The AI DM That Backfired
Jordan was excited about using AI to write LinkedIn messages. He'd read articles about how ChatGPT could "10x your outreach." So he fed the AI some basic context about his product and asked it to write a DM to a potential client.
The AI delivered a perfectly structured message. Professional tone. Clear value proposition. Zero typos. Jordan copied it, pasted it, hit send, and waited for the response.
It never came.
When Jordan looked back at the message, he saw the problem immediately. It was too polished. Too formal. It sounded like a marketing email, not a human conversation. The recipient probably read the first line, thought "automated outreach," and deleted it without a second thought.
Here's the thing: AI is incredible at drafting LinkedIn DMs—but terrible at making them sound like you. It doesn't know your quirks, your conversational rhythms, or the subtle ways you build rapport. It writes in corporate-speak unless you teach it otherwise.
The good news? You can teach it. With the right prompts and the right edits, AI becomes a powerful drafting tool that saves you time without sacrificing authenticity. In this guide, you'll learn prompt frameworks that work, see five real AI-assisted drafts (with the human edits that fixed them), and discover how to use AI ethically without sounding like a robot.
Why AI-Generated DMs Sound Robotic (And How to Fix It)
AI language models are trained on billions of text examples—most of which are formal, corporate, or generic. So when you ask an AI to write a LinkedIn message, it defaults to what it's seen most: bland, safe, corporate language.
Here's what typically happens:
- It over-explains: AI loves to be thorough, which means it often writes 3 sentences when 1 would do.
- It uses formal language: Phrases like "I hope this message finds you well" or "I would be delighted to connect" sound like they're from a 1990s business letter.
- It lacks personality: AI doesn't know your sense of humor, your catchphrases, or how you naturally build rapport.
- It misses context: Unless you give it very specific details, it makes generic assumptions about your prospect and your offer.
The fix? Better prompts + human edits. Think of AI as a first draft generator, not a final copy writer. You provide the structure and constraints upfront, then you refine the output to sound like you.
This is the same balance required in all authentic LinkedIn relationship-building: efficiency matters, but so does genuine human connection.
The 5-Part Prompt Framework for Better AI DMs
The quality of your AI output depends entirely on the quality of your prompt. Here's a 5-part framework that consistently produces usable drafts:
Part 1: Context (Who You Are & Why You're Reaching Out)
Give the AI the essential background: who you are, what you do, and why this person is relevant to you.
Example:
"I'm a marketing consultant who helps SaaS companies with content strategy. I'm reaching out to [Name], a VP of Marketing at a mid-size SaaS company, because they recently posted about struggling to scale content without burning out their team."
Part 2: Goal (What You Want From This Message)
Be specific about the outcome. Do you want a response? A meeting? A simple acknowledgment?
Example:
"The goal is to start a conversation and eventually get a 15-minute call to discuss their content challenges. But this first message should just spark interest, not ask for a meeting yet."
Part 3: Constraints (Length, Tone, and Format Rules)
This is where you prevent the AI from going off the rails. Set hard limits on length, tone, and structure.
Example:
"Keep it under 60 words. Use a casual, conversational tone—like I'm messaging a colleague, not writing a cover letter. No corporate jargon. No phrases like 'I hope this finds you well' or 'I would love to connect.' Start with a reference to their recent post, then briefly relate it to my work, then ask one question."
Part 4: Voice Tokens (Words & Phrases You Actually Use)
Teach the AI your voice by giving it examples of how you talk. This is the secret to making AI sound like you.
Example:
"Use phrases like 'I've been thinking about this too' or 'Out of curiosity' or 'This resonated with me.' Avoid words like 'delighted,' 'thrilled,' 'synergy,' or 'value proposition.' Write like I'm texting a friend, not drafting a press release."
Part 5: Specific Details (Personalization Points)
Finally, give the AI the unique details that make this message relevant to this specific person.
Example:
"Reference their post from last Tuesday about content burnout. Mention that I've worked with teams facing the same challenge. Ask how they're currently thinking about scaling content without adding headcount."
Full Prompt Example:
"I'm a marketing consultant who helps SaaS companies with content strategy. I'm reaching out to Sarah, a VP of Marketing at a mid-size SaaS company, because she recently posted about struggling to scale content without burning out her team. The goal is to start a conversation and eventually get a 15-minute call, but this first message should just spark interest. Keep it under 60 words. Use a casual, conversational tone—like I'm messaging a colleague, not writing a cover letter. No corporate jargon. Use phrases like 'I've been thinking about this too' or 'Out of curiosity.' Reference her post from last Tuesday about content burnout. Mention that I've worked with teams facing the same challenge. Ask how she's currently thinking about scaling content without adding headcount."
When you structure your prompts like this, you get drafts that are 70-80% ready to send—instead of 30% ready and full of cringe-worthy corporate speak.
5 AI-Assisted DM Drafts (With the Human Edits That Fixed Them)
Let's look at real examples. Each one shows the AI's first draft, the problems with it, and the human edits that made it sound authentic.
Example 1: Following Up on a Post About Content Burnout
AI Draft:
"Hi Sarah, I came across your recent post about content burnout and found it very insightful. I work with SaaS marketing teams on similar challenges and would be interested in learning more about your approach. How are you currently thinking about scaling content production without adding headcount? Looking forward to hearing your thoughts."
Problems:
- Too formal ("found it very insightful," "would be interested in learning")
- Reads like a template, not a real message
- "Looking forward to hearing your thoughts" is cliché
Human Edit:
"Hey Sarah—your post on content burnout hit home. We've been working with a few SaaS teams on exactly this, and it's such a common trap. Out of curiosity, how are you thinking about scaling content without just piling more onto your team?"
Why it works: Shorter, conversational, and personal. "Hit home" is more human than "very insightful." The question feels natural, not scripted. This mirrors the framework for strong first LinkedIn DMs: context, relevance, genuine curiosity.
Example 2: Reaching Out After a Job Change
AI Draft:
"Hi Michael, I noticed you recently started a new role as Head of Sales at Acme Corp—congratulations! I work with sales leaders to optimize outbound strategies and thought it might be valuable to connect. Would you be open to a brief conversation about your current priorities in this new role?"
Problems:
- "Congratulations!" feels obligatory, not genuine
- "Thought it might be valuable to connect" is vague and salesy
- Jumps straight to asking for a call without building rapport
Human Edit:
"Hey Michael—saw you just joined Acme as Head of Sales. That's a great move. I'm curious, what are the first big challenges you're tackling? We work with sales teams on outbound strategy, so I'm always interested to hear how leaders are thinking about it in new roles."
Why it works: More conversational, less transactional. "That's a great move" feels warmer than "congratulations." The question is about their priorities, not your offer. This approach aligns with how to reconnect authentically without sounding opportunistic.
Example 3: Following Up After a Webinar
AI Draft:
"Hi Laura, I attended your webinar yesterday on building high-performing teams and found your insights on psychological safety particularly compelling. I'd love to continue the conversation and explore how these principles might apply to remote-first environments. Would you be available for a quick call?"
Problems:
- "Particularly compelling" is overly formal
- "Explore how these principles might apply" sounds like a research paper
- Asks for a call too quickly without adding value first
Human Edit:
"Hey Laura—really enjoyed the webinar yesterday. Your point about psychological safety in high-pressure teams stuck with me, especially since we're navigating that with a fully remote team right now. Have you seen any differences in how that plays out remote vs. in-person?"
Why it works: Natural language ("stuck with me" vs. "compelling"). Shares context about why it's relevant to them. Asks a curious question instead of pitching a call. This is how event follow-ups should feel: genuine, not transactional.
Example 4: Responding to a Feature Request Post
AI Draft:
"Hi David, I saw your post about needing better reporting features in your CRM. This is a common pain point we hear from customers. Our platform offers advanced reporting capabilities that might address your needs. I'd be happy to show you a quick demo if you're interested."
Problems:
- Immediate product pitch feels pushy
- "I'd be happy to show you a demo" is classic sales-speak
- No actual engagement with their specific challenge
Human Edit:
"Hey David—saw your post about CRM reporting. We've heard that from a lot of teams lately. Out of curiosity, what's the reporting gap that's causing the most friction for you? Always trying to understand what's missing in the tools out there."
Why it works: Focuses on understanding, not selling. "Out of curiosity" signals you're listening, not pitching. Leaves room for dialogue instead of pushing a demo. When the time is right for an offer, you can reference smooth ways to propose meetings.
Example 5: Cold Outreach Based on Shared Interest
AI Draft:
"Hi Rachel, I noticed we both follow the same thought leaders in the growth marketing space, and I appreciate your commentary on recent trends. I'm working on some innovative strategies for B2B SaaS growth and thought you might find them interesting. Would you be open to connecting?"
Problems:
- "Innovative strategies" is vague and buzzwordy
- "Thought you might find them interesting" assumes interest without evidence
- Ends with a weak "Would you be open to connecting?" that has no clear benefit
Human Edit:
"Hey Rachel—I keep seeing your comments on [Thought Leader]'s posts, and they're always spot-on. Curious: what's your take on [specific trend]? We're experimenting with something related and I'd love to hear how others are thinking about it."
Why it works: Specific reference to their activity. Genuine question, not a pitch. "I'd love to hear how others are thinking about it" invites conversation without asking for anything. This is the essence of getting tone right in DMs.
Ethical Considerations: When AI Helps vs. When It Hurts
Using AI for LinkedIn DMs isn't inherently unethical—but it can cross the line if you're not thoughtful about it. Here are the rules:
Do: Use AI as a Draft, Not a Send-and-Forget Tool
AI should speed up your process, not replace your judgment. Always review, edit, and personalize before sending. If you're copy-pasting AI output without reading it, you're doing it wrong.
Don't: Use AI to Mass-Blast Generic Messages
AI makes it easy to crank out dozens of messages quickly. But volume without relevance is spam. If your message could be sent to 100 people with only the name changed, it's not personalized—it's automated noise.
Do: Use AI for Structure, Then Add Your Voice
AI is great at organizing thoughts and suggesting frameworks. Use it to get the bones of the message right, then add your tone, personality, and specific observations.
Don't: Pretend AI-Generated Insights Are Your Own Research
If the AI suggests a pain point or trend you didn't actually research, don't use it. Authentic outreach requires real knowledge of the person and their work. This connects to ethical LinkedIn outreach principles: respect, relevance, and honesty.
Frequently Asked Questions
Which AI tools are best for writing LinkedIn DMs?
ChatGPT, Claude, and Jasper all work well if you use detailed prompts. The tool matters less than the prompt framework and your willingness to edit. Avoid tools that promise "one-click personalized outreach"—they're usually glorified mail merge.
How much should I edit AI-generated drafts?
Plan to edit 30-50% of the AI's output. You're refining tone, cutting fluff, adding personal touches, and ensuring the message sounds like you. If you're only changing a name or two, the message is probably too generic.
Can recipients tell when I've used AI?
If you don't edit it, yes—they absolutely can. AI has telltale signs: overly formal language, generic phrasing, perfectly structured sentences. But with good prompts and human edits, it's nearly impossible to tell. The goal isn't to hide that you used AI; it's to make sure the final message sounds authentically human.
Should I tell people I'm using AI to write messages?
You don't need to disclose it, but you also shouldn't misrepresent yourself. If someone asks, be honest: "I sometimes use AI to organize my thoughts, but I always personalize and review before sending." Transparency builds trust.
Next step: Write better DMs faster, without losing your voice — Try ANDI Free.
