Elicit workflow: a literature map in one hour

What it does, and what it costs

Basic (free): paper search, summaries and unlimited chat with the full text
Pro: $49/mo ($588 annual) — standard usage
Scale: $229/mo ($2,748 annual) — five times Pro's allowance
Enterprise: contact sales

Search and summarisation being free matters. What you pay for is systematic-review-scale screening of thousands of papers, high-volume table extraction and API access. For an individual doing background research, the free tier is often enough.

An hour to the shape of the field

1. Write the question as one sentence (10 min)

'The effect of AI tools' is too broad. 'Does AI coding assistance increase productivity in software development?' names a population, an intervention and an outcome. Time spent here is repaid in every later step.

2. Search and skim the top results (10 min)

Put the question in as written. Skim twenty titles and abstracts and collect the vocabulary being used. Academic fields use terms that differ from everyday language, and this is where you find the right search terms.

3. Extract into a table (20 min)

Use table extraction to pull sample size, intervention, measured outcome and result as columns. With ten to twenty papers side by side, how the study designs differ and whether the conclusions agree becomes visible at a glance.

4. Read three originals (20 min)

From the table, pick the three closest to your question and read them properly. This step cannot be skipped. Summaries carry the conclusion and drop the conditions — 'productivity improved' can rest on thirty students implementing a simple algorithm.

What to delegate, and what not to

Step	Delegate?	Why
Suggesting search terms	Yes	Valuable precisely when you do not know the field's vocabulary
Screening large result sets	Yes	Title and abstract are enough to judge mechanically
Getting the gist of each paper	Yes	Sufficient for deciding what to read
Quoting a conclusion	No	Quoting a summary that dropped the conditions produces a false claim
Assessing a study's limitations	No	Requires reading the methods section
Supporting your own argument	No	Citing without reading the original is research misconduct

If you cite it, read it. That does not change because the tool is an AI.

Two traps in practical use

The dangerous one is reading only the summary and then writing 'research shows'. Summaries shorten by discarding conditions, and if those conditions differ from your situation, the conclusion does not transfer.

The other is absence. Journals not indexed, research published in languages other than English, and internal corporate studies do not appear in the results. On a topic rooted in one country's practice, English-language papers alone will give you a distorted picture.

Frequently asked questions

Does it cover non-English research?

Coverage is predominantly English-language academic literature. For Japanese research, use CiNii or J-STAGE alongside it — on Japan-specific topics, skipping that step means missing the important work.

Is the free plan enough for a thesis?

For searching and getting oriented, yes. Every paper you cite still has to be read in the original. Citing on the strength of a summary is inappropriate regardless of the tool.

Elicit in an hour — mapping a literature before you read it