AI tools for literature review: free vs paid (2026)

AI tools for literature review are apps that help you find, map, summarise and organise research papers. Semantic Scholar, Google Scholar and Zotero are fully free. ResearchRabbit, Connected Papers, Elicit, Consensus and SciSpace have free plans with limits, and paid plans for heavy use. No single tool is best for every job.

You will see what each tool is good at, what you get for free, and how to combine them in 5 simple steps. We checked every price and limit on the official pages on 12 November 2026. Prices change often, so check again before you pay.

What does an AI tool for literature review do?

An AI tool for literature review does one of four jobs. You need one tool for each job, not ten tools for the same job.

• Find: search millions of papers by topic. Examples are Semantic Scholar and Google Scholar.
• Map: start from papers you like and show related papers. Examples are ResearchRabbit and Connected Papers.
• Read: summarise papers, answer questions and pull out key facts. Examples are Elicit, Consensus and SciSpace.
• Store and cite: save papers and build your reference list. The main free example is Zotero.

A general chatbot like ChatGPT or Gemini is not a paper search engine. The University of Illinois Library warns that these tools can create titles and authors that do not exist. The tools in this guide search real paper databases, so each result points to a real paper. Their summaries can still be wrong.

Which AI tools for literature review are free in 2026?

Semantic Scholar, Google Scholar, arXiv and Zotero are completely free. The other tools have a free plan with limits. Here is what each one gives you, as of 12 November 2026:

• Semantic Scholar: free, with no paid plan.
• Google Scholar: free, with no paid plan.
• arXiv: free to read and free to post.
• Zotero: free. You can pay for extra online storage if you need it.
• ResearchRabbit: free plan with unlimited searches and up to 50 seed papers. ResearchRabbit+ costs $12.50 a month, or $10 a month on a yearly plan.
• Connected Papers: free plan with a monthly limit on new graphs. The limit was 5 graphs a month when paid plans started in 2022. We could not open the current pricing page, so check the number yourself.
• Elicit: free Basic plan with unlimited search, summaries and chat with papers, but limited use of the Research Agent and Research Reports. Pro costs $49 a month, billed yearly ($588).
• Consensus: free plan with 10 Pro messages and 3 Deep Searches a month. Pro costs $20 a month, or $12 a month billed yearly.
• SciSpace: free plan with limits. We could not open its pricing page, so check the current limits yourself.

Paid prices are in US dollars. To pay less, note two offers. Consensus gives students 40% off Pro for 12 months with a verified student email. ResearchRabbit says it has discount codes for many countries.

Many papers these tools find are behind a paywall. In India, ask your library if your college has One Nation One Subscription (ONOS) access. ONOS is a Government of India scheme that gives 13,000+ full-text journals from 30+ publishers to colleges and universities run by the central and state governments.

Which free AI tools are best for finding papers?

Semantic Scholar and Google Scholar are the two best free places to start finding papers. Use both, because their results are not the same.

• Semantic Scholar: free search run by Ai2, a non-profit AI research lab. It indexes over 200 million papers. For nearly 60 million papers in computer science, biology and medicine, it shows a one-sentence "TLDR" summary (still in beta).
• Google Scholar: free and very wide. Under each result, "Cited by" shows newer papers that cite it, and "Related articles" shows similar ones. You can also set an alert for new papers on a topic.
• arXiv: free preprints in AI, maths, physics, computer science and more, with over 3 million papers. arXiv does not peer review these papers, so check if a later version was published at a journal or conference.

A survey paper is the fastest way into a new topic. Search "<your topic> survey" on Semantic Scholar or Google Scholar, and read one or two recent ones first.

ResearchRabbit vs Connected Papers: which is better for related papers?

ResearchRabbit is better when you already have several good papers. Connected Papers is better when you start from one key paper. Both show papers as a map, so you can see which ones belong together.

• ResearchRabbit: you add a few "seed" papers. It suggests related papers, you add the good ones, and you repeat. The free plan allows up to 50 seed papers. ResearchRabbit+ allows 300.
• Connected Papers: you give it one paper, and it draws a graph of a few dozen similar papers. Papers sit close together when they cite the same papers or are cited by the same papers, even if they never cite each other.

A student literature review rarely needs more than 50 seed papers, so the free ResearchRabbit plan is usually enough. Litmaps, another map tool, bought ResearchRabbit in May 2025, so you may see both names together.

Elicit vs Consensus: what is the difference?

Elicit is built for systematic reviews: it screens papers and pulls data from them into tables. Consensus is built for quick answers to research questions, especially yes/no questions. Use Elicit to compare many papers in detail. Use Consensus for a fast first look at what studies say.

• Elicit: searches papers, writes summaries and lets you chat with a paper. It can screen papers, extract data into tables and write research reports. It can also import papers from Zotero.
• Consensus: gives a short answer with links to the papers behind it. For a yes/no question, its "Consensus Meter" labels the relevant papers in the top 20 results as yes, no or possibly, and shows the split.

Consensus is open about the meter's limits. When it launched the meter, Consensus said the model will label some results wrongly, and estimated this at about 10%. It also said each paper counts the same, so a large meta-analysis and a single case report get equal weight. Read the papers before you trust the meter.

SciSpace is a third option. It lets you ask questions about a paper's PDF and has an AI research agent. Try the free plan of each and keep the one that fits how you work.

What is the best free reference manager for students?

Zotero is a free, open-source reference manager, and a good default for students. It is made by an independent non-profit, and it is at version 10 as of November 2026.

Zotero saves a paper's details straight from its web page, so you do not type them by hand. It then adds citations and a reference list in Word, LibreOffice or Google Docs. Zotero is not an AI tool, but it matters most for safety. When every reference comes from Zotero, the details come from the real paper, not from a chatbot.

How to use AI for literature review: a 5-step workflow

This workflow uses only free plans. It works for a course project, a final-year project or a first research paper.

1. 1Write your question: one clear sentence, plus 3 to 5 search words. A small question is easier to review than a big one.
2. 2Find the core papers: search Semantic Scholar and Google Scholar. Start with 1 or 2 recent survey papers. Save the 5 to 10 best papers to Zotero.
3. 3Expand the map: add those papers as seeds in ResearchRabbit, or put the most important one into Connected Papers. In Google Scholar, open "Cited by" to find newer work. Save only papers that match your question.
4. 4Screen and read: use Elicit or Consensus for a first view of what each paper found. Then read the abstract, figures and conclusion yourself, using the three-pass method. Keep a table with one row per paper, with its question, method, data, main result and limits.
5. 5Write and check: write the review in your own words, grouped by theme, not paper by paper. Cite from Zotero. Then check every reference before you submit.

How to use AI tools for literature review safely

Use AI tools to find and sort papers, but never let them decide what a paper says. You are responsible for every reference and every claim with your name on it.

• Check each reference: search the exact title on Google Scholar or Semantic Scholar, check the authors and year, and open the DOI link. If you cannot find the paper, do not cite it.
• Read before you cite: an AI summary can miss a detail or get a result wrong. Read at least the abstract and the part you cite.
• Follow the rules: read your college's AI rules and the AI policy of the journal or conference you send to. The Committee on Publication Ethics (COPE) says an AI tool cannot be an author. Authors must say how they used AI, and they are responsible for every part of the paper, even the parts AI wrote.
• Ask your guide: if the rules are not clear, ask your supervisor before you use AI, not after.
• Keep data private: do not paste unpublished data, or a draft someone shared with you in confidence, into an AI tool.

In 2026, fake references were found even in papers at top AI conferences. Our post on fake citations in AI research papers shows what happened and gives a full checklist.

What do students get wrong with AI research tools?

The most common mistake is trusting a summary instead of reading the paper. Students new to these tools often make a few other mistakes too:

• Too many tools: one tool for each job is enough. Learn three tools well.
• Paying too early: the free plans above are enough for most student reviews. Pay only when you hit a limit you really need.
• Only one database: Google Scholar and Semantic Scholar give different results. Search both.
• A list, not a review: a literature review compares papers and shows what is still unknown. A list of summaries is not a review.

At APJ-RAISE, our free course on reading and reviewing papers teaches you to find, read and review papers step by step, with other students. Joining is free.