How to Turn Your LinkedIn Data Export Into an AI Knowledge Base

How to Turn Your LinkedIn Data Export Into an AI Knowledge Base

If you've posted on LinkedIn for a while, you've built up a real body of work: posts that got genuine discussion, articles you spent hours on, carousels that explained something well enough that people saved them. LinkedIn lets you download your account data through its official data export. The problem is that what you get back is built for spreadsheets, not for reading. Turning it into something NotebookLM or ChatGPT can actually use takes an extra step most people don't expect. PostToSource handles the part that matters most: converting the public posts and articles you pick out, link by link, into clean documents an AI tool can read (for a carousel, that means the post's text, not the slides).

What's Actually Inside Your LinkedIn Export

To request it: click the Me icon on LinkedIn, go to Settings & Privacy → Data privacy, then, under How LinkedIn uses your data, click Download your data (some LinkedIn help pages still call it Get a copy of your data). You can pick the specific data you want or request the larger archive covering your full account history, including connections. LinkedIn's help page says the larger download arrives by email within 24 hours, and the file stays available to download for 72 hours.

What arrives is a download of spreadsheet files organised by category, not a readable document. LinkedIn's help page lists what each category holds (your comments on posts and articles, the date and URL of items you saved, your messages, and so on). The files themselves are built for spreadsheets: one row per item, columns for dates and links, no formatting. It's a thorough record. It's also structured entirely around what LinkedIn's systems needed to store, not around what a language model needs to read.

Why the Raw Export Doesn't Work as an AI Knowledge Base

This is where most people get stuck. They request a copy of their data expecting something like a portfolio of their best writing, and instead get a stack of spreadsheets.

Gemini Notebook (formerly NotebookLM) does accept spreadsheet files (CSV and Google Sheets) as sources, so the export will upload. But your posts arrive as table rows, hundreds of them in a single source, with your actual writing wedged between date and link columns. Drop that same file into a ChatGPT or Claude project and the model can technically read a CSV, but it wasn't written to be read: you'll burn context on column headers and repeated metadata instead of the posts themselves. (For what each tool's source types actually support, see our breakdowns of NotebookLM's source limit and file upload limits across ChatGPT, Claude, Gemini, and NotebookLM.)

None of this makes the export useless. It means it's a records request, not a knowledge base. Those are different jobs.

A Better Workflow: Export for the Record, Convert for Actual Use

The fix isn't to wrangle the spreadsheets into shape: it's to treat the export and your AI knowledge base as two separate things, built two different ways.

Step 1: Keep the export as your record

Request it, unzip it, and store it somewhere safe. It's LinkedIn's official copy of your account data, useful if you ever need to prove what you posted and when, or if you're leaving the platform and want your data preserved outside LinkedIn's servers. You don't need to open it day to day.

Step 2: Find the posts and articles actually worth a knowledge base

Open the posts file in your export as a spreadsheet (or scroll back through your profile's activity) and scan for the posts that still hold up: the ones with a real framework, a story people referenced later, or an argument you'd want to build on again. Copy the links to those specific posts and articles, from the export or from the post itself on LinkedIn, rather than trying to convert everything at once.

Step 3: Convert those specific posts with PostToSource

Paste each post or article URL into PostToSource and it returns a clean PDF and Markdown file: the post text with its author (the PDF also shows the date), and for articles, the headings and lists intact. It's the same conversion it runs for LinkedIn posts generally. It only works for posts and articles that are publicly visible, so anything you shared with connections only won't convert. The free plan converts one link at a time (2 a month without an account, 5 with a free one); Pro is unlimited and takes up to 10 links at once. Markdown drops straight into NotebookLM, a Claude Project, or a custom GPT's knowledge files without the spreadsheet noise.

That gets you both halves: an official record you'll rarely reopen, and a curated, AI-readable knowledge base built from the posts that actually matter.

Real Use Cases

Turning years of professional writing into a searchable library. If you've built up a consistent point of view on LinkedIn (a niche, a leadership philosophy, a running argument about your industry), converting your strongest posts lets you ask an AI tool "what have I already said about this" instead of scrolling your own activity feed.

Preparing for a platform-independent portfolio. Job searches, speaking applications, and ghostwriting pitches often need "your best thinking" assembled somewhere other than LinkedIn's own feed. A converted set of posts, fed into a NotebookLM-based research workflow, becomes that portfolio.

Backing up before an account issue. Platform suspensions and policy changes happen, and LinkedIn is no exception. Requesting the larger export gives you LinkedIn's official copy of your data; converting your best work separately means it stays usable even if you never reopen the ZIP.

Picking up where your saved posts leave off. If you've already been saving other people's LinkedIn posts for research, your own data export is the natural second half: the posts you wrote, not just the ones you bookmarked, joining the same knowledge base.

Frequently Asked Questions

Can I upload my LinkedIn export directly to NotebookLM?

You can upload the spreadsheet files, because Gemini Notebook (formerly NotebookLM) accepts CSV files as sources. They just don't make a very useful source: each file mixes hundreds of posts with dates and links in one table, so answers point back to one big spreadsheet rather than to a specific post. Converting the posts you care about gives each one its own clean source with its author, date and link.

How long does LinkedIn take to prepare a data export?

Per LinkedIn's own help page, the larger archive arrives by email within 24 hours, and the download stays available for 72 hours. You can also request just the specific data you're looking for instead of the larger archive.

Is the data export worth requesting even if I don't build an AI knowledge base from it?

Yes. It's LinkedIn's official copy of your account data, including your posts and connections. Keep it as a record regardless of whether you convert any of it.

What's the difference between this and saving LinkedIn posts I've bookmarked?

Saved posts are other people's content you've bookmarked for later; see our guide on turning LinkedIn saved posts into an AI knowledge base. Your data export is your own posting history. The conversion workflow is the same (paste a URL, get a clean document), but the export additionally gives you an official backup, not just what you chose to bookmark.

Do I need the full export if I only want a handful of posts?

No. If you already know which posts and articles belong in your knowledge base, you can skip the full export and go straight to converting those specific URLs with PostToSource. The export is mainly useful as an official record or for rediscovering older posts you'd otherwise have to scroll back through manually.