What is Elicit?
Elicit is an AI-powered scientific discovery tool created by Ought (and later spun off independently) aimed at automating systemic literature reviews without the hallucination risks common to general-purpose chatbots like standard GPT-4. Rather than inventing citations, Elicit interacts directly with a vectorized index of over 200 million academic papers sourced from Semantic Scholar and open-access publishers.
Core Capabilities Evaluated
1. Semantic Paper Discovery
Traditional search engines require precise Boolean operators (AND, OR, NOT) and specific keywords. Elicit lets researchers ask plain-language questions such as "What interventions reduce cognitive decline in early-stage Alzheimer’s patients?" It returns relevant papers even if those exact keywords aren't in the title.
2. Structured Data Extraction Tables
One of Elicit’s standout features is its custom column extraction. When reviewing 20 papers on a topic, you can configure columns to automatically extract:
- Methodology / Sample Size: How many human participants took part?
- Outcomes Measured: Primary and secondary endpoints.
- Key Limitations: Self-reported weaknesses identified in the discussion section.
Strengths vs Limitations
Where Elicit Excels:
- Zero Citation Hallucinations: It only reports on published, indexed DOIs.
- High-Speed Synthesis: Reduces literature mapping time from weeks to hours.
- Custom PDF Uploads: You can upload your own library of paywalled PDFs for private vector analysis.
Areas for Improvement:
- Paywalled full-text papers without open access abstracts can only be partially analyzed unless manually uploaded.
- Credit usage can scale quickly during high-volume data extractions across 100+ documents.
Final Verdict
For thesis writers, principal investigators, and university students embarking on capstone research, Elicit represents the highest standard of ethical AI implementation. It does not write the paper for you; rather, it supercharges your ability to find and critique relevant academic literature.