Everyone on LinkedIn has had the obviously automated connection message: generic praise, a pitch by the third line, nothing that couldn't be sent to anyone. This guide is about doing outreach that doesn't read that way — using AI to speed up the research while you write and send the messages yourself. It also covers what LinkedIn's own terms permit, because the "set it and forget it" auto-senders sold for this are against the rules.
Deciding who to contact and whether they can pay is a separate step — that's covered in our guide to finding higher-paying clients with AI. This guide picks up once you have a shortlist.
What LinkedIn's rules actually allow
LinkedIn's User Agreement (section 8.2, "Dos and Don'ts") is specific. You agree not to:
- "Develop, support or use software, devices, scripts, robots or any other means or processes (such as crawlers, browser plugins and add-ons or any other technology) to scrape or copy the Services";
- "Use bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages … or otherwise drive inauthentic engagement";
- Create a false identity or misrepresent who you are.
In practice: the third-party tools that auto-send connection requests and messages fall under this, and relying on them puts your account at risk of restriction. Separately, LinkedIn limits how many connection invitations you can send in a week, and it has changed that number over time — check the current limit in the LinkedIn Help Center rather than assuming a figure from an older article. The rest of this guide assumes you're sending manually.
Why generic outreach fails
The tell isn't that a message was automated — it's that it's generic. A message that could go to anyone in a role reads as spam the moment it's opened, whether a person or a script produced it. So the goal isn't "make automation sound human," it's "make every message specific enough that it obviously isn't a template."
An AI-assisted workflow that stays personal
- Use AI for research, not for the message. For each person on your shortlist, gather two or three concrete, checkable facts: their most recent post, a specific line in their profile, a piece of recent company news. An AI assistant can speed up pulling this together — but verify each fact against the actual profile or source before you use it.
- Write the message yourself from those facts. Not a template with the name swapped in. Reference one real, specific thing, say briefly why you're reaching out, and make a small, low-pressure ask.
- Keep the volume low. A smaller number of genuinely researched messages is more effective than a high volume of near-identical ones, and it keeps you well inside LinkedIn's invitation limits. Ten to twenty considered messages a week is a sustainable pace for a solo freelancer.
Worked example. Say a profile shows: "Head of Marketing at a B2B SaaS company," a recent post about struggling to keep a blog publishing schedule, and a company page noting a new product line. A message built from that:
"Saw your post about the blog schedule slipping — that's usually the first thing to go when a team's shipping a new product line. I'm a freelance writer who works with B2B SaaS marketing teams on exactly that gap. Not pitching a package here, just curious whether it's a hiring problem or a capacity one for you right now."
It names a real detail, connects it to a real service, and asks a question instead of pushing a sale. That's the standard every message should meet.
Connection request, InMail, or content-led
- Connection request with a note: the default for most freelancers. Keep the note short and specific. Free, but subject to the weekly invitation limit.
- InMail: lets you message people you're not connected to, on a paid LinkedIn plan with a monthly credit allowance. Worth it if your targets rarely accept cold connection requests.
- Content-led inbound: posting useful, specific content in your niche so the right people reach out to you. Slower to start, but the highest-quality conversations, and no volume limit.
Follow-up cadence
If there's no reply, one short follow-up after about a week is reasonable — add a new specific detail or a lighter ask rather than repeating the first message. After that, stop. A third or fourth unanswered message is where outreach tips into feeling like spam.
What never to automate
The conversation after someone replies. The value of your first message was that it showed real attention; a templated reply throws that away immediately. These are the exchanges most likely to become clients, so give each one a genuine, individual response.
Measuring it
Over your first month, track response quality more than response rate. Silence or short brush-offs mean the messages still read as generic — tighten the specificity. Replies that ask a follow-up question or suggest a call mean it's working, even if the count looks modest. If response quality stays low after a few weeks of genuinely specific messages, the problem is probably who you're contacting, not how — go back to targeting and qualification.
Frequently asked questions
Are LinkedIn automation tools against the rules? Tools that auto-send connection requests or messages fall under the User Agreement's prohibition on bots and automated methods (section 8.2). Using them risks account restriction.
How many connection requests can I send per week? LinkedIn caps this and has changed the number over time. Check the current figure in the LinkedIn Help Center.
Should I use InMail or free connection requests? Start with connection requests and a short note. Move to InMail (a paid plan) only if your specific targets rarely accept cold requests.
The "don't sound like a bot" principle applies to content too — see our guide to writing SEO blog posts without sounding robotic.