Experience of using the perplexity AI-powered answering system for educational content generation

The development of educational materials for courses taught in higher education institutions is a labour-intensive process that consumes a significant proportion of academic staff time. With the advancement of text generation tools powered by artificial intelligence, educators have gained the opportunity to enhance and accelerate this process.
The purpose of the study is to analyse the capabilities and specific features of the web-based search engine Perplexity AI, which is built on a large language model, for the development of theoretical lecture content and test questions based on the generated material.
The study addresses the following objectives: to identify the capabilities and features of Perplexity AI as described by researchers in academic publications; and to analyse its potential for developing educational content, particularly theoretical material for lectures and test questions derived from it.
Research methods. The study employs the literature review method and the method of qualitative analysis of responses generated by Perplexity AI. The literature review was conducted through searches in the Scopus and Web of Science databases to analyse recent research and publications in order to identify the capacities and limitations of Perplexity AI.
As a result of this analysis, the advantages and drawbacks of Perplexity AI in tasks related to text generation and user queries were identified. The qualitative analysis method was applied to examine practical outcomes based on hands-on experience using the tool in the process of generating educational content. For the purpose of this study, the authors developed educational materials with the help of Perplexity AI, including eight lectures for the course “Immersive Technologies” and accompanying test questions based on the generated content. The relevance, accuracy, currency, and informativeness of the content were then analysed using the provided primary sources.
It has been established that the strengths of Perplexity AI include the provision of precise sources and follow-up questions alongside query responses; effective handling of general knowledge, reasoning, and evidence-based content; reliable responses supported by relevant data; extensive search capabilities; a focus on clarity and efficiency of answers; the ability to sustain long-form conversations; and maintenance of con- text in extended interactions. The identified weaknesses of Perplexity AI compared to ChatGPT-4, Google Bard (Gemini), and Chatsonic include lower accuracy; inconsistent performance across categories; reduced resistance to hallucinations; occasional use of unreliable sources such as blogs for citations; and issues with performance and readability.
The study revealed that for the effective development of educational content – particularly theoretical lecture material – Perplexity AI can be used with prompts following this sequence: an initial prompt requesting the generation of a lecture outline; individual prompts for each lecture topic with emphasis on the text style and the use of appropriate sources; and clarifying prompts to add or refine content within the generated text.
Conclusions. The development of lecture material and test questions using Perplexity AI can be considered satisfactory, even within the framework of the free plan. The majority of the generated educational content was well-structured, relevant, accurate, up-to-date, and informative, with source quality being verifiable through the hyperlinks provided by the system. Achieving such results requires carefully formulated prompts and appropriate use of the references and suggestions offered by Perplexity AI.