Analyze LinkedIn connections with ChatGPT
Anna SheronovaUpdated
ChatGPT can analyze your LinkedIn connections once you upload Connections.csv from your LinkedIn data export. The file has seven columns: First Name, Last Name, URL, Email Address, Company, Position and Connected On. That is enough to count people by company and chart when you connected, year by year. It holds no record of whose profile you opened.
The export is one of four ways to give an assistant LinkedIn data, compared in can ChatGPT access LinkedIn. The other way to bring your own history in is a connector, explained in what a LinkedIn MCP server is.
Get the file and upload it
Request the archive
On a computer: Me, Settings & Privacy, Data privacy, Get a copy of your data, then Download larger data archive. LinkedIn says it emails the archive within 24 hours.
Unzip it
Open Connections.csv. The same zip holds messages.csv and Invitations.csv.
Delete the note lines
Three lines of notes about hidden email addresses sit above the header row. Delete them so the column names are the first row.
Upload to a chat
Attach the file in ChatGPT or Claude and say what you want: counts, a list, or a chart.
OpenAI's data analysis guide asks for descriptive column headers in the first row and one row per record, which is why the note lines go. The full request steps, with the timing on one real account, are in export LinkedIn connections to CSV.
A sample Connections.csv
This sample has the real layout, with the note lines removed. Every person and company in it is fictional.
First Name,Last Name,URL,Email Address,Company,Position,Connected On
Mara,Quint,https://www.linkedin.com/in/sample-mara-quint,,Northwind Labs,Head of Product,14 Oct 2026
Jonas,Feld,https://www.linkedin.com/in/sample-jonas-feld,,Northwind Labs,Senior Recruiter,02 Sep 2026
Priya,Raman,https://www.linkedin.com/in/sample-priya-raman,priya.raman@example.com,Bellhaven Health,VP Operations,19 Jun 2026
Tomas,Reyes,https://www.linkedin.com/in/sample-tomas-reyes,,Copperline,Founder,03 Mar 2026
Elena,Brandt,https://www.linkedin.com/in/sample-elena-brandt,,Northwind Labs,Engineering Manager,27 Nov 2025
Sam,Okafor,https://www.linkedin.com/in/sample-sam-okafor,,Harbor and Pine Consulting,Partner,11 Aug 2025
Lena,Vogt,https://www.linkedin.com/in/sample-lena-vogt,,Bellhaven Health,Talent Acquisition Lead,30 Jan 2025
Owen,Marsh,https://www.linkedin.com/in/sample-owen-marsh,owen.marsh@example.com,Copperline,"Sales Director, EMEA",16 Sep 2024
Nadia,Haddad,https://www.linkedin.com/in/sample-nadia-haddad,,Studio Haddad,Brand Designer,05 May 2023
Chris,Albright,https://www.linkedin.com/in/sample-chris-albright,,Stellar Freight,Recruiter,21 Feb 2022
Yuki,Tan,https://www.linkedin.com/in/sample-yuki-tan,,Northwind Labs,Data Scientist,09 Jul 2020
,,,,,,12 Mar 2019
Ruth,Kalman,https://www.linkedin.com/in/sample-ruth-kalman,,Bellhaven Health,Chief Medical Officer,28 Oct 2018
Three details match real exports. Most Email Address cells are empty. A title with a comma, like Owen's, sits in quotes. Some rows carry only a date, with no name, URL or company, so nobody can tie them to a person.
Prompts the sample can answer
"Count my connections by Company, largest first." The sample has 13 rows. Northwind Labs has 4, Bellhaven Health 3, Copperline 2, and three companies have one each. The date-only row has no company, so ask it to count blanks separately.
"List everyone whose Position is about recruiting or hiring." Jonas (Senior Recruiter), Chris (Recruiter) and Lena (Talent Acquisition Lead). A search for the word "recruiter" alone would miss Lena, because Position is whatever the person typed.
"Chart connections per year from Connected On." The dates read like 14 Oct 2026, so say that in the prompt. The sample gives 4 in 2026, 3 in 2025, then one each in 2024, 2023, 2022, 2020, 2019 and 2018.
"Who at Northwind Labs did I connect with in the last two years?" Mara, Jonas and Elena. Company and date together narrow it down. Yuki is there too, but from 2020.
"Which rows have an email address?" Priya and Owen, 2 of 13.
For charts you can reproduce, see LinkedIn network analysis with AI.
What the file can't answer
| You ask | From Connections.csv | Why |
|---|---|---|
| How many people do I know at each company? | AnswersYes | The Company column, one value per row. |
| Which recruiters am I connected to? | AnswersYes, by job title | The Position column is free text, so ask for every title that means recruiting. |
| How many people did I connect with each year? | AnswersYes | Connected On has a date on every row. |
| Who can I email? | Mostly blankA few rows | Email Address is filled only for people who allow it. |
| Who did I talk to last, and who went quiet? | Other fileNo | Not in this file. messages.csv, in the same archive, has a date on each message. |
| Who viewed my profile, and whose profile did I open? | Not in exportNo | No file in the archive covers either. |
| Which connections react to my posts? | Not in exportNo | The archive holds the reactions you gave, not the ones you received. |
The archive is still useful beyond this one file. messages.csv has full text and a date for each message, archived conversations included, so a second upload can answer who you last talked to. Invitations.csv lists invitations sent and received, with the note and the date. LinkedIn's download page says it does not provide who viewed your profile, and the full export has no file for reactions other people left on your posts. Every file shows your account on the day you downloaded it.
When a connector fits better
If your questions are about what happens week to week, a file you re-download each time gets old fast. LinkedAsk keeps a record of your own LinkedIn activity as you use it, in Chrome: profiles you visited, people who viewed you, invites you sent and whether they were accepted, reactions, and your posts with their numbers. You add it as a connector, then ask "whose profile did I open twice this month but never invite?" and get an answer from that record. Claude takes custom connectors on every plan, with one on Free. In ChatGPT, OpenAI lists full MCP for Business and Enterprise/Edu and read-only connections for Pro, as of October 2026. Cursor takes it too.
Steps for each assistant are in connect ChatGPT to LinkedIn and connect Claude to LinkedIn.
LinkedIn connections in ChatGPT: questions
Can ChatGPT analyze my LinkedIn connections?
Yes, from Connections.csv in your LinkedIn data export. Delete the three note lines above the header, upload the file and ask by company, job title or year. It answers only from the seven columns in the file.
Can ChatGPT analyze a LinkedIn profile?
Yes. Open the profile, click More or Resources and choose Save to PDF, then upload the PDF. That works for your own profile or another member's, as it looked on the day you saved it.
Why are most emails missing from Connections.csv?
Each connection decides. LinkedIn fills Email Address only for people who allow their connections to download it, in their Email visibility setting. Your own settings cannot change that.
Can ChatGPT tell me who viewed my profile from the export?
No. LinkedIn does not provide Who's Viewed Your Profile data in the archive. A record kept from now on, such as LinkedAsk, can answer that over a connector.
Does this work in Claude too?
Yes. Claude lists CSV among the file types it works with, so Connections.csv, messages.csv and Invitations.csv can go into one conversation.
Keep reading
- Guide · 5 minLinkedIn MCP: what it is and what it can readA LinkedIn MCP lets Claude or ChatGPT use LinkedIn data. Six kinds side by side by data source, what each reads and does, and where LinkedIn's ChatGPT app fits.
- Guide · 3 minFind warm introductions on LinkedIn with AIPrompts that find who you know at a company, rank them by your last message, and draft the request, with the LinkedIn data each prompt reads.
- Guide · 4 minWho to follow up with on LinkedIn: AI promptsPrompts that rank who to message on LinkedIn today, from replies you owe to profile viewers worth a hello. Each one says what data it needs and where to get it.
Ask about what the export leaves out.
Install the extension, then connect LinkedAsk to Claude, Cursor, or ChatGPT on a plan that allows custom MCP servers. It answers from the day you install it.