ChatGPT Prompts for LinkedIn Connection Messages: A 15-Prompt SDR Library
By Daniel Okoro, Outreach Tactics. Last updated: 2026-05-30
- A generic "I'd love to connect" gets ignored, but most reps have no time to research every prospect.
- The prompt is the easy part. The clipboard-at-scale send is where accounts hit limits.
- Short, specific notes beat long pitches almost every time.
- The library only compounds when personalization is wired into a system that paces invites for you.
Why do most ChatGPT connection prompts produce generic messages?
Vague inputs produce vague output. When you type "write a LinkedIn connection request," ChatGPT has nothing to anchor on, so it returns the same hollow template every other rep is sending. The fix is structural: every good connection prompt feeds the model three data points before it writes a word.
Those three points are the prospect's role (what they own and what they are measured on), a recent trigger (a post they wrote, a job change, a funding round, an event they attended), and your shared context (a mutual connection, a comment thread, an industry overlap). Give ChatGPT those three and the draft sounds researched. Skip them and you get filler. The prompt template below is the spine the whole library is built on.
You are an SDR writing a LinkedIn connection request note under 300 characters.
Prospect role: [ROLE / what they own]
Recent trigger: [POST, JOB CHANGE, FUNDING, EVENT]
Shared context: [MUTUAL CONNECTION / COMMENT / INDUSTRY]
Write a short, warm note that references the trigger, states no ask, and sounds like one human to another. No buzzwords, no pitch.
Why it works: it forces the specifics in and caps the length out, which is exactly where most AI drafts fail.
What are the 15 prompts, organized by scenario?
The library is grouped into six scenarios SDRs actually hit. Each prompt is the spine above with the variables and instruction tuned to the situation. Fill the brackets, generate three variants, send the one that reads most like you.
Cold connect (no prior interaction)
- Role-anchored: "...reference what a [ROLE] is usually measured on this quarter, then connect on that, no pitch."
- Trigger-anchored: "...open with one line about [RECENT POST/FUNDING], make it specific enough they know I read it."
- Curiosity note (no note at all is also valid): "...write a 2-line note that earns a 'who is this' in a good way, under 200 characters."
Post-comment warm follow
- "We both commented on [POST]. Write a note that picks up that thread, under 250 characters."
- "Reference the specific point they made in [COMMENT], agree or build on it, then connect."
Mutual-connection intro
- "[MUTUAL] is a shared connection. Write a note that names them naturally without name-dropping, no ask."
- "Warm-intro tone: assume they will recognize [MUTUAL], keep it to one sentence of context."
Job-change trigger
- "They just started as [NEW ROLE] at [COMPANY]. Congratulate specifically, connect on the transition, no pitch."
- "New-in-seat note: reference one thing [NEW ROLE]s usually prioritize in their first 90 days."
Event attendee
- "We both attended [EVENT]. Reference one session or theme, ask nothing, just open the door."
- "Post-event follow: 'good to be in the same room at [EVENT],' under 200 characters, casual."
Re-engage a stale lead
- "We connected months ago and never spoke. Write a light re-opener tied to [TRIGGER], no guilt, no pitch."
- "They went quiet after [INTERACTION]. Write a one-line nudge that gives them an easy out."
Voice and variant prompts
- Variant generator: "Give me 3 versions of the note above at different warmth levels, all under 280 characters."
- Banned-phrase pass: "Rewrite this with zero buzzwords. Remove 'synergy', 'leverage', 'circle back', 'I'd love to'."
Run prompt 14 on every draft so you always have an A/B pair. For more hand-written examples to feed the model as reference, see Linked Insider's connection-request examples and the no-note connection request note guide.
Want to put this into practice?
Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.
Start Free →How do you make ChatGPT outputs sound like you and not a bot?
Calibrate the model to your voice first, then edit hard. Before generating connection notes, paste five messages you have actually sent and prompt: "This is my voice. Match the rhythm, contractions, and bluntness. Do not get more formal than this." The model copies your patterns instead of defaulting to corporate-AI cadence.
Then run three filters on every output. Cap the length: short notes beat walls of text, and a tight note reads human. Strip the banned phrases with prompt 15. Finally, run the "would I actually send this" pass, because if a line makes you cringe, the prospect will feel it too. The point is not to hide that you used AI. The point is that the message is specific and human enough that the question never comes up. The humanize-AI-outreach playbook goes deeper on the edit pass.
Can you scale these prompts without getting rate-limited?
Not by copy-pasting drafts by hand, which is precisely where reps trip the limits. A prompt produces a draft. It does not pace your sends, respect weekly invite ceilings, or stop you at the point where more volume starts costing you accepts. LinkedIn caps connection invitations on a weekly basis, and the platform's own limits are not negotiable. See LinkedIn's Help Center for the current invitation ceilings.
The deeper trap is the volume tax, not the hard cap. Across Reachium's analysis of LinkedIn outreach data, acceptance peaked at 34% for accounts sending 10-19 invites a day and fell to 30.6% at 20-29 a day. More volume produced fewer accepts. Pasting AI drafts as fast as you can generate them pushes you straight into that decline. The discipline that protects your accept rate is reviewed in why your connection requests stop getting accepted and the case against 100 requests a day.
How do you turn a prompt library into a sending system?
You bake the personalization into an automated motion that paces invites for you. The prompts solve the blank page. A sending system solves the part the clipboard cannot: consistent volume at a safe pace, personalization applied across a list instead of one note at a time, and accept rates that hold because the cadence respects the volume tax.
This is the difference between a clever prompt and a repeatable channel. An automated Outreach campaign on the verified LinkedIn API applies your personalization variables across a segment and meters the sends so you never out-run your own accept rate. The flagship numbers behind this pacing logic live in the LinkedIn outreach benchmarks for 2026. The companion piece, ChatGPT prompts for LinkedIn connection requests, covers the request-note variant of this same workflow.
Want to put this into practice?
Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.
Start Free →How do you measure whether the prompts are working?
Track accept rate by prompt and scenario, then reply rate, then meetings booked. Tag each draft with the prompt number that produced it so you can see which scenarios convert. A cold trigger-anchored note and a mutual-connection note are different bets, and the data tells you which one to run more of.
For a clean read, A/B the openers using prompt 14's variants and split your list. Reachium's data shows a 28% average connection acceptance rate across LinkedIn sequences, with about 29% of accepted connections replying, so use those as a rough sanity check on your own numbers. If you are well below, the inputs are probably too vague, not the prompts. The mechanics of a clean opener test are in how to A/B test connection messages.
FAQ
How do you use ChatGPT to personalize LinkedIn connection requests?
Feed the model the prospect's role, a recent trigger such as a post or job change, and your shared context, then cap the output length. The specifics are what make the draft read as researched rather than templated.
What variables should a LinkedIn connection prompt include?
Three at minimum: role (what the prospect owns), a recent trigger (a post, funding round, event, or job change), and shared context (a mutual connection or comment thread). Add a length cap and a no-pitch instruction so the note stays short and warm.
Can you scale ChatGPT-written connection messages without getting rate-limited?
Not by hand. A prompt produces a draft, but copy-pasting drafts at high volume pushes you past LinkedIn's invite limits and into the volume tax, where more sends produce fewer accepts. An automated campaign that paces invites is what protects the accept rate.
Do AI-written connection requests actually get accepted?
Yes, when the inputs are specific. The acceptance gap is driven by personalization, not by who wrote the draft. A generic AI note performs like any other generic note, while a specific one tied to a real trigger performs like a researched one.
Should you mention you used AI to write the message?
No need to flag it either way. The goal is a message specific and human enough that the question never comes up. Calibrate the model to your voice and run the "would I actually send this" edit pass before sending.