Large language models (LLMs) can streamline information search, enhancing user productivity and experience. LLMs have revolutionised how users interact with information. LLMs, such as OpenAI’s GPT series, have revealed ground breaking results in natural language generation and comprehension. These models can obtain valuable insights from that data to accomplish specific goals and tasks through flexible adaptation. However, LLMs have difficulty responding accurately to queries and may exhibit “hallucination”, where they provide responses that contain false information. Techniques such as Retrieval Augmented Generation (RAG) have been developed to address these challenges. A personalised customer support chatbot using the RAG model for a travel system named TravQuery is presented in this paper. The TravQuery chatbot is trained on a synthetic travel-based Question Answer dataset. A RAG architecture retrieves the most relevant document from the provided dataset and presents it to the Llama 2 LLM to generate an accurate response. This approach reduces the likelihood of hallucinations and facilitates the creation of more accurate responses. The model leveraged the RAG architecture to achieve an accuracy rate of over 70%. This demonstrates the chatbot’s ability to deliver high-quality, contextually relevant interactions within the travel domain. The outcomes of TravQuery are compared with the ChatGPT-3.5 model to evaluate the difference in the responses produced by both models.