9 Chapter 6: Searching the Literature Using AI

Knowing how to search for literature effectively is one of the first real research skills you need to develop. We are not talking here about typing a few words into Google Scholar and hoping for the best, we are talking about developing a kind of digital fluency that lets you navigate databases, recognize quality sources, and filter through the overwhelming amount of information out there. The good news is that with a bit of practice and the help of some powerful AI tools you can become surprisingly efficient at this.

Effective searching saves you time, helps you avoid unnecessary duplication, and leads you to the most relevant and influential work in your field. These days, there’s a whole new generation of AI-powered academic search engines that make this process a lot smoother than it used to be. We’ll go over some of the best tools later in this chapter, but before jumping into those, it’s important to understand a few key strategies that apply no matter what tool you’re using.

One thing that really influences your search is how familiar you already are with your topic. If it’s a brand-new area for you, the first step is simply to get your bearings. Creswell (2009) suggests starting with broader review studies, like meta-analyses or comprehensive literature syntheses, that give you a sense of what’s already been done. These types of papers pull together and analyze many individual studies at once offering a wide-angle view of the field. They can help you recognize important trends, theoretical frameworks, and commonly used terminology, all of which are useful for narrowing your focus and refining your search terms.

A great follow-up strategy is what’s often called the snowball method. As you read those initial review articles, pay close attention to their reference lists. When certain papers or authors keep showing up across different sources, that’s a signal they’re likely foundational to the field. Go find those original works and read them yourself. This approach helps you move from secondary summaries to primary sources, an important step if you want your literature review to reflect a deep and direct engagement with the core texts.

Now, here’s where AI can be of tremendous help. Most AI-powered academic search tools offer filters that let you sort results by most cited. That’s a feature you should absolutely take advantage of. Highly cited papers tend to be the ones that shaped the field or sparked major debates, so prioritizing them helps you get to the heart of the conversation faster. Once you find some of these high-impact studies, check out the journals they were published in. A quick search of the journal’s homepage can tell you whether it’s peer-reviewed, an important marker of scholarly credibility. You might also want to glance at the journal’s impact factor, which gives a general sense of how influential it is within the academic community.

As you go through individual studies, start evaluating them with an eye on their relevance and quality. Creswell (2009) recommends focusing on key elements like the research problem addressed, the purpose of the study, the population or sample used, and the main findings. These are not only useful for determining whether a study belongs in your review, but they’ll also form the backbone of the summaries you’ll write later on. Being intentional at this stage will save you a lot of time (and frustration) down the line when you start pulling your review together.

AI Tools for Searching the Literature

The world of generative AI is evolving at an extraordinary pace. Each week seems to bring new tools, features, and breakthroughs many of which directly impact how we search for, engage with, and synthesize academic research. Over the course of writing this book, which has taken nearly a year, we’ve witnessed the release of a wide array of new AI tools.We’ve tested, evaluated, and included only those that we found genuinely useful.

One of the most notable developments during this time has been the emergence of the Deep Research category, a new class of AI tools designed specifically for extensive, complex, and multi-layered research tasks. ChatGPT was the first to introduce this functionality followed by Google’s Gemini, and later other platforms that began offering similar capabilities. As we’ll explore in the next section, Deep Research feature goes beyond basic search, it is designed to help researchers tackle the deeper cognitive demands of scholarly inquiry. That’s why we’ve chosen to begin this chapter with it.

That said, for clarity and organization, we’ve grouped the literature search tools in this chapter into two main categories:

• Deep Research tools which provide structured, AI-assisted responses to complex research questions.
• AI-powered academic search platforms to help with locating, filtering, and managing literature efficiently through intelligent discovery features.

Together, these tools represent some of the most exciting and impactful ways AI is transforming the literature search and review process.

I. AI Deep Research

Since its introduction by ChatGPT, the Deep Research feature has been added to several general-purpose AI platforms. In the section below, we discuss some of the ways Deep Research can help you in your literature search and some of the platforms where you can experiment with it.

1.1. ChatGPT Deep Research

Described by OpenAI (2025) as “an agent that uses reasoning to synthesize large amounts of online information and complete multi-step research tasks for you,” Deep Research is fundamentally different from other GPT models in that it is purpose-built for in-depth, domain-specific inquiry and not real-time multimodal conversations.

Deep Research functions like a digital research assistant. You begin with a well-crafted prompt outlining your topic or research problem. From there, the model takes over searching the web, identifying relevant sources, analyzing content in real time, and synthesizing findings into a coherent, structured report. The output includes in-text citations, a reference list with clickable links, and often visual elements like charts or graphs.

This is the first tool from OpenAI to be powered by the o3 reasoning model, which has been optimized specifically for advanced browsing and analytical reasoning.  What we find particularly impressive is the transparency built into the process. As the model works, a sidebar tracks its reasoning step by step: the sources it’s consulting, how it’s interpreting them, and the logic guiding its synthesis. This ability to “see” the tool’s thinking in action marks an important step toward more explainable AI.

Access to Deep Research is currently limited to ChatGPT Pro, Plus, and Team users. To use it, open your ChatGPT interface, click on “Deep Research” in the message composer, and enter your query. We recommend writing a detailed and focused prompt. Be specific about the type of sources you want (e.g., journal articles, academic books, official reports) and the timeframe or perspective you’re interested in. For example: “I’m researching ethical AI. Please summarize recent academic perspectives from the past five years, focusing on peer-reviewed articles and books. Prioritize scholarly consensus and include links to original sources.”

You can use ChatGPT itself to refine your prompt before submitting it to Deep Research. The tool can sometimes ask clarifying questions when your input lacks specificity which helps guide the process. You can also attach files such as articles, spreadsheets, and background materials to provide additional context. Once you submit your prompt, Deep Research gets to work. A sidebar on the right tracks its progress and reasoning, and within a few minutes, the full report is ready. You’ll receive a notification once it’s complete.

Keep in mind, however, that Deep Research can only access knowledge that is readily available online which limits the sources it draws on and for us in education and academic research we know so much of peer reviewed knowledge is gated and paywalled which AI models, allegedly, cannot access.  This limitation isn’t just a quirk of Deep Research; it actually reflects a broader challenge in the world of generative AI.

Just recently, emboldened by Trump’s “pro-growth AI policies,” OpenAI, along with other leading AI companies, formally asked the administration to authorize the use of copyrighted materials including books and academic content for training their models (Kan, 2025). This request, controversial as it is, is telling; it signals that the current reservoirs of open-access data (such as blog posts, Wikipedia entries, online textbooks) are reaching their limits, or even that ‘data size scaling is over” (Marcus, 2025). For any substantial leap in the quality and intelligence of generative AI, we argue, the models need access to richer, more authoritative training data.

In fact, OpenAI acknowledged this limitation explicitly in their announcement of Deep Research, stating that the tool has problems “distinguishing authoritative information from rumors, and currently shows weakness in confidence calibration, often failing to convey uncertainty accurately.”

And this isn’t the only limitation of Deep Research. OpenAI also highlighted several other issues, by now characteristic of generative AI technologies, including hallucinations, flawed reasoning, and incorrect inferences. We explore these problems in greater detail in the chapter on AI ethics.

That said, we have to acknowledge that Deep Research represents a significant advancement in the use of AI for academic research. As far as automating literature searches goes, it’s state-of-the-art. What would normally take hours, sometimes days, of manual sifting through databases, Deep Research can do in minutes. You simply enter your research question or topic using a well-crafted prompt, and the tool returns a structured, coherent summary of part of the existing literature. While the references it provides still need to be carefully verified and supplemented, this output can serve as a solid starting point for your own review. It can certainly point you in the right direction, helping you get your bearings and identify key sources more efficiently.

1.2. Gemini Deep Research

Following ChatGPT Deep Research, Google rolled out its own version of AI-powered research assistance: Gemini Deep Researc. Like ChatGPT’s, Gemini Deep Research helps users tackle complex research queries with AI support but it also offers extra features not found in ChatGPT’s version. The process Gemini Deep Research works is simple: you input your research prompt and Gemini takes over. It first lays out a detailed research plan showing you the steps it plans to take, the type of sources it will consult, and how it intends to approach your query. This plan is fully editable: you can remove steps, add new ones, or refine the focus to suit your needs. This level of customization is definitely an important feature and something ChatGPT currently doesn’t offer.

Once you hit “Start research,” Gemini begins browsing the web for information. You can view a live list of the sources it draws on, which adds a welcome layer of transparency. That said, when it comes to reasoning, we find ChatGPT’s step-by-step explanations more comprehensive and thoughtful than Gemini’s, especially for academic research tasks that demand nuanced logic and interpretation.

When the report is ready, Gemini presents it in a clean, readable format, complete with hyperlinked sources so you can click through and explore them directly. One unique feature here is the audio overview, Gemini can summarize its findings in spoken form, which is especially handy if you’re multitasking or prefer auditory learning. Another practical feature is the seamless integration with Google Docs: you can open the full report with a click and continue working on it right in your workspace. No surprise there, Gemini is, after all, part of Google’s ecosystem, and the experience is clearly optimized for users already embedded in that environment.

As of now, Gemini offers two levels of access. Free users can try Deep Research a limited number of times per month, which is great for occasional or light use. Gemini Advanced subscribers, on the other hand, get expanded access to Deep Research and other advanced features.

1.3. Perplexity Deep Research

Perplexity AI has also rolled out its own Deep Research feature, joining the ranks of ChatGPT and Gemini in the growing space of AI-powered research assistants. Functionally, it closely resembles ChatGPT Deep Research, especially in how it structures its responses and walks users through its reasoning. When you enter a research query, Perplexity generates a step-by-step research plan, showing the logic behind its approach and providing a live feed of the sources it draws on as it builds your report. This level of transparency is useful and aligns with the growing push for more explainable AI.

However, unlike Gemini Deep Research, Perplexity currently doesn’t offer the ability to customize or edit the initial research plan. That means you’re limited to the default structure the tool provides, without the option to refine its focus, remove unnecessary steps, or add new angles to explore. This can be a drawback, especially for more advanced users who want greater control over how their query is interpreted and handled.

Like its competitors, Perplexity scours a wide range of publicly available online sources (e.g., blogs, open-access articles, reference websites, and more) to synthesize its findings into a cohesive narrative. The final report includes clickable links to the cited sources, allowing you to verify the information or dig deeper into particular aspects of the topic. The interface is clean and easy to navigate, and the overall experience feels smooth and responsive.

Whether you’re using ChatGPT Deep Research, Gemini Deep Research, or Perplexity, these tools can be incredibly helpful when you’re just starting out on your literature search. They offer quick overviews, surfacing open-access sources, and helping to organize your initial thoughts. They’re particularly useful for scoping out a topic, identifying emerging themes, and getting a general sense of the existing discourse.

That said, these tools should be seen as starting points, not endpoints. Because they rely heavily on publicly available content, they often miss the depth, nuance, and scholarly rigor found in peer-reviewed journals, academic books, and discipline-specific databases. For any serious academic work especially literature reviews you’ll need to move beyond these AI-generated summaries and engage directly with the foundational sources in your field.

II. AI-Powered Academic Search Platforms

The platforms we share in this section enhance the traditional literature search process using AI to recommend, organize, and filter academic papers. They help you find relevant sources faster, track citations, and manage growing bodies of literature with greater efficiency.

Although we haven’t included them in this section, it’s worth noting that tools like Google Scholar, Semantic Scholar, and traditional academic databases such as ERIC, ProQuest, and even Google itself still play an important role in academic research. They remain valuable resources, especially for accessing peer-reviewed articles, institutional repositories, and grey literature. That said, the tools featured here stand out for offering smarter, more time-efficient, and highly targeted search experiences, powered by AI. Besides offering help with locating relevant studies, they also assist with summarization, synthesis, and concept exploration.

2.1 Scite

Scite is an AI-powered research platform that can help enhance the way you search for, evaluate, and manage scholarly literature. Scite is designed to help you interact meaningfully with academic content by focusing on the context of citations showing not just where a paper is cited, but how it’s cited (supporting, contrasting, or mentioning) which adds an important layer of nuance to your literature review.

Here is how it works: First, you input a research question or a general topic and Scite retrieves insightful excerpts from peer-reviewed papers, with direct links to the original sources for verification. This is incredibly useful when you’re trying to ground your literature review in evidence-based studies. You can explore citation networks, review how studies have evolved over time, and follow both the references within a paper and the “cited by” trail to discover additional, relevant literature making it easier to map out your field thoroughly.

To make research even more accessible, Scite offers a Chrome extension that allows you to verify online content instantly. Just highlight a statement, right-click, and choose “Ask scite.ai Assistant” to see related academic sources that confirm or challenge the information, an excellent way to quickly back up claims or explore deeper context while reading on the web.

Scite has also introduced its own version of Deep Research, aptly named Assistant. This AI-powered tool is designed to enhance your entire research workflow by allowing you to ask complex questions and receive detailed, reference-supported answers. What makes Assistant especially powerful is the degree of customization and control it offers. You can:

• Choose whether references are required in responses or let Assistant decide automatically.
• Specify filters like year range, publication type, journal name, or topic to tailor search results to your needs.
• Limit Assistant to draw only from papers in a dashboard collection you’ve curated or from specific journals you trust.
• Control the length and depth of responses to suit your writing or planning stage.

This level of flexibility allows you to use Assistant for both quick insights and deep, targeted literature search. You remain in the driver’s seat, with the AI working as a powerful co-pilot helping you generate ideas, summarize complex topics, and locate studies that truly align with your research goals.

2.2. Consensus

Consensus is another powerful AI-driven academic search engine to help you quickly locate and synthesize relevant research papers. It draws on the Semantic Scholar database, which includes over 200 million academic papers, to deliver clear, evidence-based answers to your research questions.

If you’re just getting started with your research, you can simply type your research problem or topic as a question, and Consensus will generate a structured, multi-source report that synthesizes findings from the literature. Each result appears in a tile format that includes a summary insight and is linked to a full page with detailed information about the cited source. You can hover over the citations to preview the source, and click through to explore further viewing the paper’s title, author, abstract, citation list, references, and, if available, direct links to the full text or its Semantic Scholar page.

One practical feature is the ‘Study Snapshot’ located at the bottom of each tile. This AI-generated summary gives you a quick overview of the study’s structure and design, highlighting details like population, sample size, methods used, outcomes measured, and the key takeaways. It’s a time-saver for quickly assessing whether a study aligns with your research focus without reading the entire paper upfront.

For those who want to dig deeper, clicking on the paper’s title opens a more detailed view, complete with quality indicators, such as journal impact scores, study type, citation metrics, and links to referenced literature. This helps you evaluate the credibility and academic value of the paper in context.

Consensus also provides a wide range of filters to help you narrow your results efficiently. You can refine your search by publication year, country where the study was conducted, journal source, study type (e.g., meta-analysis, literature review, case report), sample size, methodology, open access status, and more. These filters are especially helpful when you’re trying to target high-quality or methodologically specific studies.

Another core strength of Consensus is its emphasis on paper-level insight extraction. Instead of just listing articles, it pulls out key findings and links them directly to their sources, allowing you to follow up on leads, expand your reading list, and build your review with greater efficiency.

And for those who use ChatGPT, Consensus has gone a step further by offering a custom GPT integration. You can add Consensus GPT directly into your ChatGPT interface and chat with scientific literature without ever leaving the platform. It’s a seamless way to incorporate AI-assisted research into your workflow.

2. 3. Elicit

Elicit is an AI-powered academic search engine built to support and streamline your literature review process. Similar to Consensus, Elicit begins by asking you to frame your query as a research question. It then searches a wide range of academic sources and generates a structured report that includes both a summary of key insights and a table with topically relevant papers.

Each paper in the table is accompanied by essential information including the title, authors, journal name, citation count, a brief summary of the abstract, and, when available, a direct external link to access the full text. This layout makes it quick and easy to evaluate the relevance, quality, and credibility of the sources at a glance.

To help you narrow and refine your search, Elicit offers a robust set of integrated filters. You can sort results by most cited, publication year, study type, PDF availability, and even journal quality indicators. One particularly helpful feature is the ‘Abstract Keywords’ filter which limits results to papers whose abstracts contain specific terms you provide, great for targeting studies closely aligned with your research focus. For even more customization, Elicit allows you to filter using various predefined columns or create custom columns based on keywords, topics, or variables that matter most to your project.

2. 4. R Discovery

R Discovery is a versatile and user-friendly AI tool that can significantly streamline your literature search. It offers a powerful yet intuitive way to discover relevant academic materials for your literature review. You can search for papers using a variety of filters, including year of publication, journal name, field of study, topic, language, and more.

When you click on a paper, you’re taken to a detailed view where you can read the abstract, check for availability (such as open access or full-text PDF), view the number of citations, and access links to external sources. This quick overview helps you efficiently assess whether the study is relevant and worth reading in full.

One of the most helpful features on this page is ‘Similar Papers’, an AI-powered tool that suggests other studies related to the one you’re viewing. It’s an excellent way to expand your reading list with relevant sources you may not have discovered on your own. For literature reviews, where breadth and depth are essential, this feature can save you a significant amount of time and introduce you to foundational or complementary works.

Another feature we truly appreciate is the personalized feed on R Discovery’s homepage. When you first create an account, the platform prompts you to set your preferences by selecting your research area, followed by specific topics, journals, or keywords. Based on this input, R Discovery generates a custom feed of research papers that updates regularly to reflect your interests. The feed is divided into two tabs:

• Top Papers: Highlights the most cited and impactful studies in your field, these are often key sources you’ll want to include in your literature review.
• Recent Papers: Displays newly published research, helping you stay current with the latest developments in your area.

Both tabs are invaluable for building a literature review that is not only comprehensive but also timely. R Discovery also includes thoughtful accessibility features. You can listen to an audio version of a paper, which is helpful for multitasking or auditory learners, and you can translate content into your preferred language.

R Discovery is also available as a custom GPT for ChatGPT users, meaning, you can interact with academic literature directly within the ChatGPT interface. Additionally, R Discovery offers a Chrome extension that brings research discovery tools right into your browser for even faster access to relevant papers.

2.5. Connected Papers

Connected Papers is another research tool to help you visually explore the landscape of academic literature related to a specific topic. It’s particularly useful when you’re trying to get an overview of a research field or trace how ideas are connected across studies. With just a few clicks, you can generate a visual graph that maps out relationships between papers revealing both well-known works and potentially overlooked but relevant studies.

The process is simple: enter a paper title, DOI, or URL, then click “Build a graph.” Connected Papers will generate a dynamic graph in which each node represents a paper, and the connections between them are based on similarity in citations and references. As noted by the platform, papers with highly overlapping bibliographies are assumed to be closely related in subject matter and are placed closer together on the graph. Less similar papers appear farther apart, offering a clear visual representation of the conceptual proximity among them.

This format provides a powerful way to uncover related literature you might not find through conventional keyword searches. It also helps you see how different studies cluster around specific themes or theoretical frameworks.

Connected Papers also includes two especially valuable features:

• Prior Works: Helps you discover foundational papers that significantly influenced the topic or area of research.
• Derivative Works: Surfaces more recent studies or literature reviews that build on or are influenced by the paper you started with.

You can easily toggle between graph and list views, depending on your preference. The list view offers a more traditional layout for browsing related papers, while the graph view gives you an intuitive and interactive visual map. When you click on a paper within the graph, a summary appears in the right-hand pane, providing an overview of the study. From there, you can build a new graph based on that paper, download or share the current graph, and open the paper on Google Scholar or Semantic Scholar for further reading.

2.6. ResearchRabbit

ResearchRabbit is an AI-powered research discovery platform that enhances your literature search experience through interactive visual graphs. Similar to Connected Papers, ResearchRabbit helps you explore how studies are connected, uncovering patterns, clusters, and emerging themes within your field. Its strength lies in its ability to continuously refine recommendations based on your interests thus turning your search into a dynamic, evolving process.

To get started, you begin by adding papers that are relevant to your research. You can do this by searching for a paper using its title, DOI, or keywords, or by uploading papers directly from your computer or reference library. ResearchRabbit uses this initial input to generate a network of recommended papers, which you can explore and add to your project with a single click. As you interact with the platform, selecting relevant papers or dismissing others, ResearchRabbit creates a feedback loop, refining its suggestions to align more closely with your research focus.

Real-time updates is another interesting feature we find helpful. Once you’ve created and saved a project, ResearchRabbit will notify you whenever new papers related to your topic are published helping you stay current without the need for constant manual searching.

In addition to literature discovery, ResearchRabbit includes several powerful tools that support deeper exploration. The ‘Similar Work’ feature features collections of papers closely related to those in your project allowing you to expand your literature base with minimal effort. The ‘Explore People’ tool lets you dive into the research output of specific authors, see who has cited them, and trace the development of their scholarly impact. The ‘Suggested Authors’ tab introduces researchers whose work overlaps with your topic, helpful when building a broader understanding of the academic community surrounding your field.

2.7. Litmaps

Much like ResearchRabbit and Connected Papers, Litmaps incorporates a dynamic visual component, enabling you to see the relationships between research papers at a glance. This bird’s-eye view allows you to uncover both well-established connections and unexpected links between studies helping you think more broadly and critically about your topic.

Litmaps works by building a citation network using shared references and citations to map out how papers are interconnected. The size and density of this network will vary depending on your topic and the article you begin with.

Getting started is simple. You can either search for a paper within the platform using the built-in search bar or upload a paper directly from your computer or reference library. Once you select a starting article, Litmaps quickly generates a literature map, a visual graph that displays the key papers related to your chosen entry point. Each paper is represented by a node (circle), and you can hover over a node to see the title or click on it to view more detailed information.

Clicking on a paper gives you access to its abstract (if available), along with a visual breakdown of its top 10 references, citations, and adjacent articles, those that are closely connected within the citation network. From here, you can easily add new papers to your map, creating a growing, interactive representation of the literature around your topic.

The ‘Explore’ pane on the left side of the screen features AI-powered recommendations based on your selected paper. You can further refine your results by clicking ‘More Like This’ on articles that align with your research interests. This helps Litmaps understand your focus and tailor its suggestions accordingly.

Conclusion

As we’ve seen throughout this chapter, generative AI has dramatically transformed the way we approach literature search. What once felt like a tedious and time-consuming process has become faster, smarter, and more accessible than ever. You no longer need advanced technical skills or hours of training to navigate complex databases. With just a clear, well-defined research question or topic, you can begin searching, and in a matter of minutes, these tools can surface relevant studies, highlight major themes, point out research gaps, and even help you organize your findings.

We honestly could never have imagined, back when we were doing ours PhDs, that one day a graduate student would have access to a 24/7 research assistant, one that can suggest recent, high-impact papers, identify trends in the field, and recommend where to go next. It’s an incredible advancement, and generative AI deserves credit for opening up this kind of possibility.

That said, this progress also comes with a word of caution. These tools are powerful, but they are not a replacement for the intellectual work required in research. They can point you in the right direction, but you still have to do the work: reading, analyzing, questioning, and understanding the literature. The real insight, the meaningful contributions, come from your own engagement with the material, which is the topic of our next chapter.