Can I export all my Readwise highlights?

Yes. Readwise has a built-in export under Settings → Export. You can pull every highlight as Markdown, CSV, or JSON. The Markdown export preserves book/article structure and is the most useful format for feeding into another tool.

How many highlights does a typical Readwise user have?

Heavy readers accumulate 5,000–20,000 highlights over a few years. That's a remarkable dataset — your taste, your obsessions, the lines you keep coming back to. Most of it never gets re-read, which is the whole problem.

Why not just paste my highlights into ChatGPT?

Token limits. Even 1,000 highlights blow past ChatGPT's context window, and the model has no way to know which 30 highlights are relevant to a given question. You need a layer in between — something that pre-distills your highlights into a persona summary that fits in any prompt.

Does Readwise have an AI integration?

Readwise itself has some surfacing features (daily review, AI chat with your library). It's good. But the data lives inside Readwise and doesn't follow you when you ask Claude, Gemini, or anything else. The highlights are yours, but the AI layer over them is theirs.

What's the best way to use Readwise highlights with an AI?

Export the full library as Markdown. Drop it into a memory vault that runs distillation across all of it — extracting the recurring themes, the authors you keep returning to, the ideas you've highlighted more than once. The distilled persona is short enough to paste at the top of any AI conversation.

Does Konshus support Readwise imports?

Yes, on Founding and above. The importer takes the Markdown export from Readwise and creates atoms for the highlights that show up as patterns — repeated authors, recurring themes, ideas you've highlighted multiple times. The persona that comes out the other side actually sounds like you.

A guide · ~7 min read

Turn 10 years of Readwise highlights into an AI persona

Readwise is the most underrated dataset most knowledge workers own. Five, ten, fifteen thousand highlights — every line you cared enough to save from every book and article you've read in the last decade. It's the closest record of your intellectual taste in existence. And almost nobody is using it for anything but the daily review email.

The numbers

5,000–20,000 highlights for a typical heavy Readwise user over 3–5 years.
~95% never re-surfaced. Daily review hits a handful per day; the rest sit in storage.
2–5 MB of Markdown for a full export. Tiny by any standard.
$8–10/mo Readwise subscription — but the highlights themselves are yours, regardless of subscription state.
0 native integrations with ChatGPT, Claude, or Gemini that would let those models actually read your library.

The export, in 3 steps

Open Readwise in the browser, go to Settings → Export.
Pick Markdown as the format. (CSV works too, but Markdown preserves the book/article hierarchy.)
Download the file. It's a single Markdown document with every highlight, every source, every note you've ever added.

Why pasting it into ChatGPT doesn't work

The intuitive thing is to drop your Readwise file straight into a chat and ask the model to "use this as context." Two problems:

Token limits. A serious Readwise library is far too big to fit in any model's context window. You'd be paying to truncate yourself.
Relevance. Even if it fit, the model has no idea which 30 highlights matter for your current question. You'd get vague, average-y answers.

What works is a distillation layer: pre-process the whole library once, find the patterns (authors you return to, ideas you've highlighted multiple times, themes that cross books), and produce a short persona file. That persona fits in any prompt and actually reflects your taste.

Short answer

Export your Readwise library as Markdown (Settings → Export). Drop it into a memory vault that distills your highlights into a compact persona summary — recurring authors, repeated ideas, the themes that cross sources. Paste that persona at the top of any AI conversation and you'll get answers that actually sound like a model that's read what you've read.

The portable answer

Konshus has a Readwise importer on Founding and above. You drop the Markdown export in once and we run distillation across the whole library — finding the authors you keep coming back to, the ideas you've highlighted in different books, the themes that show up across years. The output is a persona file short enough to paste into ChatGPT, Claude, or anything else.

See also Notion as AI memory and Obsidian → portable persona if your second brain lives somewhere else.

Frequently Asked Questions

Your highlights, finally working for you.

Konshus distills your full Readwise library into a portable persona — so the books you've read can shape every AI conversation, not just the daily review email.

Meet Konshus