The LinkedIn Connection Note Is Now the Worst Touchpoint

Expandi's August 2026 benchmark report just landed, and the number every sourcing lead should be staring at is 2.2%. That is the current reply rate on LinkedIn connection notes, down from 3.5% eleven months earlier. The channel most technical recruiters skip, the post-connection DM, is now roughly 4.7x more effective than the note everyone spends their personalization budget on.

What Expandi's 13.2M-message dataset actually says

LinkedIn connection note reply rates fell from 3.5% in May 2025 to 2.2% in April 2026, a 37% relative decline in under a year, while post-connection message replies held flat at 10.4%. The playbook has inverted and most sourcers have not noticed.

The dataset is not small. Expandi's 2026 benchmark covers:

• 13,218,869 connection requests
• 6,730,447 outbound messages
• 3,766,161 accepted connections
• 13,302 active LinkedIn accounts
• May 2025 through April 2026

Platform-wide averages now sit at 28.5% connection acceptance, 3.0% connection-note reply, and 10.4% message reply. Belkins' independent 15.1M-touchpoint dataset corroborates the direction, with Tuesday still leading at a 6.90% reply rate. Two independent panels, same trend line: the front of the funnel is decaying and the second touch is not.

37%

Why the note is decaying and the DM is not

The mechanism is boring and structural. A connection note hits a prospect before they have looked at your profile, your title, or a single shared connection. There is no social contract. A post-connection DM hits after the accept, which is itself a soft yes. The prospect has already evaluated the sender to some degree. That accept, worth 28.5% on average, is doing the work the note used to do.

If you are spending 20 minutes crafting a 280-character opener, you are optimizing the touchpoint LinkedIn's own user behavior has demoted.

LinkedIn outreach benchmarks 2026, in one table

Here is the compact version of the benchmark shift, plus the two supply numbers that matter for anyone hiring recruiters right now.

Two things jump out. First, Staffing & Recruiting still leads every industry at 36.5% acceptance. LinkedIn is not dead for recruiters. It is dead for the specific tactic recruiters over-index on. Second, the Apparel & Fashion floor sits at 19.9% acceptance, an 83% spread from the top, which means the platform average is nearly useless for benchmarking a technical sourcing team. If your ops dashboard shows "we are above LinkedIn average," you are grading yourself against Apparel.

InMail vs connection request: the math changed in late 2025

Recruiting InMail now runs 18 to 25% response rate against 2.2% on connection notes, but LinkedIn quietly capped Open InMail sends to under 100 per month per account in late 2025, down from a practical limit near 800. The channel with the best economics also has the tightest volume ceiling in five years.

The numbers to memorize:

• Recruiting InMail: 18 to 25% response rate, best of any vertical
• Legal and Professional Services InMail: 10.42%
• Healthcare InMail: 9.25%
• Signal-triggered LinkedIn DMs: 13.4 to 14.21%
• All-DM average: 10.3%
• Cold email: 5.1%
• Connection notes: 2.2%

And the two policy floors that constrain the whole regime:

1. LinkedIn Recruiter flags any account with under 13% response rate on 100 or more InMails in a 14-day window and can temporarily suspend InMail privileges.
2. Open InMail is capped at under 100 sends per month per account since late 2025, an 8x cut from the prior practical limit.

Combined, LinkedIn is enforcing what benchmarks were only recommending. If you cannot hit 13% reply, you lose the channel. If you try to muscle through with volume, you cannot. The platform is now on the side of the quality regime.

Why this hurts high-volume sourcing shops most

Teams that used to absorb bad targeting with InMail volume no longer have that option. A team of ten sourcers is now capped at under 1,000 Open InMails per month, total. Every send has to earn its slot. That is a research problem, not a copywriting problem, which is why the sourcing craft (finding the right 40 people instead of blasting 400) is suddenly the constraint.

Sender seniority is a 3-point non-factor

Across the 13,302 senders in Expandi's dataset, every seniority bucket lands within a 3-point band on acceptance rate, from 26.3% for junior senders to 29.4% for C-level. The "VP ghost-write" tactic buys you almost nothing.

This kills a common excuse in recruiting Slack channels: "candidates only reply to the hiring manager." At platform scale, they do not. What moves reply rates is not the sender's title. It is:

• Whether the target is actually relevant to the role (targeting)
• Whether the message is under 400 characters (InMails under 400 chars reply 22% above average; over 1,200 chars, 11% below)
• Whether a signal triggered the send (13.4 to 14.21% for signal-triggered vs 10.3% baseline)
• Whether the touchpoint is post-accept, not pre-accept

If your engineering leaders are burning their weeks writing outbound because "candidates need to hear from an EM," the data says stop. Route their time to interviews and final closes. Push the outbound to the sourcer.

The platform is now on the side of the quality regime. The only lever left is targeting.

The supply-side story: 21,155 US sourcers, 14% with Boolean skills

Refolk's index shows 21,155 Technical Recruiters and Sourcers currently active in the US versus 697 in the UK, roughly a 30 to 1 ratio, and only 2,988 of the US cohort - about 14% - publicly list Boolean Search or Sourcing as a skill. That is why most teams get hit hardest by the note-reply decline: the sourcing craft that would let them adapt is scarce on their own bench.

The concentrations are worth naming:

• Top US employers of technical sourcers: Snowflake, Google, Blue Origin, SpaceX, Experis
• Top UK employers: Monzo Bank, Meta, Capgemini
• US sourcers with explicit Boolean/Sourcing skill (2,988): concentrated at Meta, DoorDash, Rippling, Robert Half

The 86% who do not signal Boolean or active-sourcing craft are, in practice, LinkedIn Recruiter filter operators. Their entire workflow assumes the connection-note funnel that just decayed 37%. When the front-door touchpoint stops working, they do not have a fallback motion. They have another Boolean string, run against the same saturated pool that Snowflake, Google, SpaceX, and Rippling are also hitting.

This is the specific gap Refolk closes for teams that do not have a Boolean specialist on staff: you describe the person in plain English ("staff-level distributed systems engineers who left Snowflake in the last 18 months and now work at a Series B or earlier"), and get a ranked shortlist across GitHub, LinkedIn, and the open web. The targeting stops depending on whether your sourcer happens to sit in that 14% band.

The playbook shift for technical sourcing on LinkedIn

Skip the note. Invest the personalization budget in the post-accept DM. Trigger on signals. Keep InMails under 400 characters. That is the entire 2026 recruiter outbound response rate playbook in four sentences.

Concretely, here is what to change this quarter:

1. Send blank or ultra-short connection requests. The note is worth 2.2%. The accept is worth 28.5%. Do not spend the token on the note.
2. Move all personalization to the first DM after accept. That is where the 10.4% baseline lives, and where Staffing & Recruiting leads the platform.
3. Cap InMails at under 400 characters. Expandi's data shows a 22% lift over average; InMails over 1,200 characters run 11% below.
4. Trigger on signals, not on lists. Signal-triggered DMs hit 13.4 to 14.21%. A funded-round list from three months ago does not.
5. Track your Recruiter response rate against LinkedIn's 13% floor. Below that, you can lose the channel, not just the campaign.
6. Stop ghost-writing from executives. A 3-point acceptance band across seniority is not worth an engineering VP's afternoon.

The uncomfortable implication for staffing agencies and in-house teams: throughput per seat is now bounded by targeting quality, not by send volume. The under-100 Open InMail cap makes that a math problem, not a philosophy. If you cannot describe the 40 right people before you touch LinkedIn, you cannot run this playbook.

That is the workflow Refolk is built for. You ask in plain English, I return the ranked candidates across GitHub, LinkedIn, and the open web, and your sourcers spend their capped InMails on people who actually match. The connection note is dead. The DM after the accept is not. The only remaining question is whether you know, before you send it, that you have the right person on the other end.

FAQ

Is LinkedIn outreach dead for recruiters in 2026?

No. LinkedIn is dead for the specific tactic most recruiters over-use, the connection note, which now replies at 2.2%. But Staffing & Recruiting still leads every industry at 36.5% connection acceptance, and Recruiting InMail runs 18 to 25% response rate, higher than any other vertical. The channel is fine. The default playbook, connect-then-pitch with a clever note, is decaying.

What is a good LinkedIn connection request reply rate right now?

Anything above the 2.2% platform average is technically "good," but that is a low bar and a poor use of your time to optimize. The more useful benchmarks are 28.5% connection acceptance (platform average), 10.4% post-connection message reply, and 13% InMail response rate, which is LinkedIn Recruiter's official floor before it flags your account. Aim to move the DM and InMail numbers, not the note.

How did LinkedIn's late-2025 InMail cap change sourcing?

LinkedIn capped Open InMail sends to under 100 per account per month, roughly an 8x reduction from the prior practical limit near 800. Combined with the 13% response-rate floor that can trigger InMail suspension, high-volume sourcing is now structurally constrained. Throughput per seat depends on targeting quality, not send volume, which is why teams are shifting spend toward research and signal-based tools that pre-filter who is worth an InMail.

Should executives write outbound InMails to lift response rates?

Probably not. Expandi's 13,302-sender dataset shows every seniority bucket lands within a 3-point acceptance band, from 26.3% for junior senders to 29.4% for C-level. The lift you get from an executive signature is small enough that the executive's time is almost always better spent on later-stage conversations. Move the token budget to targeting and the post-accept DM instead.