People who'd engaged with my posts accepted 65.8% of invites. A cold list, 18.6%.

What this means for you

The numbers that matter

The closest public method to ours is Reachium's LinkedIn outreach benchmarks 2026 (180,155 requests given 30 days to mature, any inbound message counts as a reply): "Of accepted connections, 27.55% produced a reply". HeyReach's 96,051-campaign benchmark puts replies per accepted connection at 18.10% for the typical campaign and says stronger campaigns reach 28.89%. Our 43.3% sits above both, with two honest asterisks. We count a reply to the connection note itself, and some of those are a short "thanks for connecting". And every one of our requests carried a note: Reachium's own connection note study found "47.8% replied when the request carried a note against 25.5% when it did not", so like for like we're in the same range, not ahead of it. Where we're not ahead is acceptance over every request: 21.1%, against 27.11% at Reachium and 20.75% for HeyReach's typical campaign. For the other public LinkedIn and email reports side by side, see every public cold email benchmark, compared.

All invitations, followed to the end

Of 706 connection requests sent between 7 July 2026 and 15 September 2026, 149 were accepted: 21.1%. 449 were withdrawn after going unanswered and 108 have no recorded outcome yet. If you only count requests that reached a clear yes or no, it's 24.9%. Of the 263 people who accepted a request anywhere in the log, 114 replied (43.3%).

The 95% range on the 21.1% is 18.3% to 24.3%. Acceptance is fast when it happens: on the 74 accepts with a live timestamp, the median was 7 hours after the request and 81.1% came within three days.

Acceptance by segment

Only segments with at least 30 invitations. The range is the 95% interval: where two ranges overlap, I wouldn't call the difference real.

One thing moved acceptance: whether the person already knew who I was. People who had liked or commented on one of my posts accepted 65.8% of the time, more than three times the 18.6% of a cold list. That group is only 38 people, so the exact figure is loose, but its range doesn't come near the cold list's.

Title made no difference: founders and CEOs accepted at the same 18.6% as other senior roles. Neither did the weekday. What I couldn't test: a note against no note, because every request in the cohort carried one, and country, because 592 of the 706 requests went to people in the Netherlands (21.8% accepted) and the other 114 are too mixed to say anything about. Acceptance also differed a lot between senders and markets, which is why the blended figure is the honest one to quote.

What happens after accept

114 of 263 accepted connections replied at some point. The bigger surprise is when: 59 of the replies I could time came before our first follow-up message, as an answer to the connection note or to the accept itself. The scheduled messages after that did much less:

Those are small numbers, so read them as a direction, not a rate: each message drew fewer replies than the one before, and most of the conversation starts at the note. On the 22 replies where both moments have a live timestamp, the median time from accept to reply was 3.9 days.

Booking links: since 15 August 2026, 72 people clicked one 145 times. That counter files every click under LinkedIn, including links sent in email replies, and it can't tell a person from a link preview, so I don't turn it into a rate.

The weekly invite limit, measured

LinkedIn doesn't publish a weekly number and it isn't fixed. When an account hits it, the API we send through returns an error called cannot_resend_yet. The name suggests re-inviting someone; in August I checked by hand in LinkedIn itself, and it said the weekly limit for invitations had been reached. We logged that error 82 times, on 27 sender-days.

In weeks where the sending account is on record, one account sent a median of 53 invitations and never more than 87. Our busiest week overall, 27 July to 2 August 2026, had 117 invitations and 94 withdrawals on our own outreach with 0 refusals, so withdrawals didn't push us into the limit that week.

Withdrawals are housekeeping. I don't want a big pile of unanswered invitations sitting on an account, so stale ones get pulled back. We withdrew 1,886 invitations in this window; 486 of them were for invitations in this log, a median of 15 days after they were sent, and the rest were older invitations from before the log started.

How this compares to vendor averages

Expandi's 2026 LinkedIn outreach benchmarks (published 19 May 2026) analysed 13,218,869 connection requests from 13,302 accounts and report "28.5% connection acceptance, 3.0% connection-note reply, 10.4% message reply." Belkins' LinkedIn outreach study (updated 29 June 2026, 15 million touchpoints) says "18.7% of invited prospects accepted the connection request"; the same page's summary also quotes a 26% connection rate.

Our 21.1% sits inside that range, and our cold list alone, at 18.6%, sits at Belkins' figure. That's not modesty, it's the denominator: we count every request, including the 449 that were withdrawn unanswered, and we send at a deliberately slow pace from a handful of accounts. Where we differ from the averages is the warm group. If the person has engaged with your content, the request is a different thing.

Safety rules I run with

• Slow cadence. A low daily invite count per account, and a full stop the moment LinkedIn pushes back.
• 72 hours minimum. At least three days between any two messages to the same person, enforced at send time, unless they've just written back.
• No cold sequence to a first-degree connection. Someone already in my network gets a human message or nothing.
• Nobody pitched twice. One person, one pitch, across every account, campaign and channel.

That's a method, not a guarantee. LinkedIn changes its limits without telling anyone. More on how this fits a bigger plan in my LinkedIn outreach strategy for SaaS founders, the signal-to-reply method, and the multichannel cadence template.

Method

• Source. Every invitation, accept, message, withdrawal and click is written to an event row as it happens. The send ledger behind the LinkedIn rows starts on 7 July 2026.
• Invitation. A request LinkedIn accepted for delivery. 459 failed attempts (profile not found, already invited, limit reached) are not counted as invitations.
• Accepted. An accept event, or a later message we sent that person (LinkedIn only lets you message a connection), or their campaign record moved past "accepted". Accepts were only logged as events from September, so the other two signals fill in the summer.
• Window. Rates use the 706 invitations sent 7 July 2026 to 15 September 2026, at least 35 days before 20 October 2026, so withdrawal or acceptance had time to happen.
• Replies. An inbound LinkedIn message after the invitation, or a campaign record marked replied. Our own drafted nudges are not replies.
• Scope. 4 companies selling B2B, one of them Sell Successfully itself.

Limitations

• Small n per segment. The warm group is 38 people. The message steps are 113 sends and fewer.
• Few accounts, one tool stack. 3 sending accounts with an ID on record, one sending API. Another stack may hit limits differently.
• "No outcome" is not "no". 108 requests have neither an accept nor a withdrawal on record. If some of them were accepted quietly, the true rate is higher than 21.1%.
• Mostly one country. 592 of 706 requests went to the Netherlands.
• Not comparable with the hub's LinkedIn figure. This page follows each invitation to its outcome. Use this page for acceptance.