How to Send AI Voice Notes on LinkedIn With Lemlist (Step-by-Step)

To send AI voice notes on LinkedIn with lemlist, add an AI Voice Message step to a campaign, clone your voice in about two minutes using the built in ElevenLabs integration, write a short script with personalization variables, connect your LinkedIn account under Settings, and set the note to fire after a prospect accepts your connection. Cloning is free, and each sent voice note costs 5 credits. The message lands as a normal voice note in the prospect's LinkedIn inbox on desktop and mobile.

Cloned voice notes are one of the sharpest outbound tactics to appear in 2026 because they solve the two things that killed manual voice outreach: it did not scale and most reps hated recording. lemlist's own reporting puts LinkedIn voice messages at roughly twice the reply rate of plain text, and its partner ElevenLabs grew outbound from about 5 percent to 30 percent of pipeline running this exact motion. At DevCommX we treat features like this as one channel inside a system you own, not a growth hack you rent, which is the same lens we bring to any B2B outbound automation stack. This guide walks the full setup, then the guardrails that decide whether it books meetings or gets your account restricted.

Why AI voice notes are the 2026 outbound tactic to watch

Text outreach on LinkedIn has flatlined. Everyone is running the same connection request, the same follow up, the same slightly reworded value proposition, and prospects have learned to swipe past all of it. A voice note breaks that pattern because it is rare, it is personal, and it forces the reader to stop and listen for fifteen seconds instead of skimming. The problem was always that recording individual voice notes for hundreds of prospects is impossible for one rep and inconsistent across a team.

Voice cloning removes that ceiling. You record your voice once, then a single script with dynamic variables becomes hundreds of individually addressed notes that still sound like you. lemlist reports overall engagement lifting by up to 20 percent among early adopters, and case studies of high ticket B2B teams show meeting booking rates moving from single digits into the high teens once a personalized audio touch is added mid sequence. The tactic is hot precisely because it combines the intimacy of voice with the reach of automation, which used to be a contradiction.

There is a timing advantage too. Right now voice notes are still novel enough on LinkedIn that most inboxes have never received one, so the pattern interrupt is real. That window will not stay open forever. As more teams adopt the motion, the bar for what counts as a good note rises, and the sloppy ones that read a generic script in a flat clone will start to hurt sender reputation rather than help it. The teams that build the discipline now, tight scripts and a clean clone, are the ones who keep the reply lift as the tactic matures.

What you need before you start

Before you touch a script, confirm the plumbing. AI voice messages are gated to specific lemlist plans, they consume credits per send, and they only work once your LinkedIn account is properly connected. The table below is the setup at a glance so you know exactly what the feature costs and where it can break.

Step 1: Clone your voice

Open any campaign and add an AI Voice Message step from the step picker. Inside it, click Record with AI, then Clone your voice. lemlist shows a short set of recording guidelines, and they matter more than they look. Find a genuinely quiet room, kill any fan or air conditioner noise, and get close to a decent microphone. The quality of this single recording sets the ceiling for every note you send, so it is worth two minutes of care.

You will read a provided practice script aloud, then record the full version, aiming for somewhere between ninety seconds and two minutes of clean audio. Read the way you actually talk, with natural pace and a little warmth, because the clone captures your rhythm as much as your timbre. When you click Clone your voice, lemlist generates the ElevenLabs model in seconds. Preview it, and be honest about whether it sounds like you. If it sounds flat or robotic, re record rather than shipping it, since a slightly off voice is more damaging than a text message. Once you are happy, name the voice and click Use my cloned voice. Cloning costs nothing, so you can iterate until it is right.

If you would rather not use your own voice at all, lemlist also offers high quality prebuilt voices. For real outbound I would avoid them. The entire advantage of this tactic is that the prospect believes a specific human recorded a note for them, and a generic stock voice quietly breaks that illusion the moment someone who has seen you on a call hears it.

Step 2: Add the voice note step to a LinkedIn sequence

The most important sequencing fact is the one people miss: LinkedIn only allows voice notes to first degree connections. You cannot open with one, and you cannot send one inside an InMail. That shapes the entire sequence. The voice note has to sit after a connection has been accepted, which means your structure looks like a connection request first, a short wait, then the AI voice message as a warmer second or third touch.

In practice, the highest converting placements are the moments where a human would naturally reach out by voice. The three that consistently work are a note on day two after a connection accepts but has not replied, a note right after someone engages with your content, and a note that follows a soft demo or call invitation. Set the delay on the step to match that logic, for example two days after the connection is accepted, rather than firing the instant it lands, which feels automated.

Confirm your LinkedIn account is connected under Settings and Integrations before you launch, and keep an eye on it, because a dropped LinkedIn session is the single most common reason voice steps silently stop sending. The note itself arrives as an ordinary voice message in the prospect's LinkedIn inbox on both desktop and mobile, indistinguishable from one you thumbed out by hand. This kind of trigger based placement is exactly what a contextual outreach playbook built on buying signals is designed to systematize, so the voice note fires on a real signal instead of a static schedule.

Step 3: Personalize the script with dynamic variables

A cloned voice reading a generic script is just a robocall with better production values. The personalization is what earns the reply. lemlist lets you drop variables like {{firstName}}, {{companyName}}, and {{jobTitle}} straight into the voice script, and its AI variables can generate an icebreaker that references a recent post, a mutual connection, or a company detail. One script then becomes hundreds of individually tailored notes that still sound spontaneous.

The structure that works is trigger, question, teaser, then call to action. Open with the specific observation that justifies the message, ask a genuine question, hint at the value you can offer, and close with a low friction ask. Something like: noticed you are scaling {{companyName}}'s outbound team fast, curious how you are handling reply rates right now, we have a way to lift them without adding headcount, worth a quick swap of notes. That is roughly twenty seconds spoken, it references something real, and it does not pitch.

Voice cloning supports scripts in English, French, Spanish, German, Danish, Dutch, Portuguese, Italian, and Greek, so a single cloned voice can run localized campaigns without you speaking the language. Preview every variant before launch, because a variable that renders awkwardly in text sounds far worse read aloud. The goal is a note that could plausibly have been recorded by hand, which is a bar we unpack in our guide on how to make AI outbound feel human.

Deliverability, limits, and authenticity guardrails

The audio is the easy part. The way to get burned with this tactic is LinkedIn's automation limits, not the voice technology. LinkedIn watches for sudden spikes in activity, and voice notes count as actions like any message. Keep total daily actions under roughly one hundred across visits, invites, and messages combined, start conservative at around fifty a day on a newer account, and ramp slowly. Note also that LinkedIn now caps custom noted connection invites at three per month, so lean on standard invites without a note to keep your connection engine running while you save personalization for the voice step.

The authenticity guardrail is simpler and non negotiable: only clone a voice you own. Cloning your own voice with consent is fine and is exactly how lemlist and ElevenLabs designed the feature. Cloning anyone else's voice, or having one rep's clone speak as a colleague, crosses a line that is both a policy problem and a trust problem the first time a prospect notices. If someone hears your stock or borrowed voice on a call after receiving a note, the credibility you built evaporates.

One operational detail worth knowing. If you delete a campaign that uses a cloned voice, lemlist enforces a seven day waiting period before you can remove the voice clone, and the campaign must not be actively using it. Plan clone management around that so you are not blocked mid cleanup. Treat the whole setup like infrastructure: monitor the LinkedIn connection, respect the action limits, keep one clean clone per real sender, and the channel stays healthy instead of getting your account flagged.

Make voice one owned channel, not a rented hack

AI voice notes are a genuinely strong tactic, but a tactic is not a system. The teams that win with this are the ones who wire voice into a signal based outbound engine they own, so the note fires when a prospect actually does something worth reacting to, and the whole sequence is monitored rather than left to drift. That discipline is what takes a fresh setup to more than forty qualified demos in around six weeks, because every touch, voice included, triggers on a real buying signal instead of a static list. If you want to build voice into outbound infrastructure your team owns end to end, book a GTM strategy call and we will map it to your pipeline.

Further Reading

• lemlist Help Center: Send AI Voice Messages on LinkedIn
• ElevenLabs Docs: Voice Cloning Guide
• LinkedIn Help: Send and Receive Voice Messages

FAQ

How do I set up AI voice notes on LinkedIn with lemlist?

Open a campaign, add an AI Voice Message step, then click Record with AI and Clone your voice. Read the two minute practice script aloud in a quiet room, generate the clone, name it, and confirm it sounds natural. Connect your LinkedIn account under Settings and Integrations, write a short script with variables, set the delay after connection, and launch. Cloning is free and each sent voice note costs 5 credits.

How long does it take to clone your voice in lemlist?

Voice cloning takes about two minutes. You record yourself reading the provided script for roughly ninety seconds to two minutes in a quiet space with minimal background noise, then lemlist generates the AI voice model using ElevenLabs. You preview the result, name the voice, and it is ready to use across campaigns. A rushed or noisy recording is the most common reason a clone sounds robotic, so it is worth a clean take.

Do you need to be connected on LinkedIn to send a voice note?

Yes. LinkedIn only lets you send voice notes to first degree connections, so your sequence has to earn the connection first. A typical structure is a connection request, a short wait after it is accepted, then the AI voice message as a later touch. You cannot fire a voice note at a cold profile or inside an InMail, which is why voice works best as a second or third step rather than an opener.

How long should a lemlist AI voice note be?

Keep it to fifteen to twenty five seconds. That is long enough to reference something specific about the prospect and make one clear point, and short enough that it feels like a quick personal note rather than a monologue. A useful structure is trigger, question, teaser, then call to action. Longer messages get skipped, and anything over about forty five seconds reads as a pitch.

Is using AI voice cloning for LinkedIn outreach against the rules?

Cloning your own voice with consent is allowed, and lemlist voice notes appear as normal voice messages in the LinkedIn inbox. The real risk is LinkedIn automation limits, not the audio. Stay under roughly one hundred actions a day, keep connection requests conservative, and only clone a voice you own. The line you should never cross is impersonating someone else, which is both a policy and a trust problem.

How much do lemlist AI voice notes cost in credits?

Voice cloning itself is free. Each AI voice note you send costs 5 credits, and the feature is available on the Free Trial, Multichannel Expert, and Enterprise plans. Because a single cloned voice can read hundreds of personalized scripts, the cost per touch stays low even at volume, which is the whole point of cloning rather than recording each note by hand.

👉 Send AI Voice Notes