<p>The chatbots work as a communication tool among machines and humans to obtain the most appropriate outcomes for human inputs. Moreover, the generative chatbots are designed by fusing the sequential models with Natural Language Processing (NLP) techniques. But, these techniques are in a sequential nature that affects the processing outcomes. Hence, it is essential to develop a novel chatbot interaction framework which supports English Language. In the beginning phase, necessary text data for the validation are collected from the benchmark database named “Healthcare Chatbot” (2·04MB) and offered to the text pre-processing phase. Then, Generalized Autoregressive pre-training for Language Understanding (XLNet)-based word embedding procedure is utilized to perform the word embedding. Next, the word-embedded outcomes are provided to the generative AI-based chatbot interaction phase, which utilized the developed Generative Tuned-Long Short-Term Memory with Region Attention (GT-LSTM-RA) model, which is tuned using an Enhanced Random Variable-based Dung Beetle Optimizer (ERV-DBO). Later, chatbot interaction outcomes are attained from the developed GT-LSTM-RA and different validations are performed to verify its effectualness. Thus, results in the reduction of Word Error Rate (WER) and Sentence Error Rate (SER) as 0.03 and 0.02 in 5^th fold when competes over other classical chatbot interaction techniques.</p>