How to Do a Literature Review from Your Highlights

A literature review is mostly memory work. You read forty papers over two months, and when you finally sit down to write, you need the one sentence from paper twelve that disagreed with paper three. This guide shows a simple way to keep that sentence findable: highlight as you read, tag by theme, and let an AI assistant pull your own notes back together when you write.

The usual way, and where it breaks

Most people annotate PDFs in a reader or keep notes in a reference manager. That works for one paper at a time. It breaks when the useful passages are spread across many PDFs and publisher pages, because there is no single place to search what you marked, and no easy way to see every passage on one theme side by side.

A highlight-first workflow

1. Highlight where you read. With the Glasp browser extension, you can highlight papers directly on publisher pages such as ScienceDirect, PubMed Central, Springer, Wiley, Nature, and arXiv, on web-hosted PDFs, and on PDFs you upload from your computer. Each highlight keeps a link to its source. (How to install Glasp, How to upload and highlight PDFs)

2. Add a note when a passage matters for your argument. A short note ("contradicts Smith 2021", "use in methods") is what makes a highlight useful three weeks later.

3. Tag by theme, not by paper. Tags like measurement, limitations, or counter-evidence let you see every passage on a theme across all your papers on one page.

4. Ask an AI assistant to work from your own highlights. Connect Glasp to Claude or ChatGPT with the Glasp MCP Connector. It searches what you highlighted and returns each passage with its title and link, so you can ask:

"Find everything I highlighted about sample size in the papers I read this month, grouped by paper, with links."

"Which of my highlights disagree with each other on the effect of sleep on memory?"

The assistant answers from passages you chose, and every answer points back to the source, which makes checking easy.

5. Export when you write. Download your highlights as Markdown, or send them to Notion or Obsidian, and draft from there.

Privacy

Highlights on web pages, including publisher pages, are public on your Glasp profile by default. PDFs you upload are always private. If you work with unpublished or sensitive material, private highlighting for web pages needs a Pro or Unlimited plan. The MCP Connector is read-only and reads only your own highlights.

FAQs

Does this work on paywalled papers?

Glasp works on pages you can open in your browser. If your library gives you access to a paper, you can highlight it.

Can I highlight PDFs, not just web pages?

Yes. You can highlight web-hosted PDFs, and upload PDFs from your computer to highlight them in your browser. Uploaded PDFs are always private.

Will the AI make things up?

The assistant searches your own highlights and returns them with links, so you can check every passage against its source. Treat its synthesis as a draft and verify before you cite.

Is it free?

Highlighting, notes, tags, and exports are free. Free accounts can upload up to 30 PDFs.

Which AI assistants work?

Claude (web, desktop, and Claude Code), ChatGPT, and any client that supports the Model Context Protocol.

Before you leave

If you have any questions, please refer to the FAQ or contact us.

👉 X

👉 Reddit

See you next time,

Glasp team