A bad phone recording can’t be saved in editing. I learned that the hard way after a 40-minute interview with a guest who was driving with the windows down. The conversation was great. The audio was road noise with a voice buried somewhere underneath. There was no fixing it, and no getting his time back.
So before you learn how to record a phone interview, know that the recording either works or it doesn’t, and you find out after the guest hangs up. About half of U.S. adults (49%) have listened to a podcast in the past year, and a lot of that content is interviews with people who were never in the same room as the host. Phone and remote interviews are how most small shows book guests. Getting the capture right is the whole game.
This guide covers the full loop: consent, the recording methods that actually hold up, keeping the audio clean, turning it into a usable transcript, and getting more than one piece of content out of a single call.
Get consent before you hit record
Recording someone without permission isn’t just rude. In some states, it’s illegal.
The rule splits by state. Most run on one-party consent, meaning you can record a call you’re part of without telling anyone. But a group of states requires all-party consent, where everyone on the line has to agree first. California, Florida, Illinois, Pennsylvania, Massachusetts, and Washington are in that group. Justia keeps a plain-English 50-state survey of recording laws if you want to check your state and your guest’s.
One rule catches people out. If you’re in a one-party state and your guest is in a two-party state, you’re usually bound by the stricter one. Interstate calls default to caution.
The fix is easy, and it takes ten seconds. At the top of the call, once you’re already recording, say: “Just confirming you’re good with me recording this for the podcast?” Get the yes on tape. That verbal confirmation covers you legally, and it saves you the awkward email later when a guest asks what happened to the off-the-record bit.
Four ways to record a phone interview that actually work
Your recording method comes down to what your guest is willing and able to use. Some guests will happily install software. Most just want to pick up a phone. Match the method to the guest, not to your ideal setup.
Remote recording apps (Riverside, Zencastr, SquadCast)
If your guest can click a link, this is the best audio you’ll get. These tools record each person’s voice locally in high quality, then upload the files separately, so a dropped connection doesn’t wreck the recording. You get isolated tracks for editing, which means you can cut the guest’s cough without cutting your question.
The catch is friction. Every extra step you ask of a guest is a chance for them to reschedule. I’ve had a CEO bail on a session because the browser mic permission confused him for five minutes. Now I send a one-line setup note the day before and stay on the line while they join.
A softphone or Google Voice number
For guests who only have a phone and zero patience for apps, a softphone number is the path of least resistance. They dial a number, and you record. Simple.
One honest limitation worth knowing: Google Voice only records incoming calls, not ones you place, and it announces the recording with an automated voice. That announcement is fine, sometimes even helpful for consent, but it surprises people who expected a silent capture. Test it once on yourself before you rely on it for a real guest.
The double-ender
A double-ender is where each person records their own side locally, then sends you the file afterward. You talk over any connection you want, but the audio each of you captures is studio-clean because it never travels through the phone line.
This is the pro move for a guest who happens to have a decent mic. It’s also the method most likely to fall apart, because it depends on your guest remembering to hit record and actually finding the file afterward. Budget for one guest per year, sending you thirty minutes of silence because they recorded the wrong input.
A recording-enabled business phone line
If you interview people by actual phone on any kind of regular schedule, a business phone system that records and transcribes calls automatically is worth a look. The call comes in over a normal line, gets recorded, and a transcript and summary are waiting for you when you hang up. No separate recorder, no file to export, no transcription step bolted on later.
Allo works this way: it records, transcribes, and summarizes every call, so the audio and a rough transcript land in the same place. It’s built as a business phone system rather than a podcast tool, so you won’t get isolated tracks the way Riverside gives you. But for a journalist or a B2B host doing five phone interviews a week, having the recording and transcript handled in one step beats stitching three tools together. Fair trade, depending on what you’re after.
Clean audio is the difference between a transcript and a mess
Every transcription tool, human or AI, gets worse the dirtier your audio is. A guest on speakerphone in a tiled kitchen, with the TV going behind them, is what turns an automated transcript into a paragraph of “[inaudible].” Fix the input, and everything downstream gets easier.
Ask your guest to do three things before the call. Find a small, soft room; a closet full of clothes beats a kitchen with hard tile. Wear wired earbuds, even cheap ones, because the mic sits closer to their mouth and kills echo. And check their signal, then stay put, since walking around mid-call is how you get audio that fades in and out.
Do your own homework too. Charge the recording device fully. Run a ten-second test clip and actually listen back before the guest is on the line. My road-noise disaster would have been a two-minute fix if I’d listened to a test clip first and just asked him to pull over.
How to transcribe the interview
Match the transcription method to how much the exact words matter.
For show notes, clip-hunting, and searching a conversation later, AI transcription is fast, cheap, and good enough. Tools like Descript, Otter, and Rev’s automated tier will turn an hour of clean audio into text in a couple of minutes for a few dollars or less. The accuracy drops with accents, jargon, and poor phone audio, so expect to clean up names and technical terms by hand.
When the words have to be exact, such as a research interview you’ll quote, anything with legal or reporting weight, pay for human or human-reviewed transcription. It costs more and takes longer, but you’re buying accuracy you can stand behind.
One trick for checking a shaky AI transcript against the audio: slow the playback down. Dropping to 75% speed makes it much easier to catch the words the software mangled, and you can slow down audio without changing the pitch so voices still sound normal. It’s the fastest way I’ve found to fix a transcript without transcribing the whole thing from scratch.
You’ll also pick between verbatim and clean-verbatim style. Verbatim keeps every “um” and false start, which matters for research where the hesitation is data. Clean-verbatim strips the filler so the text reads smoothly, which is what you want for show notes and quote cards. Pick before you order, because redoing it costs you twice.
One interview, several pieces of content
A one-hour interview is four or five pieces of content, but most shows publish it once and move on.
A single interview transcript is raw material for a lot more than the episode. Pull three or four punchy quotes for quote-card graphics. Cut 30-to-60-second audio moments for social clips. Turn the transcript into a written recap or a blog post. Grab the best one-liner for the episode title. The transcript makes all of this a copy-paste job instead of a re-listening job, which is exactly why getting the transcript right earlier pays off here.
[IMAGE: One audio file in the center branching out to five outputs — full episode, audiogram clip, quote card, blog recap, episode title] Alt text: Diagram showing one interview recording repurposed into five content formats
Keep your interview audio where you can find it
A recording you can’t find six months later might as well not exist. The workflow that kills more content than bad audio is a Downloads folder with 200 files named “recording_final_2.mp3.”
Set a naming convention before you have a mess, not after. Guest name, date, and topic are enough: smith_2026-03-14_pricing.mp3. Keep files grouped by guest or by season so you can pull an old interview when a guest comes back or a topic resurfaces. And back up the raw files to the cloud, because a dead laptop shouldn’t cost you your archive. It’s straightforward to upload audio files to Google Drive and keep a synced copy off your machine.
Hang onto the source link or original file too, not just the edited cut. When someone questions a quote a year later, you’ll want the raw recording to point to.
Frequently asked questions
Is it legal to record a phone interview? It depends on the states involved. One-party consent states let you record a call you’re on without telling anyone; all-party states require everyone to agree first. When in doubt, ask for a verbal yes at the top of the call and get it on the recording.
What’s the easiest way to record a guest who only has a smartphone? A softphone number or a browser-based recording link. The guest dials or clicks, and you handle the recording on your end. Skip anything that asks them to install and configure software.
How accurate is AI transcription for phone audio? Good on clean audio, shakier on phone calls with accents, jargon, or background noise. Fine for show notes and searching. For anything you’ll quote exactly, use human or human-reviewed transcription.
Can I record and transcribe a call at the same time? Yes. Some business phone systems record and auto-transcribe every call, and standalone tools like Otter can transcribe live. The live transcript is a draft, though; always check it against the audio before you publish a quote.
Where to start
Pick your recording method for your next guest based on what they can actually use, and run one test call with a friend this week. Record it, transcribe it, and listen back to where the audio or the transcript broke down. You’ll learn more from one flawed test than from any checklist, and you’ll never again find out about road noise after the guest has hung up.