Elicit

Elicit is what most people imagine when they hear "AI research assistant" — except it actually earns the label. Built originally by the nonprofit Ought and now a standalone company, it searches 125+ million papers and, crucially, extracts structured data from them into comparison tables. It is aimed squarely at researchers doing literature reviews, and that focus shows everywhere in the product.

What we like

The extraction tables are the killer feature and nothing else on the market does them as well. Ask a question, get thirty relevant papers, then add columns — sample size, intervention, effect size, population — and Elicit fills them in from each paper's full text, with links back to the source passage. For a systematic review's screening and extraction phases, this collapses days of work into an afternoon. Because every claim is grounded in a retrieved paper, the fabricated-citation problem that plagues general chatbots mostly disappears. The systematic review workflow added in recent years, with inclusion/exclusion screening, makes it viable for real published reviews, not just exploratory reading.

Where it falls short

The credit system is the main friction: heavy extraction burns through lower-tier allowances quickly, and the jump to Pro pricing stings for students. Extraction accuracy is good but not trustworthy enough to skip verification — expect to correct a field here and there, especially from papers with unusual formatting. And Elicit is clearly optimized for empirical research; if your field is theoretical, legal, or humanities-based, the structured-extraction premise fits poorly.

Verdict

If you do evidence synthesis — in medicine, psychology, economics, or any empirical field — Elicit is the single most useful AI research tool you can buy, and the ~$12/month Plus tier is an easy recommendation. Casual readers who just want quick answers should try Consensus first; it is cheaper for that job. But for serious review work, Elicit is the benchmark.