LinkedIn outreach: how to do cold outreach on LinkedIn in 2026

Why LinkedIn outreach is different from cold email

Two steps, your real name, and roughly double the replies.

The two-step. Cold email is one send: the message lands in an inbox and lives or dies there. LinkedIn outreach is two sends: first a connection request, then, only if the prospect accepts, the actual message. Each step has its own numbers and its own failure mode, and each gets its own section below.

Your identity is attached. A cold email can come from an alias on a secondary domain. A LinkedIn message comes from your face, your name, and your work history, and the prospect reads that profile before deciding on the accept. This cuts both ways: a credible founder profile outperforms any subject line, and the account behind every send is your own, which matters once automation enters the picture.

Replies run higher. In the public platform datasets, LinkedIn message replies run about double cold email replies in the same studies. The channel earns that gap the hard way: every message goes to someone who already accepted you.

Expandi's platform data across 70,000+ campaigns. The full set of LinkedIn and email studies, with sources and caveats, is on the cold outbound benchmarks page.

The catch. Those replies come at the cost of scale. LinkedIn caps how many invites you can send, and the cap is the single most important fact about the channel, so the arithmetic comes first.

The connection-request math: weekly invite limits are real

The cap makes targeting the whole game.

The cap. LinkedIn does not publish a number, but the weekly invite limit is real: send enough requests and the "you've reached the weekly invitation limit" wall goes up until your rolling seven-day window resets. The commonly reported cap is around 100 connection requests a week for most accounts; automation vendors report 150 to 200 for some Sales Navigator and long-established accounts, and lower effective ceilings for new or quiet profiles. Treat every one of those figures as observed, not official.

The arithmetic. Run the public averages through a maxed-out week. The Belkins and Expandi joint study measured 18.7% of invited prospects accepting, a 7.2% message reply rate, and 1.3% of connected prospects booking a meeting; the study is on the benchmarks page. At those averages, 100 invites a week produces about 19 new connections, and about one booked meeting a month. That is the whole channel at average quality: real replies, tiny volume.

What the cap changes. Email rations attention and scales sends; LinkedIn rations sends. Every invite spent on a poorly chosen prospect is one you cannot spend on a good one that week, so list quality moves a LinkedIn campaign more than any message edit will. The targeting discipline is the same as email, and the order of operations is in our B2B SaaS outreach guide.

Housekeeping. Withdraw invites that have sat pending for a few weeks. A large pile of ignored invites is feedback about your targeting, and it is commonly reported to lower the effective ceiling LinkedIn gives you. The next three sections are the tips founders actually ask for: the note, the first message, and the cadence.

Connection note or no note

The public data is mixed; the reasoning is not.

The short answer. Send no note rather than a templated note, and a real note only when you can name the specific reason you want this specific person.

What the data says. The automation platforms have published tests pointing both ways: some found blank invites accepting as well as or better than noted ones, others found genuinely specific notes outperforming both. Where the tests agree is the floor: a templated note performs worst.

Why blank beats template. A blank invite reads as neutral; people connect on LinkedIn without stating reasons all the time. "I came across your profile and was impressed" reads as the first line of a pitch, because it always is, and the prospect declines the pitch by declining the invite.

The note cap. Free accounts are commonly reported to be limited to a handful of personalized notes a month, five to ten in most reports, while Premium accounts have no note cap. On a free account the default is chosen for you most of the week anyway.

The first message after an accept

The accept is permission to be specific. It is not interest.

Do not pitch the accept. The most common failure on this channel has a name, the pitch slap: a paragraph of product pitch seconds after the accept. An accept is a neutral act. Treat it as permission to say something specific, not as a buying signal.

What to send. The same first-line rule as cold email: open with the one observation that proves you looked at their company, then make one ask. On LinkedIn the message reads in a chat window, so two or three sentences is the whole message; the first-line test itself is in the B2B SaaS outreach guide.

When to send it. Not within seconds; an instant message after an accept reads as automation, because it usually is. Within a day or two, while the accept is still fresh in their notifications.

How often to follow up on LinkedIn

Fewer than email, and each one earns its send.

Fewer touches. Email tolerates two or three follow-ups in a thread. LinkedIn tolerates one, maybe two. The chat thread stacks: every unanswered bump sits on the prospect's screen above the reply box, permanently, in your name.

What earns the send. The same rule as email: a new observation, a sharper ask, or a piece of work done since the last message. "Just bumping this" adds nothing and reads as automation.

The total across channels. Four to six touches across LinkedIn and email combined, spread over two to three weeks, then stop. Silence after a good sequence means the list or the message, and the fix is upstream, not a seventh touch.

LinkedIn outreach tools: what automation does, and what it risks

The tools are real, and so is the account on the line.

What the tools do. LinkedIn outreach automation tools (Expandi, Waalaxy, PhantomBuster, and dozens like them) queue connection requests and messages, run them as multi-step sequences, pace sends to stay inside the limits, and report accepts and replies in a dashboard. Functionally, they are sequencers for a channel that does not officially have one.

The fact underneath. There is no separate sending infrastructure on LinkedIn. Email lets a secondary domain absorb the reputational damage of a bad campaign; a LinkedIn send, automated or hand-typed, goes out from a person's real account. Whatever a tool sends, it sends as you.

The risk, stated plainly. LinkedIn's user agreement prohibits third-party automation software, and LinkedIn restricts accounts it decides are automated. Vendors reduce the odds with human-like pacing and conservative per-account limits, and plenty of people run these tools for long stretches without trouble, but nobody publishes a restriction rate, and the account at stake carries your name, your history, and your network.

The tradeoff in one sentence. Automation buys volume on a channel whose own math punishes volume, and stakes your account to get it. If volume is genuinely your constraint, that is what email is for, which is the routing question next.