Elicit

Elicit is an AI research assistant built specifically around the systematic review workflow rather than around chat. You give it a research question, it searches a large corpus of papers, and it returns a structured table in which each row is a study and each column is something you asked it to extract — sample size, intervention, outcome measure, effect direction, limitations.

What it does

The table is the product. Instead of a summary paragraph you cannot audit, Elicit gives you extracted values with the source passage attached, so every cell can be checked against the paper it came from. It supports screening against inclusion and exclusion criteria, deduplication, and export to the formats a review actually needs. Reported time savings are substantial — Elicit has been credited with reducing systematic review timelines from weeks to days.

Where it fits

Elicit is the right tool when the question is well defined and the work is volume: several hundred abstracts to screen, a consistent set of variables to pull from each included study. It is a poor fit for open-ended exploration of an unfamiliar field, where a citation-graph tool such as Research Rabbit or a broad evidence engine such as Consensus will serve you better.

Strengths

• Structured extraction into auditable columns, not opaque prose summaries.
• Source passages attached to extracted values, so claims can be verified.
• Genuine time savings on screening at scale.
• Purpose-built for systematic review rather than retrofitted from a chatbot.
• Usable free tier for evaluating it on a real question.

Limitations

Extraction accuracy varies with how clearly the underlying paper reports a value, and it degrades on complex study designs — every extracted cell still needs human verification, and treating the output as final is the failure mode to avoid. Corpus coverage is strong but not exhaustive, so it does not replace a database-specific search with properly constructed syntax. Serious use requires a paid plan. It is an accelerant for the screening and extraction stages, not a replacement for methodological judgement.

Verdict

The strongest tool in its category for structured review work, and worth a paid month the next time you face a few hundred abstracts. Verify the cells. The point is that with Elicit you can — which is more than can be said for a chatbot summary.