I’ve been running campaigns for about 3 weeks now, and I’m hitting a wall. My reply rates are stuck around 3-4%, which tells me something’s off with how my messages are landing. The thing is, I’m using LiSeller’s personalization features, pulling in details about prospects’ recent activity, their company changes, stuff like that. But when I read back through what’s actually being sent, there’s still this… cadence to it. It feels like someone wrote a formula and then filled in blanks.
I’m wondering if the issue is that I’m not giving the AI enough context about who I’m actually talking to, or if I’m overthinking the personalization and it needs to be more conversational and less “research paper.”
Have any of you found the sweet spot between looking researched and actually sounding like you wrote it naturally? Or is there a specific way you’re structuring your message prompts in LiSeller to get it to feel less… template-y?
This is the most common mistake I see, and honestly it’s not about the AI—it’s about what you’re feeding it. Most people throw in way too much research. “I noticed you were promoted to VP of Sales in March, and I saw your company just opened an office in Austin…” That’s not personalization, that’s detective work. Prospects can smell it instantly.
The hook needs to be about them or their problem, not your homework. Try this: lead with ONE relevant observation that connects to what you actually do. Not five. Then ask a genuine question. The AI will sound way more natural if you’re not asking it to be a research machine. What specific result or pain do you think your prospect is experiencing? Start there, not with their LinkedIn timeline.
Have you looked at your message template structure in LiSeller? The way you’re prompting the AI makes a huge difference. If you’re using generic variable insertion like {{First_Name}} {{Company}} {{Recent_News}}, that’s going to feel stiff no matter how good the AI is.
Try feeding LiSeller a more human prompt: something like “Write this as if I’m texting a colleague—casual, specific, one main point.” Also, are you A/B testing different tone variations? Run one batch that’s conversational and short, another that’s slightly more formal. You might find one resonates way better. I pipe my top performers into HubSpot automatically, so I can segment by message tone and see which actually converts.
In tech recruiting, I’ve learned that “personalization” high-level candidates trust is super specific and brief. They don’t want you to prove you stalked their profile. They want to know why YOU think they’d care about what you’re offering.
Maybe try this: instead of listing facts about them, reference something they actually said publicly—a blog post they wrote, a comment on a post, a podcast they were on. That’s not creepy, it’s attentive. Then connect it to your value in one sentence. The AI will sound way more natural if the hook itself is strong and genuine.
I had this exact problem last month. Turned out I was trying to “perfect” every message and ending up with these long, overthought things. Switched to shorter, more casual openings—like 2-3 sentences max. Sounds dumb, but my reply rate jumped to 6.5%.
Also started reviewing my top performers and reverse-engineering what made them work. Usually it was something simple: they led with curiosity, not credentials. “Quick question—are you guys dealing with [specific problem]?” That’s it. Let me know if you want to share a few examples and I can give you honest feedback on where they’re probably tanking.
This is a critical insight: perceived authenticity in cold outreach isn’t about saying more, it’s about specificity. Generic research reads as robotic. Authentic outreach references one concrete thing and connects it to one concrete outcome.
The pattern I’ve seen across high-performing campaigns is this: (1) Reference something specific that took less than 30 seconds to find, (2) One sentence about why it matters to them, (3) One question. That’s it. The AI should be helping you execute that structure cleanly, not adding unnecessary detail.
What’s your current word count per message? If it’s over 150 words, I’d bet that’s where the robotic feeling is coming from.