT?
ChatGPT is an artificial intelligence chatbot developed by OpenAI and launched in November 2022. It is capable of generating text in a wide range of styles and for different purposes, ranging from blog posts (I could probably have gotten it to write this one) to math equations to essays with remarkable detail and flow. It’s designed with a strong focus on interactive conversations, which means the user can tailor and fine-tune answers with follow-up questions and redirection as needed.
According to AssemblyAI, the creators used a combination of both Supervised Learning and Reinforcement Learning to fine-tune ChatGPT, “but it is the Reinforcement Learning component specifically that makes ChatGPT unique. The creators use a particular technique called Reinforcement Learning from Human Feedback (RLHF), which uses human feedback in the training loop to minimize harmful, untruthful, and/or biased outputs.”
I’ll admit: all the above still sounded confusing, so I went directly to the source and asked ChatGPT to explain itself to me in simple terms. Here’s the answer:
“Sure, I'd be happy to explain ChatGPT in simple terms!
ChatGPT is an artificial intelligence language model that has been trained on a vast amount of text data to understand and generate human-like responses to text-based queries. It uses a neural network architecture to analyze the structure and meaning of language and generate responses that are contextually relevant and linguistically coherent.
When you ask a question or make a statement to ChatGPT, it analyzes the text you've provided and uses its understanding of language to generate a response. It's designed to learn from its interactions with users, so the more it's used, the more accurate and helpful its responses become. Overall, ChatGPT is a powerful tool for natural language processing that can assist with a wide range of tasks, from answering simple questions to engaging in complex conversations.”
General ChatGPT limitations
Sounds almost too good to be true? Well, there are definitely limitations. For instance, ChatGPT can sometimes write plausible-sounding answers that are in fact nonsensical, which the user would only realize if they had a firm grasp on the material at hand. And the system can be sensitive to tweaks in phrasing; much like you might have to phrase a question several times to your phone’s AI or household AI (think Siri or Alexa) to get the ‘correct’ answer, the same can happen with ChatGPT.
In using ChatGPT, I also found the program to repeat or overuse certain phrases and be verbose. Currently, it only ‘course corrects’ when the user asks follow-up questions or directions, whereas it would be helpful if the chat program asked clarifying questions of its own, to narrow down topics and ‘get it right the first time’. According to OpenAI, continued user feedback will help alleviate some of these shortcomings.
So, how does ChatGPT perform when answering specific online hospitality reviews?
I will admit ChatGPT makes a good first impression. In screenshot number one, I input a real hotel review and asked it to respond to the guest.
As shown above, the response is well-written with proper grammar, and even the tone is impressive. The problems?
1. Echo. For the most part, ChatGPT repackages what the reviewer shared within the review. As a general rule, there’s nothing wrong with echoing the guest’s sentiments and mirroring their praises or feedback, but in using AI, you lose the opportunity to expand upon what the guest is saying, using their comments as a launching point for further detail. For instance, in this example, the guest does not say that the hotel is centrally located in the heart of the city, so ChatGPT won’t ‘add’ that fact. The workaround: you can give the AI follow-up prompts, such as ‘mention the hotel’s downtown location’, or ‘touch upon the central location of Hotel X’. But then you’re defeating the purpose of using a chatbot to save time and creative resources.
2. Nothing new. The bot also can’t make statements, or statements that make reasonable assumptions based on what the reviewer shares. These are valuable to pepper into review responses. For example, Hotel X has a wonderful wine bar, but since the guest didn’t refer to it in the review, the response does not include something like, ‘We hope you found time to linger in our wine bar to further enjoy our views’. This, too, would be content that would need to be added by a human responder.
3. Fake news. And as mentioned above, ChatGPT doesn’t fact-check, so some details might be completely made up. Or, if the reviewer included incorrect information, such as hating the off-site parking, when in fact, Hotel X does not have off-site parking, this information will not be taken into account in the response. Whereas a human responder might redirect and correct with a reply like, ‘We apologize if our valet parking option was not clearly marked for you,’ or, ‘We wish you had utilized our attentive valet services’, the AI will simply spit out something like, ‘we apologize that our off-site parking did not deliver’…echoing an incorrect statement about the hotel. ChatGPT won’t know the other side of the story, or how to explain things like parking charges being at market rates, or the process for incidental deposits. Again, a human responder with extensive knowledge of the hotel would need to check the content, defeating the purpose of utilizing an AI bot.
4. Bot voice. Next, there’s the matter of voice, which varies by hotel brand. Some hotel responses may need to sound formal and highly professional, whereas some brands may prefer a more casual, approachable touch. I played around with changing the voice of a response, as seen in the second screenshot below (which is utilizing a different review than Example #1).
As you can see, it did a pretty good job of ensuring the tone of the response was causal. When I asked it to become more formal, however, more repeated phrases occurred.
Additional problems with ChatGPT
When there are grammar inconsistencies and punctuation errors in the original review, which there often are, ChatGPT has a harder time intuiting intent or meaning than the human eye. At times, a human responder is needed to interpret what the guest meant. For example, when a review referred to the excellent service provided by Rene and Carson, but the reviewer punctuated this as ‘Rene, Carson’, the AI spit back out: ‘We appreciate your praise for Rene Carson’. As an expert on the hotel, I know the reviewer meant Rene and Carson, and corrected the response accordingly.
In addition to having to monitor and correct these types of errors, the operational task of copying and pasting multiple reviews into the ChatGPT system every day is still time-consuming, especially when you consider the time spent proof-reading responses to make sure facts are correct and reviewer intention is understood.
While a helpful tool, ChatGPT is not ready to replace the human touch when it comes to hospitality review responses There is simply still too much room for error, and error can lead to increased guest dissatisfaction in negative reviews, as well as a lost opportunity to celebrate the positives of a property in positive reviews. Hotels and restaurants need reviews answered by humans well-versed in each property and the art of the English language.
To learn more about how we can help your hotel, restaurant or other consumer-facing venture create, implement, and manage a reputation management strategy, we invite you to send us an email through our secure contact form or call us at 855-979-6800.