Launch bonus Yearly plans 20% 40% off

Literature Review

A literature review needs three things from your reading: what the papers say, where they agree, and where they do not. This guide builds all three into one mindmap, with every claim pinned to the page it came from, so you can write the review from the map instead of from memory.

The screenshots come from a real run on a library of ten open-access papers about social media use and adolescent mental health. Every step uses the default model, so it works the same on the Free plan.

Before you start

  • Put the papers in one library. The agent can only cite what your library holds, so a missing key paper becomes a missing branch. Upload PDFs or send a Zotero collection, see the Quickstart and Zotero Integration.
  • Pick a question with edges. "Social media use and adolescent mental health" works. "Teenagers and technology" is too wide for one map.
  • Create a new mindmap in the project that holds the library.

Indexing the ten papers here cost 61 credits, about 6 per paper.

Step 1: Map the topic

An empty chat box shows a row of Ready-made prompts. Each one inserts a tested instruction into the chat box, with the parts you fill in shown in square brackets. You can also reach them later from the + button next to the chat box.

Pick Map a topic and replace [topic] with your question:

Map social media use and adolescent mental health from my library.

Build one branch per aspect: what it is and its core idea, the competing positions and how they differ, the main arguments and cases, and the criticisms and open problems. Cover every distinct position my library contains, not only the best-attested one.

Send it. The agent searches the library, writes claims, and attaches a citation with the author, year, and page to each one. This run took a few minutes and cost 2.7 credits.

Step 2: Read where the papers disagree

Zoom into the branch on competing positions. Each position is its own claim with its own citation, instead of one blended summary. This is the part of the review that is hardest to get from a chatbot, and the part your examiners will read most closely.

Click any citation chip to see its passage in the sidebar, then click the reference to open the PDF at that page. If a claim says more than the passage does, edit the claim or ask the agent to fix it.

Step 3: Add the counter-evidence

A review that lists only the findings reads like a summary. Click a branch to select it, so it appears under Context Items in the chat, then pick Add counter-evidence and replace [branch] with the branch name:

Find evidence in my library that challenges or qualifies the claims in the branch Main arguments and cases.

Add each objection as a sibling of the claim it targets, with citations. Keep the existing claims as they are.

Each objection lands right under the claim it qualifies, so you read a finding and its caveat together. Here a projected 14.7% drop in depression sits directly above the same authors' warning that the causal role is still contested.

Step 4: Follow the agent's leads

Every run ends with a short summary in the chat: what changed, what the library did not cover, and a few suggested next steps. Answer one of them in the chat, or ignore them and give your own instruction.

Step 5: Go deeper, one branch at a time

Work through the map branch by branch. A narrow instruction gets a focused search and a result you can check before moving on.

  • Expand a branch adds the specific studies, cases, and data under each claim.
  • Fill the gaps compares the map with your library and adds what the library covers and the map lacks. Run it after you add new papers.
  • Cite my claims checks your own claims against the library, keeps your wording, and tells you which claims the sources contradict or do not address.
  • Write a passage turns a branch into one cited paragraph, placed under that branch.

When the map holds the review you want to write, continue with Outlining a Chapter or Paper.

More prompts to try

These are not ready-made, so type them into the chat box. Select the branch first when a prompt names one.

Group the claims in the branch branch by study design: experiments, longitudinal studies, cross-sectional surveys, and reviews. Keep the citations.

For each competing position, add a claim that names the strongest piece of evidence for it and the paper it comes from.

Find the claims in this map that rest on a single paper, and tell me which ones my library could support with a second source.

Add a branch on how the papers measure concept, and show where different measures lead to different results.

You can also ask the library directly without changing the map. The Chat page of your project has its own ready-made prompts, including Literature overview and Map the debate. Its answers are cited too, and nothing is added to a mindmap.

Tips

  • Read the citations as you go. A claim is only as good as the page behind it. Checking ten claims after each run is faster than checking two hundred at the end.
  • Keep disagreements as separate claims. If two papers conflict, the map should show both. Do not ask the agent to resolve a debate your sources have not resolved.
  • Save a version before a big change. The Snapshots menu at the top of the map saves the current state, so you can compare or restore it later. See Version History and Provenance.

What's next