Who to follow up with on LinkedIn: AI prompts

This page is part of the AI prompts for your LinkedIn network. If you want to understand how an assistant reads LinkedIn data over a live connection, read what a LinkedIn MCP is first.

Get the data first

Prompts 1 to 3 and 6 run on your LinkedIn data archive.

1. In LinkedIn, open Me, then Settings & Privacy, then Data privacy, and click Get a copy of your data.
2. Choose Download larger data archive and request it. LinkedIn says the larger download arrives by email within 24 hours.
3. From the ZIP, keep messages.csv, Connections.csv and Invitations.csv. Connections.csv opens with a few lines of notes above its header, so tell the assistant to skip them.

LinkedAsk, a Chrome extension that records your LinkedIn activity, can stand in for the files once connected to your assistant, but for people you saved only, since it keeps message threads for them alone. For a weekly pass by hand inside LinkedIn, tracking follow-ups without a spreadsheet uses Mark as unread and Archive.

Conversations and invitations waiting on you

messages.csv holds the date, sender, content and profile URLs of each message, archived conversations included. The prompts below match people by those URLs, since names repeat.

1. Replies you owe

This one reads the sender and date columns of messages.csv. LinkedAsk answers it from the reply state of each person you saved.

Using messages.csv, find conversations where the other person sent the last message in the past 30 days.
Group the rows by conversation and match people by profile URL.
List them oldest first, with how many days ago they wrote and a one-line summary of their last message.

Example answer (fictional people): "Daniel Okafor wrote 12 days ago to ask whether you are free for a call in November. Sofia Marin wrote 4 days ago and attached the deck you asked for."

2. Conversations that stalled after your message

Same file, read the other way. You sent the last message and are waiting on them.

Using messages.csv, find conversations where I sent the last message between 5 and 30 days ago.
For each one, show how many days it has been and what I last asked or offered.
Put the ones where my last message asked a direct question at the top.

The window copies the rule LinkedAsk uses for its own reply states, where a conversation counts as waiting on the other person from 5 to 30 days after your message.

Example answer (fictional): "Mira Svensson, 8 days. You asked whether her team has set its budget for next quarter."

3. Accepted invitations with no message since

Invitations.csv lists invitations you sent and received, with a Direction column, the invitee's profile URL and your note. It has no column that says whether an invitation was accepted, so the prompt checks each invitee against Connections.csv. In one archive downloaded in October 2026, the file reached back just under six months, so older invitations may be missing.

LinkedAsk keeps the status of each invitation it saw you send. It holds message threads only for people you saved, though, so for anyone else it can tell you an invitation was accepted but not whether you wrote afterwards.

Using Invitations.csv, Connections.csv and messages.csv, list people I invited in the last 90 days
(rows where Direction is OUTGOING) who are now my connections and with whom I have exchanged no message since the invitation date.
Skip the notes lines above the header in Connections.csv.
Match people by profile URL and show the note I sent with the invitation, if any.

Example answer (fictional): "Tom Reyes accepted your invitation of October 3. No messages since. Your note mentioned the product meetup."

People showing interest

Prompts 4 and 5 need data your archive does not hold. LinkedIn's download page says it does not provide Who's Viewed Your Profile data, and your archive has no list of profiles you opened. Both prompts run on LinkedAsk, from the day you install it.

4. People who keep viewing your profile

LinkedAsk holds the viewers LinkedIn showed you, each with LinkedIn's rough label such as "2d ago", so a repeat view counts only if both visits were recorded. Anonymous viewers have no name and drop out. LinkedAsk also records who reacted to or commented on your posts, which lets the prompt combine the two signals.

List people who viewed my profile at least twice in the last 30 days, counting the views on record,
or who viewed it and also reacted to or commented on my posts.
Skip anonymous viewers. For each person, say whether we are connected and whether I saved them.

5. Profiles you opened and then left alone

This one combines the profiles you visited with the invitations you sent.

Find people whose profile I opened three or more times in the last 60 days
who are not my connections and to whom I have sent no invitation.
Show their headline and when I last opened their profile.

6. Draft the follow-up

Your archive has the full text of each message in messages.csv. LinkedAsk holds previews of about 50 characters, for people you saved, plus the note you keep on each, so the draft leans on your note.

For each person I pick from the list, draft a follow-up of 50 words or fewer.
Refer to what we last talked about and end with one easy question.
Give me the drafts to edit.

For variations, ask for a version for someone who went quiet months ago, or one that matches the tone of your last message.

Run the pack on your live LinkedIn activity

An archive shows one day, and prompts 4 and 5 need data it never has. LinkedAsk records your profile visits, viewers, invitations you send, reactions and comments on your posts, your own reactions, and the connections and followers it sees in your lists as you use LinkedIn. It keeps message threads for the people you save. You can connect it to Claude or ChatGPT and run these prompts on it. Prompts 1 to 3 then cover your saved people only.

The connection includes a ready-made prompt, "Who should I follow up with today". It reads the people you saved whom you owe a reply or who owe you one, and returns them in order. Each entry says why the person is due now and suggests an opening sentence. Setup is described in connecting Claude to LinkedIn and connecting ChatGPT to LinkedIn.

Pick by situation

• A backlog of old conversations: your archive with prompts 1 to 3. In the same October 2026 archive, messages.csv went back to 2019.
• A short list every morning: LinkedAsk and the ready-made follow-up prompt.
• Job hunting or selling: prompt 4 shows who is looking at you, and prompt 5 shows whom you keep looking at.
• You want a way into a company instead: the warm introduction prompts.