What happened
Belkins' 2026 LinkedIn outreach study, run with Expandi across 15.1 million outreach touchpoints, publishes benchmark reply rates alongside message-level performance data. Overall reply rate lands at 7.2 percent, with an industry range of 4.2 to 10.5 percent. Sending a personalized note with a connection request pushes reply rates from 5.44 to 9.36 percent, a 72 percent lift, even though acceptance rates barely move at all. AI-written first messages reply at 4.19 percent versus 2.60 percent for non-AI openers, a clear win for AI at the top of a sequence. Follow-ups tell the opposite story: human-written follow-ups outperform AI-written ones consistently across the dataset. Southern Europe leads with an 11.81 percent reply rate, well above any other region tracked.
Why AI-written LinkedIn openers work but follow-ups do not
MagnetizeX builds founder visibility systems for B2B firms.
An opener is a pattern-recognition problem. AI is good at fitting a relevant hook to public information about a prospect at scale, which is exactly what a strong first message needs. A follow-up is a memory problem. It has to reference the specific thread, the last thing said, and the reason to still care two days later. AI drafts follow-ups that read like a new cold pitch stapled to an old one, and prospects who accepted the connection are more attuned to that mismatch than someone seeing you for the first time.
By the numbers: 15.1 million outreach touchpoints, 72 percent reply lift from a personalized connection note, 4.19 percent AI opener reply rate against 2.60 percent for non-AI openers, and 11.81 percent reply rates in Southern Europe.
- Route AI to the opener onlySet your outreach tool to generate the connection note and first message with AI, using prospect signals from LinkedIn activity, funding events, or hiring pages. Then hard-code every follow-up as a human template with variables limited to first name, company, and the specific thing they said. Measure the split for two weeks.
- Rewrite follow-ups around what changedA good follow-up references something new since the last message: their post, their promotion, a competitor announcement, or a shared connection who commented. Ban the phrases just following up and wanted to check in. Both signal you have not paid attention, and they cost more replies than any subject line change earns.
- Pick your geography carefullyAn 11.81 percent reply rate in Southern Europe against low single digits elsewhere is not marketing noise. If your ICP has any presence in Italy, Spain, or Portugal, run a two-week campaign into those markets before expanding. The same message will earn a materially different pipeline volume for the same amount of effort.
What to do this week
Open your outreach tool and audit your last 50 sent messages by stage. Split them into openers and follow-ups, then flag which were AI-generated. Reply rate is likely inverted from what you assumed: AI is helping at the top and hurting on second and third touches. Rewrite your follow-up templates by hand, tied to one specific trigger per template. Ship the new sequence to a small segment first, then compare reply rates against the old one after two weeks of live data. Log the results in a simple weekly note so you can catch the moment AI follow-up performance shifts if the underlying model changes. That habit turns a static playbook into an active one and keeps your reply rate compounding while everyone else's decays.