How to Write a LinkedIn Connection Request

What a connection request is for

A connection request is a two second decision. The person sees your name, your headline, your photo, and your note, and they answer two silent questions: do I understand why this arrived, and does accepting cost me anything. The note answers the first question. The absence of an ask answers the second.

Selling in the note breaks both answers at once. It tells the reader the request arrived because they matched a filter, and it asks them to make a buying judgment before they have any reason to spend attention on you. Nobody evaluates an offer from a stranger in a 300 character box. They just decline, or worse, they accept and never open your next message.

The right mental model is that the connection request buys you one thing: the right to send a message that will actually be read. That is all. Everything you try to squeeze in beyond that reduces the odds of getting it.

• Good outcome: they accept, and they remember roughly why.
• Bad outcome: they accept out of habit with no memory of you, which makes your first message a cold message with extra steps.
• Worst outcome: they mark it as spam or click "I don't know this person", which is widely believed to be among the strongest negative signals an account can accumulate. LinkedIn does not confirm how it weights them.

Do the research before you write: the five to seven fact rule

Sales Connector founder Wes Lemos teaches a rule that makes the writing part almost trivial: before you type a word, find five to seven facts about the person that are true of them and of nobody else on your list. Not five to seven adjectives. Facts. Things with dates, numbers, names, or specific words in them.

Once you have those facts, you are no longer trying to be clever. You are choosing which true thing to lead with. Most bad connection requests are not a writing problem. They are a research problem being solved with adjectives.

There is a second benefit to the rule, and it is the more valuable one. If you cannot find five facts about a person in about four minutes, that person does not belong on this list yet. The rule quietly audits your targeting. A list that resists research is a list built on filters that do not describe a real buyer.

• Counts as a fact: something they posted or commented in the last quarter, the way their About section is written in first person, a funding or acquisition announcement, a product launch, a podcast or webinar appearance, the specific job they left and when, a conference session they spoke in, a change to the company's homepage positioning, open roles on their careers page, a public review or complaint they left about a tool.
• Does not count as a fact: their job title, their industry, their company headcount, their city, the fact that they are on LinkedIn, or anything a filter produced.
• Best single source: their own comments, not their own posts. Comments are less curated and people remember writing them.

The shape of a note that gets accepted

Three moves, in this order. First, the observation: one specific thing about them, stated in enough detail that they know you actually saw it. Second, the reason: why that thing made you stop instead of scrolling. Third, the close: a plain statement that you are connecting, with no request attached.

The constraints matter as much as the structure. Stay well inside the character ceiling so the note is not truncated. No links. No calendar mention. No question that requires work to answer. No compliment inflation, because "huge fan of your work" from a stranger reads as either false or slightly unnerving. And no "quick question", which every reader on LinkedIn now correctly translates as "pitch".

One test before you send. Read the note out loud and ask whether it could have been sent to anyone else on your list. If it could, it is not finished. Then ask whether a colleague of theirs, reading over their shoulder, would find it embarrassing. If yes, cut the flattery.

Example connection requests, with the reasoning

These are written as examples, not as reports of real campaigns. Names in brackets are placeholders. Read the reasoning under each one rather than copying the words, because the words stop working the moment enough people copy them.

Example 1, when they have published something. "Your write-up on killing discovery calls before demos went the opposite direction from everything else I read this year, and the part about close rates stuck. I have been arguing the same thing with worse evidence. Connecting to see the follow-up." This works because it names one specific claim, states a real reaction instead of praise, and admits the sender's own position is weaker, which is disarming. It fails if the post is two years old, or if you skimmed it and got the argument slightly wrong. Referencing something you did not read is worse than sending no note at all.

Example 2, when there is nothing published, which is the common case. "You have owned RevOps at [Company] through the Series B and the move from two products to four. My guess is the CRM you inherited was built for the old catalog. Connecting either way." This works because it uses two verifiable facts and then makes a falsifiable guess. A guess gives the reader something to correct, and corrections are the easiest replies in the world to write. It fails when the guess is so safe it reads like a horoscope, or so wrong it proves you do not know the space.

Example 3, when you share a room. "We were both in the pricing session at [Event], the one where half the room pushed back on usage based billing. You were the person who said it made forecasting worse. Connecting." Shared context is the strongest currency on the platform because it is verifiable and it implies you are a real participant, not a list. It weakens fast when the shared context is a 40,000 member group nobody reads.

Example 4, through a mutual connection. "[Name] mentioned you when I asked who had actually rebuilt their onboarding rather than talked about it. She said you cut it to one week. Connecting on the strength of that." This works because it borrows real credibility and repeats a specific detail. It only works if the person named genuinely knows you and would be comfortable being quoted. Name dropping someone who does not know you is the fastest route to being reported, and it deserves to be.

Example 5, after a job change. "Congratulations on the move to [Company]. You are the third person I have watched leave [Old Company] for a smaller team this quarter, which is either a coincidence or a trend. Connecting to find out which." Job changes are a real trigger, but everyone knows it, so the window is crowded. Being specific matters more here, not less.

• Before: "Hi [Name], I help B2B founders book more meetings with a proven LinkedIn system. Open to connecting?"
• Why it fails: it describes the sender, it could go to ten thousand people unchanged, it asks for a decision, and "proven" carries no evidence.
• After: "You are hiring two SDRs and a RevOps lead at the same time, which usually means the founder is still doing the outbound. I have watched that handover go badly more than once. Connecting."
• Same sender, same business, no pitch, and now the note could only have gone to one person.

Why "I help X do Y" does not work

"I help [role] do [outcome]" is a description of your company wearing the costume of a message. It puts the burden of relevance on the reader, who now has to work out whether they are an X and whether they want Y, on your behalf, for free, in two seconds.

It also has a recognition problem that has nothing to do with quality. The pattern is so common on LinkedIn that it triggers categorization before comprehension. The reader does not evaluate your specific claim. They file the message under a type they have already decided about.

The fix is not a better version of the same sentence. The fix is to make the first sentence about them and let your offer earn its turn later. If your positioning is sharp, it will land far harder in message three, spoken to someone who has already told you something about their situation, than it ever would in a cold 300 character box.

What to do when you have nothing specific to say

Sometimes the research comes up dry. The person has a quiet profile, no posts, no talks, no press. You have three honest options and one dishonest one.

Option one is to send the request with no note at all. A blank request costs the reader nothing to evaluate, so they fall back on your profile and your mutual connections, and for well matched targets that often performs perfectly well. Do not treat that as an excuse to skip research forever, but a blank request beats a generic one every time. A weak note is not neutral. It tells the reader what kind of sender you are.

Option two is to fix your profile so it does the persuading when the note cannot. Your headline should say what you do for whom in plain words, your banner should not be a stock photo, and your featured section should show one real thing. When there is no note, the profile is the message.

Option three is to take the person off the list. Not every account on a filtered search is a buyer, and a list that produces nothing to say is a list that will produce nothing else either.

The dishonest option, which is to invent shared context or fake familiarity, is the one to avoid completely. It works on a small number of people once. It damages the account and the brand permanently.

How to tell whether your notes are working

Measure acceptance by segment, never in aggregate. Acceptance varies enormously by seniority, by industry, by how well known your company is, and by how many other people are currently messaging that same title. An overall number blends all of that into something you cannot act on.

Change one variable at a time and give each version a real cohort before you judge it. Then read the diagnostic. If acceptance is healthy and replies are near zero, the note is fine and your first message or your offer is the problem. If acceptance is low across every segment, suspect the list and the profile before you rewrite the copy. Copy is the lever people reach for first and it is rarely the biggest one.

Resist comparing your number to a figure in somebody's blog post. Published acceptance and reply benchmarks are almost always self reported, drawn from unnamed samples, and calculated on denominators nobody defines. That case is made in full on LinkedIn outreach response rates. Your own trailing numbers, split by segment, are the only benchmark worth managing against.

If you would rather not run this yourself, this is the work Sales Connector does for clients. On the Assisted plan at $595 a month the team builds the targeting, writes the copy, and runs the campaigns while the client answers their own inbox. On the Managed plan at $1,195 a month the team also answers every reply in the client's voice and books the meetings. Both are month to month with no contract.