Daniel's been fighting with a pharmacy bureaucracy all week — the kind where you call, they promise something, you call back, nobody knows what you're talking about. And his instinct, which I think is the right one, is to reach for a recorder. He's done this before — he once recorded a landlord committing to a repair that the guy later denied ever promising. That recording saved him. But here's what he's actually asking: the law says you can record, fine — but how do you actually do it well? How do you capture both sides of a conversation when modern phones seem designed to stop you, and how do you get the other person to say something specific and useful on the record without them clamming up?
This is the part nobody talks about. Everyone fixates on the legality — is it one-party consent, two-party consent, what state am I in — and then they assume the rest is just pressing a button. It's not. Recording for evidence is a craft. It's technical, it's conversational, and honestly it's psychological. You're managing a social interaction while operating a covert recording system, and if either half fails, you've got nothing.
So today we're going to talk about the unglamorous, deeply practical art of recording for evidence. Not the law, not the ethics — the craft. What gear works, what doesn't, and how to steer a conversation so the recording actually proves something.
And we should do the legal frame quickly, because the rules set the boundaries for everything else. In Israel, the Wiretap Law from nineteen seventy-nine establishes one-party consent. If you're part of the conversation, you can record it without telling anyone. The Bismuth bill, which passed recently, actually reinforced this — it clarified and strengthened the one-party consent provision.
The key phrase being "if you're part of the conversation." You can't bug a room and leave. You can't record two other people talking without being in it yourself. And there are exceptions — if you're recording with criminal intent, that's a different thing entirely. But for the kind of situation Daniel's describing — a landlord, a government clerk, a pharmacy that keeps losing prescriptions — you're in the clear.
Admissibility is a separate question from legality, and that's worth flagging. A court can decide a recording is legal but still not admit it as evidence if the chain of custody is sloppy or the authenticity is questionable. Which is exactly why the technical side matters so much. You're not just capturing audio — you're building a piece of evidence that has to survive scrutiny.
So you've got the legal green light. But the hard part is actually getting the audio. Let's talk about the technical trenches.
The first thing to understand is that modern smartphones are actively hostile to call recording. Apple and Google have both locked it down hard. On iOS, there's no native call recording at all — the API simply doesn't expose the audio stream during a call. On Android, it's slightly more variable, but Google has been tightening the screws for years. There used to be apps that could tap into the call audio, and most of them are dead now.
Which is genuinely strange when you think about it. A phone is a device for talking to people, and recording that talk is somehow the one thing it won't let you do.
The stated reason is privacy — protecting the person on the other end who might be in a two-party consent jurisdiction. The practical effect is that everyone who needs to record calls for legitimate reasons has to use workarounds. And the workarounds are clunky, but they work.
Let's break this into the two scenarios Daniel asked about. In-person recording, and phone calls. They're completely different problems.
In-person is actually the easier one, technically, once you understand the physics. If you just hit record on your phone and set it on the table, the person across from you sounds like they're at the bottom of a well. Your own voice is fine because you're close to the mic, but theirs is distant, echoey, buried in room noise. That recording is useless.
The ten-inch rule.
Microphone placement is everything. Every time you double the distance, you lose about six decibels of signal. So if someone is three feet away and you're recording from your pocket, their voice might be fifteen or twenty decibels quieter than yours. That's the difference between "I can kind of make out what they said" and "a transcription service will produce gibberish."
So what's the fix? Nobody's going to hold a microphone in someone's face during a covert recording.
Lapel mic. A small wired lavalier microphone that plugs into your phone, with the phone in your pocket and the mic clipped inside your shirt or jacket, pointing outward. These cost maybe fifteen to thirty shekels for a basic one, and they transform the audio quality because the mic is now on your chest instead of buried in your pocket. It picks up your voice clearly, and it also picks up the person across from you much better because it's higher and less obstructed. And it doesn't look like anything — a tiny black clip on a shirt collar, nobody notices.
The other trick is positioning. If you can place your phone on the table between you, face down, with the microphone pointing toward the other person, that's significantly better than having it in your pocket. It looks natural — people put their phones on tables all the time. Just make sure the mic isn't covered by a case or a napkin.
If Daniel's using a lapel mic, the phone stays in his pocket and he just stands normally. If he's relying on the phone's built-in mic, he wants to set it on a counter or table between them, ideally with the bottom edge — where the mic usually is — facing the landlord. And he wants to be closer to the phone than the landlord is, because his own voice will still dominate, but at least the landlord won't sound like he's shouting from another room.
What about a second device as backup?
Non-negotiable. This is the two-device principle. Devices fail. Batteries die. Apps crash. Storage fills up. You hit record and the app silently doesn't start. If you're recording something that matters — a landlord committing to a repair, a government worker giving you a reference number — you cannot rely on a single device. Use your phone as the primary and something else as the backup. An old phone with no SIM card works fine. A cheap voice recorder from any electronics store. Even a second app on the same phone, though that's riskier because if the phone dies, both are gone.
I've done this with a small digital recorder I bought for maybe sixty shekels. It's the size of a thumb drive, runs on a single AAA battery for something like twenty hours, and the audio quality is shockingly good. It lives in my bag.
And that's the thing — dedicated voice recorders haven't gotten worse just because smartphones exist. They've actually gotten better and cheaper. A fifty-shekel recorder from a brand like Sony or Olympus will capture better audio than a two-thousand-shekel phone, because it's designed to do exactly one thing.
Phone calls are where it gets difficult.
The shift away from call recording on smartphones is the biggest obstacle. For years, Android had a workable call recording API, and apps like ACR — Another Call Recorder — worked reliably. Google killed that. Apple never allowed it. So now, if you want to record a phone call, you have basically four options.
Walk me through them.
Option one, the low-tech fallback: speakerphone plus a second recorder. You put the call on speaker, set your phone next to a voice recorder or a second phone running a recording app, and capture the whole thing through the air. The audio quality isn't great — it's speakerphone quality, which is compressed and tinny — but it's reliable and it works on any phone with no setup.
And it's obvious what you're doing if someone walks in, which is a different problem. Option two?
VoIP services that include recording. Google Voice lets you record incoming calls by pressing four — it announces that recording has started, which kills the covert aspect, but for outgoing calls you can enable recording in settings and it won't announce. Skype has recording built in. These work well if you're initiating the call from the service, but they don't help if someone calls your regular number.
Option three?
Hardware. There are call recorder dongles — little devices that plug into your phone's headphone jack or lightning port and sit between the phone and your headphones. They tap the audio directly from the line. These work regardless of software restrictions because they're intercepting the analog audio signal, not the digital call stream. The downside is they're conspicuous — you've got a thing plugged into your phone — and they're not cheap. Figure a hundred to two hundred shekels.
And option four?
A second phone on the call. If you've got a three-way calling setup or a conference line, you can dial in a second device that's recording. This is getting into elaborate territory, but for something like a call with a government agency where you really need the evidence, it's worth considering.
The speakerphone method is underrated, honestly. It's what I'd recommend to someone who needs to record a call tomorrow and doesn't want to buy anything. The quality is mediocre, but mediocre is infinitely better than nonexistent.
One thing to watch with speakerphone: test it first. Call a friend, put them on speaker, record with your backup device, and listen to the playback. Some phones have terrible speakerphone mics. Some rooms have awful echo. You want to know what you're working with before the real call.
Audio quality isn't just about clarity — it's about whether the recording is usable as evidence. If a transcription service can't make out what the other person said, or if a judge has to strain to hear the key phrase, you've lost. Background noise is the enemy. Air conditioners, traffic, other people talking, the hum of a refrigerator. All of that competes with the voice you're trying to capture.
The ten-inch rule applies here too. If you're recording a phone call on speakerphone, get the recording device as close to the speaker as possible. Don't put it across the room. Put it right next to the phone.
There's a deeper problem that's been creeping in over the last few years. A raw recording isn't automatically trusted anymore. Deepfakes have changed the evidentiary landscape. Anyone can generate a convincing fake audio clip of anyone saying anything. So if you show up with a recording, the first question isn't "what does it say" — it's "how do we know it's real."
This is where metadata becomes critical. A recording file straight from a voice recorder app will usually have a timestamp embedded in the file properties — creation date, modification date, sometimes device information. Don't strip that. Don't edit the file in any way that alters the metadata. If you need to trim the recording, keep the original untouched and work from a copy.
Cryptographic hashing is the next step up. You generate a hash of the file — a unique digital fingerprint — immediately after recording, and you store that hash somewhere independent. Later, if someone challenges the recording's authenticity, you can re-hash the file and show it matches. Any alteration, even a single bit, produces a completely different hash.
For most people in most situations, this is overkill. A landlord dispute isn't going to turn on cryptographic verification. But if you're dealing with something that might actually end up in court, it's worth knowing that the standard for authenticity is rising, and raw files with intact metadata are the baseline.
Okay, you've got a clean recording of both sides. But if you didn't ask the right questions, it's just noise. Here's how you actually get the evidence you need.
This is the part I find fascinating, because it's not about technology at all. It's about conversational architecture. You're designing the interaction so the other person voluntarily says the thing you need them to say, on the record, without realizing they're doing it.
The most common failure mode is people thinking the recording itself does the work. They hit record, have a normal conversation, and then later discover that the other person never actually said anything specific. They said "we'll take care of it" or "don't worry about it" or "I'll look into that." None of that is evidence of anything.
Vague assurances are the enemy. Your job during the conversation is to convert vague assurances into specific commitments. And the tool for that is the clarifying question, framed as helpfulness.
Give me an example.
The landlord says "I'll fix the leak." You say, "Great, thank you — so just so I'm clear, you'll have someone here by Friday, and you're covering the parts and the labor?" You've done two things. You've pinned down a date and a financial commitment. And you've phrased it as a confirmation, not an interrogation. The landlord is almost certainly going to say yes, because saying no would require explaining why, and most people don't want to do that in the moment.
And now you have a yes on the record. A specific yes to a specific question.
This is the mirroring technique. You paraphrase what they said, adding the specifics they left out, and you ask them to confirm. If they correct you — "no, I said next Tuesday, not Friday" — that's even better, because now they've volunteered a specific date unprompted. Either way, you've moved from "I'll fix it" to "I'll fix it by Tuesday, at my expense." That's the difference between a recording that's useless and a recording that proves something.
What if they deflect? "We'll see what we can do," or "It depends on the schedule."
Then you ask the next question. "What schedule are we looking at — is this a few days or a few weeks?" You're not arguing. You're not accusing. You're just asking for clarification, over and over, until you get something concrete or until the deflection itself becomes the evidence. A recording of someone refusing to commit to a timeline, after being asked directly three times, is itself useful.
Silence is a tool here. Most people are uncomfortable with silence in a conversation, and they'll fill it. If you ask a direct question and they give a vague answer, just... wait. Don't jump in. Don't help them out. Let the silence sit. More often than you'd expect, they'll add something — a detail, a concession, a qualification — that they wouldn't have volunteered otherwise.
The clinical parallel is striking. When I was practicing, you learned very quickly that patients will tell you the important thing in the pause after they've finished their initial answer. The first thing they say is the rehearsed version. The second thing — the thing that comes out to fill the silence — is often the truth.
That's... slightly dark, Herman.
It's not dark, it's just how people work. We're social animals. Silence feels like a breakdown in the interaction, and we rush to repair it.
The emotional trap is the other big failure pattern. You're frustrated. The landlord has been dodging you for weeks. The pharmacy has lost your prescription three times. You want to sound angry because you are angry. But the moment you sound accusatory, the other person's defenses go up. They stop volunteering information. They switch to short, guarded answers. The recording becomes useless.
Frame everything as clarification, never as accusation. "I just want to make sure I understand" is a magic phrase. "Help me get this straight" is another one. You're positioning yourself as someone who's trying to be thorough, not someone who's building a case. Even if you are building a case.
The bad faith scenario is where this gets trickier. You suspect the other person is lying or being deliberately evasive. Your instinct is to catch them in the lie. But a direct confrontation — "you said X last week and now you're saying Y" — usually just produces denial or hostility.
The better approach is to ask the same question in different ways, separated by other topics. You ask about the timeline early in the conversation. Then you circle back ten minutes later and ask again, phrased slightly differently. If they give inconsistent answers, the inconsistency is captured on the recording without you ever having to point it out. Later, when you play it back, the contradiction is right there.
"So the repair will be done by Tuesday?" — pause, talk about something else — "And just to confirm, the plumber's coming next week?" If the answers don't match, you've documented the inconsistency without triggering a confrontation.
Getting them to state their name and the date on the record is a small thing that pays off enormously. At the start of an in-person conversation, "Sorry, remind me of your full name for my notes?" At the start of a phone call, you can reference the date naturally: "Thanks for calling me back about the August fourth issue." Now the recording has identifying information baked in, and nobody can claim it was a different conversation on a different day.
Daniel's prescription situation is a perfect case for this. He's dealing with a bureaucracy where different people give different answers, nobody takes responsibility, and promises evaporate. The recording is his only way to pin down who said what.
In that scenario, I'd want to capture the person's name, their department, the specific action they're promising, and the reference number or case number if there is one. "So you're saying the prescription will be ready for pickup by Thursday, and I should ask for Miriam at the front desk — is that right?" Every piece of that is ammunition when he calls back and someone says they have no record of it.
And if they refuse to give a name, which happens — some government workers are trained not to — you still have their voice, the time stamp, and the content of what they said. The recording itself establishes that someone in that office made that commitment.
Post-recording discipline is where a lot of people lose the plot. You've got the file. Now what? First, back it up immediately. Cloud storage, email it to yourself, put it on a thumb drive — do at least two things before you do anything else. Files get deleted. Phones get lost. If the recording matters, treat it like it's fragile, because it is.
Second, transcribe it. Even a rough transcript — just typing out the key exchanges while they're fresh in your mind — makes the recording vastly more usable later. Two weeks from now, when you need to find the part where they promised the refund, you don't want to be scrubbing through a forty-minute recording trying to find it.
Third, document the context in a separate file. Date, time, location, who was present, what the conversation was about, why you were recording. This is the chain of custody. It doesn't have to be formal — a note on your phone is fine — but it establishes that you documented the circumstances contemporaneously, which matters if the recording is ever challenged.
The deepfake problem loops back in here. If you've got the original file with intact metadata, a backup, a transcript, and a contemporaneous note, you've built a pretty solid authenticity package. Anyone claiming the recording is fabricated has to explain how all of those pieces line up.
And for the vast majority of disputes — landlord-tenant, consumer complaints, bureaucratic snafus — that package is more than enough. Nobody's running spectrographic analysis on a recording of a plumber promising to show up on Thursday.
Before we wrap up, Hilbert's been pacing in the control room. He's got a problem with all of this.
Hilbert: I do. The whole episode has been about how to do this thing well, and the unspoken assumption is that doing it well is a good thing. But I keep thinking about what happens when this becomes normalized. Daniel records his landlord. Great, the landlord gets held accountable. But now the landlord knows tenants are recording. So the landlord starts recording every conversation with every tenant. And the landlord has better equipment, and a lawyer on retainer, and experience being deposed. The power imbalance doesn't disappear — it just shifts to whoever is better at the game.
That's fair. But the landlord already has power — that's the starting condition. The recording is an equalizer, not a new weapon.
Hilbert: Is it, though? The pharmacy Daniel's fighting — they almost certainly record all their calls already. Every customer service line does. "This call may be recorded for quality assurance." So Daniel recording them isn't leveling the field — it's just adding a second recording to a conversation that was already being recorded by the more powerful party. And the more powerful party has a compliance department that knows exactly what to say and what not to say on a recorded line.
That's actually an argument for recording, not against it. If they're already recording, and they're using that recording to protect themselves, then the individual needs their own copy of the interaction. Otherwise it's entirely their word against an institution that has the audio and can selectively quote from it.
Hilbert: I'm not saying it's never justified. I'm saying the frame of "this is a tool for the little guy" papers over something uncomfortable. In a one-party consent world, everyone is eventually recording everyone. Trust erodes. Conversations become performances. The landlord who might have been willing to work something out informally now insists on everything in writing because he knows anything he says could be used against him. That's not a better world — it's just a more lawyered one.
I think the line is intent. Recording a casual conversation with a friend because you might need leverage later — that's corrosive, and I wouldn't defend it. Recording a landlord who's already broken promises, or a government agency that's already lost your file three times — that's defensive. You're not creating the adversarial relationship. You're responding to one that already exists.
Hilbert: And who decides when the relationship has become adversarial enough? Everyone thinks their situation is the exception. The guy recording his neighbor over a fence dispute thinks he's being defensive too.
I don't have a clean answer for that. I think you're right that there's a slippery slope, and I think Corn's right that there are situations where recording is the only way to protect yourself. The prescription case — Daniel's not trying to trap anyone. He's trying to get his medication. The recording is a memory aid as much as anything.
Hilbert: I'll grant that the prescription case sits differently than the landlord case. But the episode spent forty minutes on technique and about ninety seconds on "maybe think about whether you should." That ratio feels off to me.
It's a fair hit. The technique is what Daniel asked about, but the ethical weight is real. I still land on the side of recording as a shield — in an adversarial context, the alternative is he-said-she-said, and the institution usually wins that game. But I take your point that a world where everyone's running a covert recorder is not a world I want to live in.
Hilbert: I'm not asking you to solve it. I'm just saying the question is bigger than microphone placement.
It is. And I think the question of where the line is — that's going to get harder, not easier, as the tech improves. When recording is effortless and invisible, the social norm either adapts or breaks. We're in the middle of that adaptation right now, and nobody knows where it lands.
That's the open question, really. As AI makes audio manipulation easier — deepfakes that are indistinguishable from real recordings — does the one-party consent recording lose its evidentiary value entirely? If anyone can generate a fake clip of anyone saying anything, then a recording proves nothing. We might be in a brief golden age where recordings still carry weight, and that window is closing.
The counter to that is the authenticity package we talked about — metadata, hashing, chain of custody. A recording with a verifiable provenance is still hard to fake. But most people won't do that work, and most disputes won't merit it. So the practical value of a raw recording probably does decline over time.
The craft of recording, though — the part where you ask clear questions and get specific answers — that doesn't lose value. That's just good communication. Even without a recorder, pinning someone down on what they're actually promising is a skill that pays off. The recording is the backup; the clarity is the real work.
The most common misconception people have about all of this is that one-party consent means you can record anything, anywhere, and it'll hold up. The reality is narrower. It only applies to conversations you're part of. It doesn't guarantee admissibility. A poorly captured recording with no context and no clear commitments is legally worthless even if it was perfectly legal to make. The consent rule is just the starting line — everything after that is craft.
And the craft is learnable. That's the encouraging part. You don't need to be a natural interrogator. You just need to go into the conversation knowing what you need them to say, and then ask the question that makes it easy for them to say it.
Thanks to Hilbert Flumingtop for producing, and for keeping us honest on the ethical side of this. He's right that the question is bigger than technique, even if I still mostly land where Daniel does.
This has been My Weird Prompts. If you've got a weird prompt — or a recording horror story — send it to us at show at my weird prompts dot com.
We'll be back soon.
