AI for Researchers
Academic work has the best-developed AI tooling of any profession on this site, because the corpus is public and the tasks are well defined. Literature discovery and screening are transformed. Analysis is faster. Writing help is everywhere and is where journal policy, and the fabricated-citation problem, need your attention.
Literature discovery and screening
The part that used to take weeks. These search semantically rather than by keyword, which finds the paper you needed and did not know how to name.
- ElicitFREEMIUM~$12/mo
Extracts trial and study characteristics into a comparable table — the single biggest saving in a systematic review workflow.
- ConsensusFREEMIUM~$12/mo
Answers a research question directly from published papers and shows how many support each direction.
- Semantic ScholarFREE$0
Free, enormous, and the citation graph is the fastest way to see whether a finding held up.
- Connected PapersFREEMIUM~$5/mo
Visual map of what surrounds a paper. Finds the adjacent literature a keyword search will never surface.
- ResearchRabbitFREEMIUM~$10/mo
Builds outward from papers you already trust and keeps alerting you as the field moves, on a free tier that covers a real review.
Reading, notes and evaluation
Getting through the pile, and judging what is in it. The good tools here point at the page they took a claim from.
- SciSpaceFREEMIUM~$12/mo
Explains dense passages in context, which makes reading outside your subfield viable rather than exhausting.
- NotebookLMFREE$0
Upload your corpus and question those papers only, with citations to the page. The most reliable way to interrogate a reading set.
- SciteFREE TRIAL~$20/mo
Shows whether later work supported or contradicted a claim, rather than just that it was cited.
- PerplexityFREEMIUM~$20/mo
Quick orientation on an unfamiliar area before you know the right search terms.
Data and analysis
Analysis without a week of boilerplate. All of these show the code, which is the only version that belongs in reproducible work.
- Julius AIFREEMIUM~$20/mo
Ask questions of a dataset in English; it writes and runs the code and shows both, so the method stays inspectable.
- DeepnoteFREEMIUM~$39/mo
Collaborative notebooks with AI assistance — the natural home for analysis a supervisor or co-author will read.
- HexFREEMIUM~$36/mo
Notebooks and reporting for analysis that needs to be reproducible and shared with a team.
- PowerdrillFREEMIUM~$20/mo
Fast exploratory passes over several files when you are still working out what the data contains.
Writing, editing and submission
Drafting help with a policy question attached. Check the target journal's disclosure rules before you use any of it on a manuscript.
- ClaudeFREEMIUM~$20/mo
Best of the general models at long structured argument — restructuring a discussion section, tightening a methods paragraph.
- GrammarlyFREEMIUM~$12/mo
Consistent language correction, and the highest-value tool here if English is not your first language.
- QuillBotFREEMIUM~$10/mo
Rephrasing your own sentences when they will not sit right. Note the plagiarism point in the warning below.
- NotebookLMFREE$0
Also earns its place here: ask what your own sources say while writing, instead of hunting through PDFs.
General models fabricate references that look correct — real journal, plausible authors, DOI that resolves to something else. Every citation must be opened and checked, and the tools that cite by design (Elicit, Consensus, Semantic Scholar, NotebookLM) are safer for exactly this reason. Paraphrasing tools carry a second risk: rewording someone else's argument without attribution is still plagiarism, and a paraphraser makes it easy to do accidentally. Third, disclosure — most journals and funders now have explicit rules on AI use in manuscripts and in peer review, and they differ. Read the target venue's policy before submission, not after a query. Using AI on a manuscript you are reviewing for someone else is prohibited nearly everywhere.
How to pick AI tools for research
Prioritise tools that cite. In academic work an answer without a traceable source is worthless, and the split between tools that retrieve from a real corpus and tools that generate from training data is the most important distinction on this page. Elicit, Consensus, Semantic Scholar and NotebookLM all show you where a claim came from; a general chatbot does not, and will produce a convincing reference that does not exist.
Then match the tool to the stage. Discovery wants semantic search and citation graphs — Semantic Scholar, which is free, and Connected Papers, whose free tier is enough. Screening wants extraction into a table, which is Elicit's whole purpose. Close reading wants a document-scoped tool over your own uploaded corpus. Analysis wants something that writes visible code. Using one general assistant for all four is the common mistake and produces the worst version of each.
Check your venue's policy early. Journal and funder rules on AI in manuscripts have tightened and vary, and the rules on using AI while peer reviewing someone else's work are close to universal prohibition. Knowing the policy before you draft is much easier than unpicking it at revision.
Prices on this page are per month at single-user rates. The remarkable thing about this page is how little you need to spend: Semantic Scholar and NotebookLM cost nothing outright, and Connected Papers, ResearchRabbit, Elicit and Consensus all have free tiers that cover a real review.
AI for researchers: common questions
Not sure which of these fits your work?
Describe the job in one sentence. You get three matches back, with reasons and real prices.
Find my AIAI for other professions
Looking for tools by what they do rather than who uses them? Start from the categories, search the full catalog with price filters, or describe the job on AI Match and get three suggestions with reasons.