LinkedIn released an automated follow-up feature in 2025. The promise: if candidates don't respond to your initial InMail within a set timeframe, LinkedIn automatically sends a follow-up message. According to LinkedIn, this increases InMail acceptance rates by 39%.

We enabled it on a Friday afternoon.

By Saturday morning, our inbox had exploded.

By Sunday, we were manually apologizing to 847 people.

Here's what happened.

"Just Checking In!" x 847

The automated follow-up feature works like this:

1. You send an InMail to a candidate
2. If they don't respond within X days (we set it to 3 days), LinkedIn sends an automatic follow-up
3. The follow-up message is AI-generated based on your original InMail
4. Profit??? (Narrator: there was no profit)

We thought: "Great! We send a lot of InMails. A lot of people don't respond. Auto-follow-ups will catch the people who just missed the first message."

What we didn't anticipate: LinkedIn's AI would send follow-ups to everyone who hadn't explicitly responded, regardless of context.

Within 24 hours of enabling the feature, LinkedIn sent 847 follow-up messages.

All of them said some variation of:

"Hi [Name], just checking in on my previous message! Would love to connect about the [Role] opportunity at [Company]. Let me know if you're interested!"

Who Received These Messages

Here's the breakdown of the 847 people who got auto-follow-ups:

Active Candidates We Were Already Talking To (143 people)

These were candidates who had responded to our initial InMail, were in active conversations with our recruiters, and in some cases had already completed first-round interviews.

But because they hadn't responded to the specific InMail thread (they'd responded via email or phone instead), LinkedIn's AI thought they were ignoring us.

So they got a "just checking in!" message while actively interviewing.

Sample response we received:

"Uh... I had a phone interview with your team yesterday. Did something go wrong? Are you not moving forward with me?"

Cue panicked clarification emails from our recruiters.

People Who Already Accepted Offers (11 people)

This is where it gets really embarrassing.

We'd hired 11 people in the past month who'd originally been sourced via LinkedIn InMail. They'd accepted offers. Some had already started.

But because the original InMail thread was never formally "closed" in LinkedIn's system, the AI sent them follow-up messages.

Sample response:

"I literally started working at your company on Monday. I'm sitting in orientation right now. Did you mean to send this to someone else?"

Even worse: one of them forwarded the message to their new manager asking if there was "some kind of system error" because they were concerned it looked like we were still recruiting for their role.

People Who Declined Weeks Ago (67 people)

These were candidates who had explicitly responded to our initial InMail with "thanks, but I'm not interested" or "I just accepted another offer."

The AI sent them follow-ups anyway.

Sample response:

"I told you 3 weeks ago I'm not interested. Please stop contacting me."

Fair.

People Who Asked Us to Stop Contacting Them (1 person, but it was bad)

One candidate had responded to our original InMail with:

"I've asked your company multiple times to stop contacting me. I am not interested in opportunities at [Company]. Please remove me from your recruiting lists."

This person had a note in our ATS: "DO NOT CONTACT."

The AI sent them a "just checking in!" message anyway.

Their response included the phrase "I'm considering legal options" and cc'd LinkedIn's support team.

We are very, very sorry, [Name Redacted].

People Who Never Received the Original InMail (93 people)

This one confused us until we dug into the data.

Apparently, if an InMail bounces (invalid LinkedIn account, privacy settings, etc.), LinkedIn's system still logs it as "sent." And if the person doesn't respond (because they never received it), the AI sends a follow-up.

So we sent follow-ups to 93 people who never got the original message.

Sample response:

"I have no idea what you're talking about. I never got a message from you. Also, I don't even work in recruiting—I'm a veterinarian. Why are you contacting me?"

We... have no answer for that one.

Completely Random People (532 people)

The largest category: 532 people who had been sent InMails by our team at some point in the past 6 months, never responded, and got auto-follow-ups.

This includes:

• People who were never a good fit for any of our roles (the AI didn't filter for relevance)
• People who'd been contacted about roles that were already filled
• People who work at companies we'd never hire from
• One person who is literally our competitor's CEO (very awkward)

The Inbox Explosion

By Saturday morning, our recruiting team's inbox had 300+ responses.

Some were confused ("What role are you talking about?").

Some were annoyed ("Stop spamming me.").

Some were angry ("This is the third message from your company in 24 hours.").

And one was from our own Head of Talent, who'd apparently been sent an auto-follow-up for a role he'd been InMailed about before he worked here.

His Slack message: "Why is LinkedIn asking if I'm interested in working at the company I already work for?"

The Manual Apology Marathon

We spent the entire weekend:

1. Turning off the auto-follow-up feature (should have been step one, honestly)
2. Manually reviewing all 847 follow-ups to understand what happened
3. Sending personalized apology emails to everyone who'd responded with confusion or anger
4. Updating our ATS to flag people who definitely should not have been contacted
5. Calling the person who threatened legal action to grovel (they accepted our apology, thankfully)

The Head of Talent spent Sunday writing a postmortem titled "Why We Will Never Trust LinkedIn Automation Again."

What We Learned

1. "Set it and forget it" doesn't work for recruiting automation: We assumed LinkedIn's AI would be smart enough to detect context (active conversations, closed deals, people who'd already declined). It wasn't.

2. AI has no concept of "please stop contacting me": Even explicit requests to be removed from recruiting outreach didn't stop the AI from sending follow-ups.

3. Auto-follow-ups need human oversight: We should have set the feature to draft follow-ups for human review, not send them automatically.

4. Three days is way too aggressive: In retrospect, a 3-day follow-up window was insane. We should have set it to 7-10 days minimum.

5. Always test automation on a small sample first: We enabled the feature for our entire recruiting team, across all roles, all at once. We should have tested it on 10-20 candidates first to see what could go wrong. (Spoiler: everything)

Where We're at Now

The auto-follow-up feature is permanently disabled.

We've implemented a rule: no automated outreach without human review.

We've sent 847 apology emails.

And we've learned that "39% higher InMail acceptance rates" means nothing if those acceptances are from people saying "STOP EMAILING ME."

If LinkedIn wants to offer automated follow-ups, that's fine. But maybe—just maybe—the AI should be smart enough to detect when someone has:

• Already responded via a different channel
• Already accepted a job offer
• Already declined
• Already asked you to stop contacting them
• Already started working at your company

Until then, we'll be doing our follow-ups the old-fashioned way: manually, with context, like humans.

At least humans know not to send "just checking in!" messages to people who are literally sitting in your company's new hire orientation.