Artificial Intelligence (AI) and the literature review process: Tools
Generative AI tools have benefited doctoral research students in three key ways (Kumar and Gunn, 2024):
- facilitating discovery and connections - helping you to discover key works and researchers and to see how these are connected. Tools also help to identify themes and trends.
- increasing efficiency - saving you time by identifying highly-cited articles and often summarising them, enabling you to prioritise your reading.
- making the process less intimidating - providing overviews of a topic, breaking down the literature and presenting complex information in a more accessible way helps to ensure you are not overwhelmed by the literature.
This page lists some of the AI tools used by researchers when conducting a literature review:
Click the use-cases links for instructions and examples of prompts.
Academic Research Tools
These tools assist researchers with searching, analysing and visualising academic publications:
Follow the use cases links for guidance and examples applying to conducting a literature review.
Connected Papers
Connected Papers (Smolyansky, 2020) is based on the Semantic Scholar database. Its premise is that two papers that have highly overlapping citations and references are presumed to have a higher chance of treating a related subject matter. The graphs that are produced are designed to highlight the most important and relevant papers immediately. You will only get 5 graphs per month for free.
Use cases: visualising search results, finding similar work.
Limits: You will only get 5 graphs per month for free.
Elicit
Elicit will do this for you saving you time and allowing you to then synthesise the information. Its own user survey found that 10% of respondents said that Elicit saves them 5 or more hours each week and that, in pilot projects, Elicit saved research groups 50% in costs and more than 50% in time by automating data extraction work they previously did manually (Elicit, 2023).
The free basic version allows you to extract data from papers and upload your own papers. However, only priced versions of the product will give you summaries of papers and allow you to extract the information into csv and bib formats.
Use case: Data Extraction.
Inciteful.xyz
Inciteful.xyz has pulled information and/or inspiration from Semantic Scholar but also three other data sources: OpenAlex, CrossRef and OpenCitations. It has used these sources to help researchers get up to speed on a new topic, to find the latest literature or to work out how two ideas are connected (Weishuhn, 2024). There are two tools that are under active development: Paper Discovery and the Literature Connector. Paper Discovery builds a network of papers from citations, analyses the network, and allows you to get up to speed on a topic. by finding the most similar papers, important papers as well as prolific authors and institutions. Literature Connector allows you to enter two papers and it will give you an interactive visualization show you how they are connected by the literature.
Use case: visualising search results.
LitMaps
LitMaps is also based on the Semantic Scholar database. As with the above, it uses the citation network to construct graphs to visualise the research landscape of your topic. Its Discover tool enables you to find gaps in your own literature reviews and to upload existing literature reviews to find not only more recent papers but also papers that the reviews may have missed.
Use cases: visualising search results.
Rayyan AI
Rayyan AI is a web-based automated screening tool, developed by Qatar Computing Research Institute (QCRI), which launched in 2014.
It uses text mining techniques to identify relevant information using statistical pattern learning that recognises patterns in the data.
Use case: Screening.
Research Rabbit
Research Rabbit also uses Semantic Scholar to search for papers but also uses PubMed to find biomedical and life sciences papers. It is committed to remaining free to researchers.
Use cases: visualising search results.
SciteAI
SciteAI, designed by Nicholson et al. (2021), categorizes citations according to whether the articles were contrasting, supporting or mentioning. Scite.ai is very keen for you to take a 7 day free trial but will provide you with a certain amount of free information that you can use for critical evaluation.
Use case: Synthesis.
Semantic Scholar
Semantic Scholar is a free, AI-powered search and discovery tool that helps researchers discover and understand scientific literature most relevant to their work.
Semantic Scholar sources its content via web indexing and from partnerships with scientific journals, indexes, and content providers. These include leading publishers such as Springer Nature, Taylor & Francis, Wiley and Sage and institutions such as the IEEE and ACM. It does not contain content from Elsevier, Emerald or Oxford University Press. There is a list of its publisher partners.
Semantic Scholar does not support Boolean searching or wildcards. Its tutorial pages tell you that Semantic Scholar covers papers across all subjects including biomedicine, computer science, business, history, and economics. Its Ask This Paper and Generative Term Understanding features employ generative AI.
Use cases: Searching.
Generic AI Chatbots
AI companies initially adopted a chat interface to allow the public to interact with language models from a Web browser. The intuitive interface, the ability to generate human-like answers to questions posed using natural language, rapid response times and the high accuracy of the responses has encouraged high use of these chatbots by students. A survey in Spring 2023 of 5894 Swedish university students found over half (55.9%) had a positive attitude towards their use but a similar number (54.2%) expressed concern about their future impact on students' learning (Stohr et al., 2024). AI chatbots can help with research tasks by generating article or content summaries, by giving you information about a topic, by suggesting areas for further research and by helping you to analyse and structure the material. They can be used with speech-to-text or text-to-speech software to help students with visual impairment. They can assist in the development of language skills for international students and can assist with programming, report writing and project management skills. They have the potential to provide a wide range of benefits and opportunities for students (Kasneci et al., 2023).
Below, we review popular AI chatbots' main characteristics and use cases in the context of a literature review:
- Microsoft Copilot (Open AI GPT 4)
- Chat GPT (Open AI GPT 3.5 and 4)
- Claude.ai (Claude 3)
- Gemini (Google Gemini Pro)
Any information you enter into free AI tools may be disclosed publicly, become the property of the website, or be shared with third parties.
Look for confidentiality notices or consult the terms and conditions or privacy policies before uploading copyrighted or personal materials online.
Microsoft Copilot
Input: Text, voice, image files.
Notable features: Connected to the Internet, references sources in its answers, 3 levels of verbosity.
Use cases: abstract and title creation, scoping, searching.
Cost: free or included in Microsoft 365 plans.
Limits: Number of allowed chats per day varies depending on the load and MS 365 plan.
History: Microsoft's Copilot (former Bing Chat) is an AI-based research assistant powered by OpenAi's GPT-4. Copilot also integrates generative AI features into Microsoft 365 apps such as Word, Excel, PowerPoint, Teams and ClipChamp.
Chat GPT
History: ChatGPT launched in November 2022. It was been built on top of OpenAI’s GPT-3 family of large language models (Generative Pre-trained Transformer) and was fine-tuned using both supervised and reinforcement learning techniques (OpenAI, 2022).
It is estimated to have reached 100 million monthly active users in January 2023, just two months after launch, making it the fastest-growing consumer application in history (Hu, 2023).
Version 4
Input: Text, images, audio and video.
Notable features: Although it is trained on data until June 2024, ChatGPT is designed to stay current by accessing the web for real-time information when necessary, reducing its tendency to hallucinate, with links to the sources it references. Its Canvas model allows you to edit any word, sentence or paragraph and request changes saving you time from asking ChatGPT to rewrite its entire response. Canvas also allows you to structure your interactions with ChatGPT providing you with more visual and organised conversations. Uses Dall-E to create or analyse images.
Use cases: abstract and title creation, scoping, searching.
Cost: The free version gives you access to GPT-4o mini. It gives standard voice chats and limited access to GPT-4o. The Pro subscription gives you unlimited access to o1, o1-mini, and GPT-4o and unlimited voice chats. The o1 models are a new series of AI models from OpenAI which have been designed to spend more time thinking about their responses to solve complex and critical problems (see the OpenAI o1 hub)
Limits: Open AI documentation > Rate limits > Usage tiers
History: OpenAI launched ChatGPT4 in March 2023. GPT-4 is a model based on transformer architecture pre-trained to predict the next token in a document, using both publicly available data (such as internet data) and data licensed from third-party providers. The model was then fine-tuned using Reinforcement Learning from Human Feedback (RLHF). As the technical report states, "given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training computer,dataset, construction, training method, or similar" (OpenAI, 2023). GPT-4o (o for omni) was released in May 2024 allowing you for the first time to input any combination of text, audio, image, and video, generating any combination of text, audio, and image outputs.
Claude.ai
Input: Text, files, images
Notable features: generates safe answers, supports large prompts, accepts text files attached to a prompt.
Use cases: abstract and title creation, scoping.
Cost: Claude.ai is free, Claude Pro requires a subscription.
Limits: The daily message limit varies depending on the demand and the subscription tier (Anthropic, 2024) .
History: Claude is a foundational model created by Anthropic to be "helpful, harmless, and honest" (Askell et al., 2021; Bai et al., 2022). The character of Claude has developed since then. Anthropic's 8 June 2024 press release explains some of the thinking behind the construction of Claude's character and how these traits have been trained in the latest versions of the model.
Claude 3.5 can now interface with computers. It can set up software, move a cursor round the screen following command prompts, click on relevant locations and input information via a virtual keyboard (see Anthropic press release, 22 October 2024). This is state-of-the-art for models that use computers in the same way a person does but still far below human skill levels.
In November 2024, Anthropic announced Amazon's $4 billion investment in the company. It will use Amazon Web Services (AWS) as its primary training partner and will use AWS Trainium and Inferentia chips to train and deploy its largest foundation models (see Amazon press release, 22 November 2024).
Gemini
Input: Text, images, voice.
Notable features: Optimised for summarising, reasoning, coding, and planning. Dataset continuously updated from the Web. Gemini is supplied with all of Google's Pixel phones and can be downloaded on all Pixel devices with 2 GB of RAM or more and Android 10+. The Gemini iPhone app was launched in November 2024. Developers use Google Cloud applications such as Google AI Studio and Vertex AI to build interactive and immersive applications with Gemini as demonstrated in its showcase.
Use cases: abstract and title creation, scoping.
Cost: Gemini is free, Gemini Advanced requires a subscription.
History: "Gemini is Google's most advanced AI model, built from the ground up to be multimodal, which means it can generalize and seamlessly understand, operate across and combine different types of information including text, code, audio, image and video. You will need a personal Google account that you manage on your own to access Gemini" (see Google's 25 May 2023 blog). Previously known as Bard, the Gemini family was launched in December 2023. One year later, Gemini 2.0 was launched with Gemini 2.0 Flash available to developers and Gemini Advanced users having access to Deep Research, a new research assistant tool (see Google's 11 December 2024 blog).