| ResearchRabbit | Literature search | Students and researchers mapping a literature. | Free search is capped at 50 starting articles | Yes | Free tier: yes. RR+ $12.50/mo or $120/yr. (checked Oct 11, 2026) |
| Semantic Scholar | Literature search | CS and biomed researchers who want free discovery and an API. | Coverage is weaker in humanities, social sciences, non-English work and books | Yes | Free (search and API). (checked Oct 11, 2026) |
| Consensus | Literature search | Students and researchers checking what the evidence says. | Free gets just 10 Pro messages and 3 Deep reviews | Yes | Free tier: yes. Pro $20/mo or $144/yr. Deep $65/mo or $540/yr. (checked Oct 10, 2026) |
| Elicit | Literature search | Researchers doing preliminary literature scans. | One published study found search sensitivity of ~33–40% vs traditional searching | Yes | Free tier: yes. Plus $11/user/mo (annual), Pro $39/user/mo (annual). (checked Oct 10, 2026) |
| Litmaps | Literature search | Researchers tracking a field over time. | Free is capped at 20 inputs, 2 maps and 100 articles per map | Yes | Free tier: yes. Pro from $10/mo (education, annual). (checked Oct 11, 2026) |
| Scite | Checking citations | Researchers checking how a paper has been cited. | An independent evaluation found low recall on supporting and contrasting labels. Treat them as triage, not verdicts | No | Free tier: no (7-day trial). $20/mo or $144/yr. (checked Oct 11, 2026) |
| NotebookLM (Gemini Notebook) | Reading and synthesis | Students and researchers digesting a pile of documents. | Usage is compute-based, refreshing every 5 hours up to a weekly limit | Yes | Free tier: yes. Higher limits via Google AI Pro ($19.99/mo). (checked Oct 10, 2026) |
| Paperpile | References | Researchers who write in Google Docs. | No free plan, just a 30-day trial | No | Free tier: no (30-day trial). Academic Regular $4.15/mo, Expert $5.75/mo (annual). (checked Oct 11, 2026) |
| arXiv | Preprints | Following cutting-edge AI research at the source. | Most papers aren't peer-reviewed, so results can be wrong or overstated | Yes | Free. (checked Oct 11, 2026) |