DEEP RESEARCH AND THE ARTIFICIAL INTELLIGENCE INFORMATION REVOLUTION ChatGPT, from its introduction in November 2022,1 started a revolution in the knowledge landscape similar to that of the information revolution Google2 brought about in the early 2000s. Google and other search engines made all the information in the world available at our fingertips through the internet. This, however, contributed to an infodemic-an information epidemic, i.e., an abundance of information impossible to consume and process within a human lifetime. Generative Artificial Intelligence (AI) Agents such as ChatGPT,3 Gemini,4 and Claude5 are now at the forefront of the next information revolution. These advanced AI systems possess the ability to consume, process, and reiterate information at a scale far beyond human capacity. The emergence of conversational AI assistants like ChatGPT has effectively placed a powerful information resource–akin to an “information genie”–at the user’s disposal. These AI models have made high-quality information easily accessible to anyone with a computer and the internet. Even though paywalls do exist for some features, most can be accessed freely within reasonable limits. However, a significant challenge lies in the fact that AI systems often operate without human oversight and are prone to generating hallucinations; that is, they present fiction as facts with such confidence that it could mislead even the most discerning users. This has led to a situation where the AI can never be fully trusted, and its answers require validation. This element of human oversight is essential and unlikely to be eliminated anytime soon, as AI is not yet truly intelligent. Semantic understanding and the ability to reiterate information do not always translate to conceptual understanding, which the AI often lacks. However, with modern developments in the AI space, AI chatbots like ChatGPT have come up with a new feature called Deep Research.6 This allows the AI to provide reference links to the claims it makes in its answer. This greatly increases the credibility of answers generated by ChatGPT. Even though it might still make mistakes, at least now it is easier to cross-check the AI’s claims by reading the linked articles. This feature could easily be used to create a properly referenced narrative review on any topic, purely using AI. Before this, AI augmented search tools like Scispace7 and Perplexity8 did exist, which provided short answers to user queries based on journal articles or other sources, and provided the reference links for the same. However, the integration of similar capabilities directly into widely used AI tools such as ChatGPT that can create long essays on a topic has transformed what was once a niche feature into a mainstream functionality, readily accessible to a large audience. Even though there are currently limits to the number of Deep Research tasks performed by ChatGPT (up to 5 per month for the free user and 25–250 tasks per month for paid users).9 This could increase and even be removed in the future as the technology improves. Other generative AI agents like Gemini, DeepSeek, and Perplexity have also integrated a Deep Research-like feature into their systems. Due to the fierce competition and huge investments in AI technology, such AI tools would keep on improving with time. This will lead to increased accessibility for the masses by driving down the price. DEATH OF THE NARRATIVE REVIEW This AI information revolution would eventually lead to the death of narrative reviews, which are an important aspect of journal articles. Narrative reviews are often just a compilation of information from already existing articles, added with the author’s perspective and biases. The narrative reviews are characteristically different from the Systematic Reviews (and Meta-analyses),10 where specific criteria are used to select quality studies which are weighed based on design, population studied, and quality, and conclusions are drawn using statistical methods. Systematic reviews and meta-analyses are not currently under the direct threat of AI, even though AI tools like Rayyan AI11 have become a valuable digital aid while doing them. However, the case of narrative reviews is different. An AI tool like ChatGPT with its deep research feature can produce an almost good enough narrative review within a few minutes. This could lead to a surge in narrative review submissions to journals by authors who may overuse this feature to increase their publication counts. Not to be mistaken with junk articles, these narrative reviews may be on interesting topics and could compile information and perspectives that could be helpful to other researchers. Nevertheless, with further development of AI, narrative reviews will eventually lose their relevance. If anyone can access a detailed, well-referenced review on any topic within minutes at their fingertips, it would dissuade people from taking journal subscriptions and spending time and money going through narrative reviews that could have been written with AI assistance. Hence, the days of narrative review publications are numbered. AI can almost always do a better job in compiling and processing information and reproducing it across various writing styles–whether academic, formal, casual, or essayistic. Human narrative reviews often take up time, resources, and introduce personal biases and perspectives that may alter the review narrative. Furthermore, they are limited or restricted by their language proficiency and writing styles that are not adaptable or malleable depending on the journal’s tone or objective. AI is also not completely free of bias, but often retraining AI for an identified bias is easier than identifying a bias in humans and reeducating them. Claims like AI reviews lack the human touch and flavor are no longer valid arguments, as today, there already exist AI chatbots that can mimic human conversation and style so well that they can often have specific personalities and act as AI girlfriends and personal secretaries. Hence, the so-called “Turing Test” is definitively and invariably passed.12 The ethics of using AI in publishing could not be relevant with regard to narrative reviews, as they are not original research and are just a cross-referenced compilation of knowledge. And with time, when narrative review itself becomes irrelevant and redundant, due to lack of readership, any discussion on the ethics of using AI for writing reviews would no longer be relevant. Currently, there could be challenges where the AI agent’s database may not be updated on time, unlike the journals, which are published at more frequent intervals. Furthermore, the fact remains that most journals hidden behind a paywall are somewhat inaccessible to ChatGPT’s Deep Research. However, this could easily change in the future, considering the current pace of AI development. CONCLUSION Similar to how Google eliminated the need for website directories and dictionaries, Wikipedia13 eliminated the need for encyclopedias; AI Chatbot agents are going to replace narrative reviews eventually. Embracing the change would help us transition better into this new world of AI-assisted information than resisting the inevitable. This could free up human creativity, imagination, and time to be spent on better fruitful avenues of research and development than just compiling and reiterating information. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Chalissery et al. (Wed,) studied this question.