How to Export LinkedIn Posts for Analytics and AI That Writes Like You
July 14, 2026
You want your LinkedIn posts out of LinkedIn, and usually it comes down to one of two reasons. The first is to understand what's working, so you can put your posts next to their engagement and stop guessing at what to write next. The second is to feed your writing to Claude or ChatGPT so it can draft in your voice. When you go looking for the export, the common wisdom is that LinkedIn doesn't offer one.
That's only half true. LinkedIn will give you a CSV of your entire post history. I requested mine while writing this, and it came back with 3,538 posts going back years. What the export leaves out is how any of those posts performed, and it covers your own account only. There is no way to pull someone else's posts through it.
Those two limits are what decide which tool you actually need. So before you export anything, it's worth being clear on two questions:
- Whose posts do you need: your own, or someone else's?
- What are they for: analytics, or an AI that writes in your voice?
The rest of this article works through the options for each.
What LinkedIn's export actually gives you
LinkedIn's native export is real, free, and better than its reputation. You'll find it under Settings & Privacy, in the Data privacy section, listed as "Get a copy of your data." Request the archive and LinkedIn emails you a zip file, usually within a few minutes and at most within a day.
The file you want inside it is Shares.csv. It holds your complete posting history: the date and link of every post, the full text you wrote, any URL or media you attached, and the visibility you set. If your goal is to have a record of everything you've published in one place, you already have the tool, and it belongs to LinkedIn.
The same archive includes two files that look like engagement data but aren't. Reactions.csv and Comments.csv record the reactions and comments you gave to other people's posts, not the ones your posts received. That difference is easy to miss and it matters a great deal, which brings us to what the export can't do.
Where the free export stops
Two things are missing from that CSV, and they happen to be the two things that make post data worth having.
The first is performance. Shares.csv tells you what you wrote but nothing about how it did. There are no reaction counts, no comments received, and no measure of reach. You can reread every post from 2023 without learning which of them landed. Impressions and audience demographics are absent as well, and no tool can recover them, because LinkedIn keeps that data behind its login wall.
The second is everyone else. The export is limited to your own account. If you want to study a competitor, learn from a writer you admire, or read up on a prospect before a call, the native archive gives you nothing to work with.
Both limits have the same solution. I built Flux after running into them one too many times. It pulls the full post history of any public profile or company page, yours or anyone else's, and it brings the engagement along with it:
- The date, type, and full text of every post
- Total reactions, broken down by type: like, celebrate, support, love, insightful, funny
- Comments and reposts
- Hashtags, mentions, links, and media
- Author details
Like the export, it asks for nothing but a public profile URL. There is no login and no browser extension. What comes back is the performance data the native export omits, for any profile rather than only your own. The Pulse plan runs $3 a month, or $30 a year, and keeps that history refreshed daily with a weekly summary of your top posts, your best times to publish, and engagement grouped by topic and format. For most analytics work, your own or a competitor's, that is the whole job.
Exporting for AI is a different problem
More and more, the reason people want their posts isn't analytics at all. It's to teach an AI to write like them. The usual approach is to export the archive, paste the entire thing into Claude or ChatGPT, and ask for the next post in that style. It rarely works well, and the reasons are worth understanding, because they point straight at what does work.
The first problem is that handing over everything gives the model no way to prioritize. Your archive holds your strongest posts and your weakest, your voice from years ago and your voice today. Asked to imitate all of it, the model averages it, and the average is exactly what you were trying to improve on. Because the free export carries no engagement data, there is nothing in the file to tell the model which posts were worth imitating in the first place.
The second problem is length. A few thousand posts is a large amount of text, and even models that technically accept it will retrieve from it poorly. The one post that should anchor today's draft sits somewhere in the middle of a very long paste, competing with hundreds of posts that have nothing to do with the topic.
The third problem is that the file is out of date the moment you download it. Your writing keeps developing, and the post most worth learning from is often the one you published last week, not one sitting in a CSV from three months ago.
What actually helps is retrieval rather than volume. When you sit down to write about pricing, the assistant should be able to reach for your own past posts about pricing, and specifically the ones that performed, and use those to guide both the content and the phrasing. A handful of well-chosen examples produces a far better draft than the entire archive at once.
This is the job the Flux Edge plan was built for, at $10 a month. Edge makes your post history, refreshed daily and with engagement attached, available to Claude or ChatGPT as an MCP server. MCP, the Model Context Protocol, is an open standard that lets an AI assistant query an outside data source directly rather than working from whatever you happen to have pasted in. Connecting your history that way improves the writing in a few concrete ways.
The assistant can query the history on demand, so a request like "find my hiring posts that beat my average engagement" becomes a live lookup rather than a manual search through a spreadsheet. Because Edge indexes posts by meaning, asking "what worked when I wrote about pricing?" surfaces the right posts even when they never used the word pricing. The engagement data travels with each post, so the model draws on what performed rather than on your average, and Edge weighs older posts against the audience you had at the time. Nothing goes stale, since yesterday's post is already indexed and there is no file to re-export. And once you've drafted something, Flux can score it before you publish, which puts a number on whether it reads the way your best posts do.
The same connection works for profiles other than your own. You can point the assistant at a writer whose style you want to study, or at a prospect's history before you reach out. Pulse includes the first two weeks of Edge, so you can try the MCP workflow before committing to the higher tier.
Which tool fits your case
The two questions from the start, answered:
| Whose posts, and what for | Use | Price |
|---|---|---|
| Your own posts, text only | LinkedIn's native export | Free |
| Your own posts, with engagement | Flux Pulse | $3/mo |
| Someone else's posts, for research | Flux | $3/mo |
| An AI that writes in your voice | Flux Edge, with the MCP connection | $10/mo |
What to do with the data once you have it
Whichever route you take, the value is in the patterns rather than the file. With posts and engagement side by side, three questions do most of the work:
- Group your posts by topic and compare the average engagement of each. The topics that consistently beat your baseline are the ones your audience is asking for more of.
- Compare formats against one another. Text, image, and carousel posts often perform very differently, and the data has a way of correcting assumptions you didn't know you were making.
- Line your post dates up against business outcomes such as inbound messages, booked calls, or signups. That connection is the analysis that matters most, and it's the subject of a separate piece on LinkedIn analytics tools for creators.
Frequently asked questions
Can I train ChatGPT or Claude on my LinkedIn posts?
You can, and it doesn't require fine-tuning a model. The practical version is retrieval: give the assistant access to your post history and let it draw on relevant examples as you write. A simple version is to paste twenty or thirty of your best posts into a project and reference them. A more complete version is to connect Flux Edge over MCP, which lets the assistant query the full history, with engagement data, on demand.
Does LinkedIn's free export include how my posts performed?
No. The archive's Shares.csv contains your post text, dates, links, and media, but none of the reactions, comments, or reposts your posts received. The Reactions.csv and Comments.csv files in the same archive record activity you took on other people's posts, not activity on yours. Performance data has to come from a third-party source such as Flux.
Can I export someone else's LinkedIn posts?
Not through LinkedIn's own export, which is limited to your account. Their public posts can be pulled through Flux using any profile or company page URL. Competitor analysis and researching people before reaching out are the two most common reasons to do it.
How far back does the data go?
To the beginning. LinkedIn's own analytics only show the last 90 days, but the posts themselves stay public, so both the native export for your posts and Flux for any public profile reach back to the first one.
The short version is this. If you only need the text of your own posts, LinkedIn's export is all you need, and it costs nothing. Once you need the engagement behind those posts, or anyone else's posts, or an AI that writes in your voice because it can see what you've actually published and what worked, that is the gap Flux was built to close. Pulse covers the analytics at $3 a month, and Edge adds the connection to Claude and ChatGPT at $10. Both grew out of the same frustration that produced this article, which is that the export button was never going to be enough on its own.