I’m getting ready to send my first batch of outreach, and I keep running into the same problem: the messages I’m generating feel… fine, but not great. Like, they’re personalized (the AI pulls in their job title, company, recent activity), but there’s still something that feels a bit stiff about them. Not quite robotic, but not quite like something I’d actually write either.
I think I understand the idea of hyper-personalized AI messaging, but I’m struggling with the execution. Are the best templates super short and casual? Do I need to vary my tone dramatically between prospects, or is there a baseline tone that just works? And honestly, is there a difference between a message that looks personalized and a message that actually sounds like I’m a real person who cares?
I’ve tried a few iterations—some messages reference the prospect’s recent LinkedIn activity, some mention their company’s industry, some have a subtle ask at the end. But when I read them back, I can’t quite tell if they’re hitting that sweet spot of being authentic without sounding like I’m forcing a relationship.
How do you actually write AI-powered messages that don’t feel like a bot wrote them, even though technically an AI is helping? Are there any tricks for keeping the tone believable while still hitting the personal touches that make opens and replies actually happen?
This is the million-dollar question, and honestly, most people get it wrong. The issue isn’t that AI personalization sounds fake—it’s that most people use it as a crutch to avoid doing the real work of writing a compelling hook.
Here’s what separates a message that feels human from one that feels robotic: specificity + brevity + a reason for reaching out that benefits them, not you.
Robotic: “Hi [Name], I noticed you work at [Company]. We help companies like yours scale their outreach. Let’s connect.”
Human: “Hey [Name], saw you recently posted about [specific challenge]. We helped [similar company] solve this in 3 weeks—might be worth a 15-min chat?”
The difference is the hook. The AI can pull in their company, title, and activity, but YOU need to do the hard part: explain why you’re reaching out specifically to them today. Not “we help people like you” generic. Specific.
My advice: write 3-5 different message frames for different scenarios (e.g., prospects who recently changed jobs, prospects posting about growth challenges, executives in high-growth companies). Then let the AI fill in names and specific details. But the core message—the hook, the ask, the beneficiary—that’s you.
Do that, and it’ll read human every single time. Mess it up, and even the best personalization AI can’t save you.
What kind of prospect are you reaching out to first? That’ll help me suggest a better frame for your hook.
One more thing: short is always better than long. Most marketers overthink this. A 2-3 sentence message that hits the hook is better than a 5-sentence message that sounds more polished. Humans are lazy readers, especially on LinkedIn. Respect their attention.
Also, tone should stay consistent. Don’t write formal for one person and super casual for another. Pick a voice (I recommend friendly but professional—a tone you’d actually use if you called them), and stick with it. The personalization comes from the research, not from changing how you sound every message.
In recruiting, this is everything. I’m reaching out to senior engineers and CTOs, and if my message sounds like a bot, I get 0% response. If it sounds human, I get 15-20% response. That’s the gap.
What works for me: I use LiSeller’s AI to pull in one or two details about their background (usually their last job, a speaking engagement, or a GitHub project), and then I write a hook that’s genuinely respectful of their time. Something like: “Hey [Name], I saw you led [specific project] at [company]—that’s the kind of system design we’re looking for. Would be curious to chat, but no pressure if recruiting isn’t your thing right now.”
The magic is in that last part—“no pressure if recruiting isn’t your thing.” It tells them you see them as a person, not a target. That psychological move alone makes the message feel human, even though the AI helped me find the person in the first place.
For your situation: let the AI do the research and pull the details. Then write your own hook that shows you actually understand why you’re reaching out to them, not to anyone else with their job title. The difference is respect, and respect reads as human.
Okay, so I’ll be real with you: I was overthinking this way too much at first. I was trying to make every message feel like I’d spent 30 minutes researching the person, which is insane.
What actually works: I write my core message once. It’s personal but not overly intimate. Then LiSeller’s AI fills in the variable spots (name, recent activity, company)—but I keep the tone and the ask the same because consistency is what makes it feel real.
People can smell it when you’re trying too hard to be their friend. So I just ask a genuine question or mention something I genuinely can help with, keep it short, and let that be enough. Like: “Hey, interesting angle on [their recent post]. I’ve worked with some companies doing similar—would be worth 15 mins?”
That’s it. Not robotic, not best-friends energy either. Just a real person saying “I see you, I might be able to help, let’s talk.”
I tested like 10 different versions of my hook, and the simplest one got the best replies. The moral: less is more, and “real” usually means “not trying too hard.”
How many connection requests are you planning to send in your first week? That might change how much you need to vary your messaging.
Great question, and I see this challenge come up all the time. The truth is, LiSeller’s hyper-personalized AI can pull in details about someone’s background, recent activity, and company, but the actual tone and hook—that’s all you.
Here’s how I think about it: the AI is your research assistant, not your copywriter. Let it find the insights (“Oh, they just got promoted to VP” or “This person shares content about [topic]”), and then you write a message that’s genuinely you.
One practical tip: in LiSeller, you can set up message templates with placeholders. So you write something like: “Hey [Name], saw your post on [topic]—we actually built something similar for [company type]. Would be interesting to get your take on it.” Then the AI fills in the variables in a way that feels natural, not like a mail merge.
The key settings to play with: Keep your message variable (the part that changes per person) to one or two points max. More personalization can actually feel less human because it looks like you’re trying too hard to be relevant. One well-placed reference beats three mediocre ones.
Try this: write 3 different core messages (3 different approaches or angles), test them in small batches, and see which one gets the best response rate. The one that wins is probably the one that sounds most like you.
This is a conversion psychology problem, not a personalization problem. Here’s the framework:
The Human-Sounding Message Formula:
- Reference — One specific thing about them (their job, a recent post, their company’s growth)
- Relevance — Why you’re reaching out to them specifically (not “I help people like you”)
- Ask — What you want them to do, clearly and briefly
Robotic happens when you skip step 2. You reference something and make an ask, but you don’t explain why those two things connect. That gap is where it feels like a bot.
Human sounds like: “Hey [Name], I saw you recently joined [Company] in an ops role. Given their expansion into [market], I’ve actually worked with 3 competitors on resource planning—curious if it’s on your radar?”
Robotic sounds like: “Hey [Name], nice to connect. We help ops teams scale. Keen to learn about what you’re building.”
One is specific and relevant. One is generic and looks like it was copy-pasted.
The AI personalization should give you the raw data (their background, their company’s moves, their content). Then your job is to connect those dots in a way that sounds like a human who’s been thinking about their world. That’s the secret.
Are you planning to use different message templates for different prospect segments, or one universal template you’re tweaking slightly?
From a safety standpoint, here’s what matters: messages that sound human are less likely to trigger LinkedIn’s spam filters. Robotic messages, even if they’re technically personalized, get flagged because the pattern looks repetitive.
So psychologically, being authentic actually improves your account safety too. Win-win.
My advice: keep your personalization genuine and light. LinkedIn’s system can detect when you’re over-personalizing (too many specific references in a short message), and it actually can look suspicious. So genuinely referencing one thing about them—their recent job change, a company growth announcement, or something they posted—is plenty.
From there, just be casual and direct. “Hey [Name], quick question about [topic]—would be good to get your thoughts.” That reads human, it’s safe, and it works.
The trap a lot of people fall into: over-personalizing to compensate for the system feeling cold. Don’t. Trust that one genuine reference + a real question + brevity = they’ll respond if it’s relevant.