27 ChatGPT Prompts for LinkedIn Connection Requests (That Get Accepted)

By Daniel Okoro, Outreach Tactics. Last updated: 2026-05-30

• ChatGPT writes a fluent first draft, then drops the same three tells (over-polish, "I came across", a soft pitch) into every note.
• The prompt only works if you feed it real context: the role, the recent post, the shared connection, the trigger.
• The 300-character limit has to live inside the prompt, or ChatGPT writes a note that gets truncated mid-sentence.
• More invites per day does not buy more accepts. The data shows the opposite.

Why do AI-written connection notes get ignored?

AI-written connection notes get ignored because ChatGPT optimizes for fluent, safe, generic language, which is exactly what a busy decision-maker has learned to skip. Out of the box, the model writes notes that are grammatically perfect and emotionally flat: "I came across your profile and was impressed by your work in the SaaS space." Nobody accepts that note because it could have been sent to ten thousand people, and it usually was.

There are two fixable failures. The first is the missing context. ChatGPT can only be specific if you hand it specifics, so a prompt that says "write a connection request to a marketing director" produces filler, while one that names her last post and the three tools she stopped trusting produces something a human might open. The second is the character limit. LinkedIn caps the connection note at 300 characters, and ChatGPT, left alone, writes 600 and gets cut off. The other reason notes fail has nothing to do with words: volume. Reachium's data across 316,703 LinkedIn outreach sequences on the verified API shows acceptance peaked at 34% for accounts sending 10-19 invites a day and fell to 30.6% at 20-29 a day. A perfectly prompted note still loses if it is buried in a 100-a-day blast, the failure mode covered in why you should stop sending 100 connection requests per day.

What variables make a connection-request prompt work?

A connection-request prompt works when it pins down four variables: who you are, who they are, the one specific thing connecting you right now, and the single low-friction ask. Strip any of them and ChatGPT defaults to filler.

The reliable structure is a system instruction plus a context block. Tell ChatGPT the format constraints once (under 300 characters, no pitch, no compliment, one specific reference, plain language a person would text) and then paste the prospect's context. The character cap does the most work, because the short box forces ChatGPT to cut the throat-clearing it loves. The second rule is to ban the pitch: the connection request is not where you sell, and prompts that ask ChatGPT to "introduce my service" produce the exact notes that get declined. For why the note field is worth using at all, see the case for the LinkedIn connection request note, and for the structure behind specific openers, these connection request message examples.

What are the best ChatGPT prompts for cold connection requests?

The best cold prompts give ChatGPT a fixed format and a real trigger, then ask for variants you can choose from. Cold means no shared connection and no prior interaction, so the prompt has to manufacture relevance from public signals.

Use this as your base system prompt for every cold note below:

You write LinkedIn connection request notes. Rules: under 300 characters,
no pitch, no flattery, plain conversational English, one specific reference
to the person, end with a low-friction reason to connect. Give me 3 variants.
Avoid the phrases "I came across", "I'd love to", "reach out", and "synergy".

Why it works: it bans the four AI tells up front and demands variants so you pick the least robotic one.

1. "Write a note to {name}, a {title} at {company}, who just posted about {topic}. Reference the post, not the company."
2. "They published an article titled '{title}'. Write a note that reacts to one specific argument in it."
3. "We both follow {creator}. Write a note that uses that overlap as the reason to connect."
4. "They list {skill} in their headline. Write a note that asks one genuine question about how they apply it."
5. "They just changed jobs to {new role}. Write a congratulations-adjacent note that does not sound like every other one."
6. "Write a note referencing that we are both in {niche group} without naming a product."
7. "They posted a take I disagree with politely. Write a note that opens a respectful discussion."
8. "Rewrite this draft to sound like a text message, not a press release: {paste draft}."
9. "Make this note 40% shorter and remove every adjective: {paste draft}."

How do you prompt ChatGPT for warm and event-based requests?

For warm requests, the prompt should anchor on the shared context so ChatGPT does not invent a relationship that does not exist. Warm and event-based notes accept at a higher rate because the relevance is real, not manufactured, so the job is to name it cleanly.

10. "We met at {event}. Write a note that references one thing we actually discussed: {detail}."
11. "We are both attending {conference}. Write a note suggesting we connect before it."
12. "They commented on my post about {topic}. Write a note that continues that exact thread."
13. "{mutual} introduced us over email. Write a short note that references the intro."
14. "They downloaded my {lead magnet}. Write a note that follows up without selling."
15. "We worked at {company} years apart. Write a note that uses the alumni angle."
16. "They asked a question in {group} that I can answer. Write a note offering the answer."
17. "They liked three of my recent posts. Write a note that acknowledges it lightly, not creepily."
18. "Write a note to a podcast guest I just heard, referencing one specific thing they said."

Event-based notes carry the highest natural relevance, which is why they belong in your highest-priority queue. For a head-to-head on when a note even helps versus going straight to InMail, see InMail vs connection request in 2026.

How do you prompt ChatGPT to personalize at scale without going generic?

To personalize at scale, you give ChatGPT a template skeleton with bracketed variables and one strict rule: every note must change at least one full clause, not just swap a name. Mail-merge personalization (just the first name) is the single biggest reason "personalized" outreach reads as spam.

19. "Here is a template and 5 prospect profiles. Rewrite the template per profile so each opening clause is genuinely different: {paste}."
20. "Generate 10 variants of this note where only the specific reference changes and the rest stays human: {paste}."
21. "From this prospect's About section, extract the one detail most worth referencing, then write the note: {paste}."
22. "Score these 5 drafts 1-10 on how generic they sound and rewrite the lowest two: {paste}."
23. "Flag any sentence in these notes that could apply to anyone, and replace it: {paste}."
24. "Turn this case study into a one-line, no-pitch reference I can drop into a note: {paste}."
25. "Write a note that references their company's recent {funding, launch, or hire} from this press snippet: {paste}."
26. "Convert this voice memo of why I want to connect into a 280-character note: {paste transcript}."
27. "Audit this note for AI tells and rewrite the ones that scream ChatGPT: {paste}."

Scaled personalization lives or dies on the "one full clause must change" rule. Without it, ChatGPT produces ten notes that share an identical skeleton, and prospects pattern-match that skeleton in two seconds. Volume discipline matters just as much, because the math compounds against you fast: read what 1,000 connection requests really get you before you decide to scale, and what to do when you hit the connection limit if you already have.

How do you keep a ChatGPT note from reading as AI-written?

You keep it human by editing one detail back in after ChatGPT drafts, because the model cannot supply a genuine opinion or a real memory. The tells are consistent: over-polished cadence, the phrase "I came across", a symmetrical two-sentence rhythm, and a closing that softly pitches. Recipients have learned all four, and an obviously AI note now signals low effort, which is the opposite of what a note is for.

The workflow that holds up: prompt ChatGPT for three variants, pick the least generic, then change one thing only a human would know or feel. Replace "I enjoyed your post on retention" with "your line about churn surveys lying to you stuck with me." That single edit is what survives the human filter. Run the read-aloud test on every note, because anything you would not say out loud to a stranger reads as automated on the screen too. The deeper point: ChatGPT is a first-draft engine, not a send button. Treat it that way and acceptance holds; automate the send and skip the edit, and you are back to generic.

FAQ

Can ChatGPT write LinkedIn connection requests that actually get accepted?

Yes, when you feed it the prospect's real context and edit one human detail back in. ChatGPT's raw, context-free output reads generic and gets declined, so the prompt and the post-edit do the real work.

How do you keep a ChatGPT note inside the 300-character limit?

Put the cap inside the prompt itself ("under 300 characters") rather than trimming afterward, because the constraint forces ChatGPT to cut its usual throat-clearing. A note that fits in two sentences accepts better than one that fills the whole field.

Should I send a note at all, or connect with no message?

It depends on relevance, and the research is genuinely mixed, so test both. A specific, low-pitch note tends to help with cold decision-makers, while a no-note request can outperform a bad note; the trade-off is laid out in InMail vs connection request in 2026.

Will sending AI-personalized requests at volume get my account restricted?

Volume and method are the risk, not AI drafting. Browser-automation tools that scrape and blast trigger detection, while verified-API platforms calibrated to about 25 invites a day stay inside LinkedIn's tolerances, which is why send discipline protects acceptance and the account alike.

Sources

• Reachium
• LinkedIn Help: Personalize an invitation to connect
• LinkedIn Sales Solutions: State of Sales
• LinkedInsider: LinkedIn Outreach Benchmarks 2026
• LinkedInsider: the LinkedIn connection request note
• LinkedInsider: why you should stop sending 100 requests a day