Free mini-course: five short steps in your browser
Check a suspicious voice message or call in 5 stepsthe ask, a pause, a second channel, saved recordings, and a dated record
TimeBinder adds a layer of protection against AI and human fraud and disputes. A five-step routine for a voice message or call that asks you for something. Progress stays in this browser.
No checker on this page can reliably identify synthetic audio. A TimeBind does not identify an AI voice or verify a recording. The steps are a routine, not a verdict on any one voice.
Start the five steps
Each step is short and ends with one or two checks. Your progress is saved in this browser only. Nothing you choose is sent to TimeBinder. The course uses no audio samples and no real person’s voice.
Your progress is saved in this browser only. Choices stay on this device and are not sent to TimeBinder or tied to your name.
Step 1 of 5. 0 of 5 complete.
Step 1. Name the ask, the voice and the channel
Suspicious voice messages usually come with a story and a deadline: a relative in trouble, a boss who needs a payment, a bank “security” call. Before you judge how the voice sounds, write down the ask in one line (what are you being asked to do, and by when?) and note the channel it arrived on: a call from an unknown number, a voice note in a chat app, a voicemail, a clip on social media. A voice can sound exactly right and the ask can still be a scam, and a strange-sounding voice can still be a real person on a bad line. Keep the three apart.
Check 1. Match each part to what it means
Tap a part, then tap the meaning that fits it best.
Check 2. A voice note asks you to send money today
What is the best first move?
Step 2. Pause: do not trust your ear, and do not be rushed
Voice-cloning tools exist, and a short public clip of a person can be enough for some of them to imitate that voice, so a convincing voice is no longer a safe test. People also misjudge real voices on poor lines. Treat “it sounds like them” as a feeling, not a check. Urgency, secrecy (“do not tell anyone”), unusual payment methods and requests for one-time codes are the pattern to look for. The simplest protection costs nothing: stop, do not pay, do not read out codes, and move to Step 3.
Check 1. Which statement is the honest one?
Check 2. The caller says “do not tell anyone, and do it now”
What does that tell you?
Step 3. Verify on a second channel you already trust
End the call, or leave the message unanswered, and contact the person using a number or account you already had before this contact, not the number in the message and not a number the caller gives you. For an organisation, use the number on your card, your statement or the official site you typed yourself. A family “safe word” agreed in advance can be one extra check, but it can be forgotten or leaked, so it is a layer and not a guarantee. If the person cannot be reached, ask someone else who would know, and wait. A real emergency survives a ten-minute check.
Check 1. Which of these are sound verification moves?
Tick all that apply.
Check 2. The “caller” gives you a number to call back
What do you do?
Step 4. If it is a saved recording: source, claimed speaker, labels
A voicemail, a voice note or a clip someone sends you is a file, and a file raises different questions from a live call. Where did it come from, and who sent it first? Does the person it claims to be confirm they said it, on a channel you trust? Does the platform show a label, or does the file carry Content Credentials (C2PA) data? Those are clues: they record what a system recorded, if still present, and metadata is easily lost when audio is re-encoded or shared. Online audio detectors give a score or a guess, not a verdict, and none can reliably identify synthetic audio. Do not decide from a score alone, and do not share the clip as “proven fake” or “proven real”.
Check 1. A detector says “93% AI” about a voice note
Check 2. What carries the most weight about a saved voice note?
Step 5. Decide, report, and keep a dated record
Honest outcomes include “confirmed by the person on a trusted channel”, “could not be confirmed, so I did not act”, and “looks like a scam, reported”. If you already paid or shared a code, contact your bank or provider straight away and report it to your country’s fraud-reporting service. If you may need to show later what you received, keep the exact voicemail or voice-note file you were sent, plus screenshots of the call log or chat, and write notes of what you checked. Rules about recording calls differ by place, so check yours before recording anyone. A TimeBind records that an exact file existed no later than a recorded time, and that a later copy still matches its SHA-256 fingerprint, anchored to Bitcoin. It does not show who spoke, that the audio is genuine, or whether it is synthetic. First Pooled TimeBind is free, then from US$5. An ASAP TimeBind is US$15.
Check 1. What can a TimeBind show about a voicemail file?
Check 2. You could not confirm the message. What now?
- SHA-256 file hash checker
- A law firm and voice-cloned payment instructions
- Glossary: existed no later than
You have finished all five steps
Progress stays in this browser. You can revisit any step above. If you keep a voice file you may need later, date the exact file you saved.
How do you check a suspicious voice message or call? The short answer
Work through five steps: write down what you are being asked to do and how it reached you; pause, because a convincing voice is not proof; end the contact and verify on a number or account you already had; for a saved recording, trace the source and ask the claimed speaker; then decide honestly, report it if it looks like a scam, and keep the exact file and your notes.
None of this identifies whether a voice is synthetic, and an online detector score is a clue at most. A TimeBind of the exact file you were sent can show that it existed no later than a recorded time and still matches. It does not show who spoke, or that the audio is genuine.
Why this course, in November 2026
Search suggestions checked on 13 November 2026 return full sets for “how to tell if a voice is ai”, “how to tell if audio is ai generated” and “how to tell if a voicemail is ai”, including variants for voice notes and voice messages. X’s recent-post counts for the week to 12 November show 1,107 English-language posts that mention voice cloning or AI voices together with scams, fakes, calls or safe words (112 to 180 a day on the complete days from 6 to 11 November). These counts measure interest, not accuracy, and are modest next to the video topic, so we treat this as a steady need rather than a spike. None of the figures says that a voice can be settled by ear.
This is the third course in the series. The first, Spot AI fakes in 5 short lessons, has a short lesson on voice scams among other things. The second, Verify a viral video in 5 steps, is a workflow for a single clip. This one is a workflow for a voice message or call that asks you to act. A fourth course, Verify a screenshot or fake document in 5 steps, covers screenshots and documents. A fifth course, Verify an AI-generated photo of an event in 5 steps, covers photos of events. A sixth course, Verify a viral news clip in 5 steps, covers news-style clips. A seventh course, Verify a leaked-looking chat or email screenshot in 5 steps, covers leaked-looking chat and email screenshots. An eighth course, Verify a fake product review or testimonial in 5 steps, covers product reviews and testimonials. A ninth course, Verify an AI-generated receipt or invoice in 5 steps, covers AI-generated receipts and invoices, and a tenth, Check a suspicious profile photo or dating profile in 5 steps, covers suspicious profile photos and dating profiles. An eleventh course, Verify an AI-edited rental or property listing photo in 5 steps, covers rental and property listing photos, and a twelfth, Check a suspicious celebrity or expert video ad in 5 steps, covers celebrity and expert video ads. A thirteenth course, Check a photo, video or audio file offered as evidence in 5 steps, covers photo, video or audio files offered as evidence, and a fourteenth, Check a suspicious video call or remote job interview in 5 steps, covers suspicious video calls and remote job interviews. A fifteenth, Check a suspicious website or login page in 5 steps, covers suspicious websites and login pages, and a sixteenth, Check a map or satellite image that may be AI-made or edited in 5 steps, covers maps and satellite images that may be AI-made or edited. All sixteen sit under the public guide to protecting yourself from AI forgeries.
What each step can show, and what it cannot
The honest job of this routine is to slow a snap judgement and leave you with notes you can show. See also synthetic media and existed no later than in the glossary.
| Step | What it can show | What it cannot show |
|---|---|---|
| 1. Ask, voice, channel | What you are being asked to do, by when, and which channel it used. | Who is really speaking. |
| 2. Pause | Whether urgency, secrecy or unusual payment requests are part of the pattern. | Whether a voice is synthetic. A convincing voice proves nothing, and neither does an odd line. |
| 3. Second channel | Whether the person you already know confirms the request, using a number or account you already had. | Anything, if the “trusted” channel was supplied by the caller. A safe word is an extra layer, not a guarantee. |
| 4. Saved recording | Where the file came from, what a platform label or a C2PA / Content Credentials manifest recorded, if still present, and whether the claimed speaker confirms it. | Truth or authorship. A missing label settles nothing, and a detector score is not a verdict. |
| 5. Decide and keep | What you checked and found, and that an exact saved file existed no later than a recorded time, with a later copy still matching its SHA-256 fingerprint. | Who spoke, that the audio is genuine, or whether it is synthetic. |
If you may need to show what you received
If you were sent a voicemail or voice note and may need to explain it later (to a bank, an employer, a platform or a lawyer), a dated record of the exact file helps:
- Keep the exact file you were sent, not only a re-encoded forward or a screen recording.
- Make a TimeBind of that file. The fingerprint is made in your browser and the file stays on your device. You can later download a Proof of Time Certificate.
- Store the file, the certificate, screenshots of the call log or chat, and your notes together. Use the file date and metadata inspector or the file-changed answer to compare a copy later.
First Pooled TimeBind free, no card, then from US$5. An ASAP TimeBind is US$15. A TimeBind does not identify an AI voice or verify the contents of a recording.
First Pooled TimeBind free, no card, then from US$5. An ASAP TimeBind is US$15.
What a TimeBind shows, and what it does not. It shows that an exact file existed no later than a recorded time and that a copy shown later still matches its fingerprint. It does not show who made the file, who owns it, that its contents are true, or whether AI was used to make or edit it. This is general information, not legal advice.
Check a suspicious voice message: FAQ
Can this course tell me whether a voice is AI-generated?
No. No checker on this page, and no tool on this site, can reliably identify synthetic audio. The steps help you check the ask, the channel and the source, and verify with the person on a channel you already trusted. “Could not confirm, so I did not act” is a valid result.
Can I tell a cloned voice by listening?
Not reliably. Voice-cloning tools exist and real voices also sound odd on poor lines, so “it sounds like them” is a feeling, not a check. Verify with something other than the sound of the voice, such as calling back on a number you already had.
Is a family safe word a good idea?
It can be one extra check, if you agreed it in advance and keep it private. It can be forgotten, guessed or leaked, so treat it as one layer next to calling back on a trusted number, not as a guarantee.
Does a TimeBind verify a voice message or show it is not AI?
No. A TimeBind shows that an exact file existed no later than a recorded time and that a later copy still matches its fingerprint. It does not show who spoke, that the audio is genuine, or whether it is synthetic.
Should I record a suspicious call?
Rules on recording calls differ by country and region, and some require consent from everyone on the call. Check the rules where you are before recording anyone. If you receive a voicemail or voice note, you can keep the exact file you were sent.
Where is my progress saved?
In this browser only, using localStorage. Choices are not sent to TimeBinder and are not tied to your name. Clearing site data, or using another browser, starts the steps again.
Is the course free?
Yes. The five steps and the checks run in your browser at no charge. If you later make a TimeBind, the first Pooled TimeBind is free, then from US$5. An ASAP TimeBind is US$15.
Related guides and tools
- Spot AI fakes in 5 short lessonsThe first mini-course: labels, reverse image search, voice scams and more
- Verify a viral video in 5 stepsThe second mini-course: a workflow for a single clip
- How to protect yourself from AI forgeriesThe public AI-fakes guide this course sits beside
- Scam voice red flags checkerA situation check for a suspicious voice contact
- How to tell if audio is AI generatedWhat can and cannot be shown about audio
- Content Credentials checkerWhat a C2PA manifest can and cannot show
- AI watermark and label decoderWhat a platform label or watermark result can show
- GlossaryPlain definitions of C2PA, SHA-256 and existed no later than
Date the exact files you keep
Add a layer of protection against AI and human fraud and disputes: a dated, checkable record of the exact file, kept while you still have it.
- Open the app and sign in with email or Google.
- Add your files. They stay on your computer.
- Seal them and download your Proof of Time certificates.
First Pooled TimeBind free, no card, then from US$5. See pricing.
Related glossary entries: digital replica and liveness detection.