Rémi KokabiNovember 17, 20267 min read

Answering machine detection in cold calling, how it works and when it helps

Your dialer connects a call. Is a person on the line, or is it a voicemail greeting about to eat 30 seconds of your rep’s time? Answering machine detection is built to answer that question before your rep says a word.

This matters more now that power and parallel dialers can fire off dozens of calls an hour. Without detection, reps burn through time listening to recordings instead of talking to prospects. With it, the dialer screens the pickup and only hands off live humans.

In this article, you’ll get a plain explanation of how answering machine detection works, why it pays off more with parallel and power dialing, and where it still gets calls wrong. You’ll also see how to handle a detected voicemail and how to wire detection into a calling sequence so your reps spend their hours in conversations, not greetings.

What answering machine detection is

Answering machine detection (AMD) is a feature built into outbound dialers that listens to the first seconds of audio after a call connects and decides whether a live person or a voicemail system picked up. The software analyzes patterns in speech, silence, and timing, then decides whether to connect the call to a rep, leave a voicemail, or end the call.

Here is how the classification works in practice. Voicemail greetings tend to have a consistent cadence of multiple words without long pauses, while a live human typically says a few words and then waits for a response. AMD algorithms can also process a live transcription and match keywords like “please leave your message after the beep” to categorize the call as unanswered.

Because the dialer needs a moment to listen and classify, there is a short pause before the rep hears anything. That pause, sometimes a second or two, is the tradeoff for filtering out machines. The rep does not dial, wait through a ring, and then sit through a 30-second voicemail greeting. Instead, the dialer handles that and only passes through calls where someone is on the other end.

Why it matters more with parallel and power dialing

When a rep dials one number at a time, hitting voicemail is annoying but manageable. When a power dialer auto-dials the next prospect the moment you hang up, or a parallel dialer calls up to five numbers at once, voicemails multiply fast. Without voicemail detection, a rep on a parallel dialer could connect to three voicemail greetings and zero humans in a single batch.

AMD changes the math. The dialer screens every line, drops the machines, and only routes live pickups to the rep. The result is more conversations per hour and less dead time.

lemlist internal data from parallel dialer beta teams (October to November 2026) shows the difference between dialing on three lines versus one:

Metric

3 lines (parallel)

1 line

Dials per hour

84

26

Conversations per hour

6

2

Meetings per 2-hour block

1.4

0.5

Calling time spent talking

58%

21%

Connect rate

13.1%

13.8%

The connect rate stays close because AMD filters voicemails on both setups. The gap is in volume: more dials per hour means more conversations, which means more meetings, without the rep sitting through recordings.

Where answering machine detection gets it wrong

AMD is not perfect, and the errors go in two directions.

False positives happen when the dialer classifies a live person as a voicemail. The call gets terminated and the live person hears nothing. That prospect picked up, heard silence, and hung up. You burned a dial and left a bad impression. Chasing ultra-low machine rates by shortening detection windows can spike false positives, where real prospects get disconnected during their greeting.

False negatives go the other way. The system incorrectly detects a live person when it is an answering machine, and the rep gets connected to a recording. This is less damaging. The rep hears a voicemail greeting, drops a message or hangs up, and moves on.

Short, customized voicemail greetings are the main culprit. A greeting like “Hello, I can’t talk right now” sounds like a live response. Poor call quality, background noise, or network delays can also lead to false positives.

What to do about it:

  • Favor settings that reduce false positives over settings that reduce false negatives. A rep hearing a voicemail greeting for two seconds is a minor cost. Hanging up on a live prospect is a major one.

  • Review call logs periodically. If you see a pattern of short calls (under three seconds) that end without a conversation, your detection may be dropping live pickups.

  • Make sure your phone numbers have good caller ID reputation. Prospects who see a suspicious number may answer with a hesitant, short greeting that the dialer misreads.

What to do when you hit a voicemail

Not every voicemail is a dead end. You have three options, and the right one depends on where the prospect sits in your sequence.

Drop a pre-recorded voicemail. A voicemail drop lets you leave a consistent, well-rehearsed message in one click without spending 30 seconds recording it live. lemlist’s power and parallel dialers include one-click voicemail drop, so the rep taps a button and moves to the next call while the message plays. This works well on a first or second attempt when the prospect does not know you yet.

Skip the voicemail and call back. If you are early in the sequence and plan multiple attempts, sometimes no voicemail is better. A missed call with a recognizable local number can prompt a callback on its own. lemlist internal data shows that reaching a prospect takes persistence: 1 attempt reaches 9.9% of prospects, 2 attempts reach 17.8%, and 5 or more attempts reach 31.2%. Spacing your calls across days gives you more chances without flooding someone’s voicemail box.

Leave a voicemail and follow up on another channel. Pair the voicemail with an email or LinkedIn message the same day. The voicemail creates name recognition, and the written message gives the prospect something to respond to on their own time. For message ideas, the cold call voicemail script guide walks through what to say and what to skip.

Whichever option you pick, log the outcome. Knowing how many times to call a prospect before moving on keeps your list fresh and your reps focused on reachable contacts.

How to set it up in a calling sequence

In lemlist, calls run inside multichannel sequences alongside email, LinkedIn, and SMS steps. Here is how to wire up answering machine detection in a sequence:

  1. Create a new sequence and add your leads. lemlist’s 650M+ contact database with waterfall enrichment across 14+ providers finds phone numbers for around 77% of contacts, so most of your list will be callable.

  2. Add a call step on the day you want the first dial. Place it after an email or LinkedIn touchpoint so the rep has context when the prospect picks up. The call step pulls up the lead’s profile, previous interactions, and any notes from earlier steps.

  3. Choose your dialer. The power dialer ($50 per seat per month) auto-dials the next prospect when you hang up and includes voicemail detection and one-click voicemail drop. The parallel dialer ($150 per seat per month) calls up to five prospects at the same time and connects the rep to the first person who picks up. Both are add-ons on the Multichannel and Enterprise plans. Multichannel starts at $87 per user per month on annual billing. Prices checked November 2026.

  4. Turn on voicemail detection. With detection enabled, the dialer screens pickups and only connects the rep to live humans. When a voicemail is detected, the rep can drop a pre-recorded message or skip and move on.

  5. Add retry call steps. Schedule follow-up call steps on different days. Spreading attempts across the sequence increases your reach. A second attempt nearly doubles your contact rate (from 9.9% to 17.8%), and five attempts get you to 31.2% of your list (lemlist internal data).

  6. Mix channels between attempts. Add an email step the day after a missed call, or a LinkedIn visit before the next dial. The prospect sees your name in multiple places, which makes the next pickup more likely.

Call recordings, transcripts, and AI summaries sync to your CRM (HubSpot, Salesforce), so nothing from the conversation gets lost.

Reach prospects everywhere

Combine email, LinkedIn, and calls to book more meetings.

Frequently asked questions

Accuracy varies by provider, call quality, and the types of voicemail greetings your prospects use. No system catches every machine or avoids every false positive. The practical goal is to minimize false positives (hanging up on real people) even if that means occasionally letting a voicemail through to the rep.

Yes. The dialer needs a moment to listen and classify the audio, which creates a short pause before the rep hears the prospect. This pause is typically one to two seconds. Most prospects interpret it as a normal connection delay, but it is worth knowing it exists.

It depends on your sequence. On a first attempt, a short voicemail drop builds name recognition. On later attempts, skipping the voicemail and calling back on a different day can be more effective. Pair either approach with a follow-up on email or LinkedIn the same day.

Yes. lemlist’s power dialer and parallel dialer both include voicemail detection. When the dialer detects an answering machine, the rep can drop a pre-recorded voicemail in one click or skip to the next prospect. The rep only gets connected to live pickups.
Rémi Kokabi

Rémi Kokabi

Co-founder @GTM Club

Share this post