Last updated 31 August 2026
Probbe records spoken interviews. If you have been sent a link and asked to talk, this page tells you what happens to your voice — before you decide whether to use it.
Probbe is built and operated by Owjar (owjar.co). Owjar is the company responsible for this service and for the data described on this page. If you want to reach a person about any of it, write to pius@owjar.co.
Some researchers recruit through a panel — Prolific, Respondent and others — which pays you for taking part. If you arrived that way, the link carried the id that panel uses for you, and we store it with your answers. It is the only way the researcher can tell the panel you took part, which is what gets you paid.
That id means something to the panel, not to us: we cannot turn it into your name, and we do not try. When the interview ends we send you back to the panel so it records you as finished.
Because these studies pay people, we also keep three one-way marks, each a scrambled value that cannot be turned back into anything about you:
None of them can be turned back into your address, your location or your device. They can only be compared with each other, which answers "did the same person answer twice" and nothing else. The researcher sees whether two responses match; they never see the marks themselves.
We also compare the wording of answers within a single study, because two responses that say nearly the same thing are usually one person. Words are only ever compared inside the study you took part in.
The marks reach one step further: they are compared across the studies run by the same researcher, so that researcher can see you have answered their work before. Never further than that. One researcher never learns anything about another's studies, and nobody learns anything about you beyond "this mark has appeared here before". Answering several studies is not held against you — panels have regulars, and it is shown as context rather than as a problem.
The marks are kept for a year and then deleted. Your answers are not affected by that; it is only the ability to recognise the mark again that expires.
All of this is shown to the researcher as a reason to look again, never as a decision. Nothing here rejects anybody automatically.
We do not analyse your voice to recognise you. Your recording is transcribed and then deleted; no voiceprint is made of it, and we could not identify you from your voice if we wanted to.
If you opened a link someone shared with you — the ordinary case — none of this applies. No id arrives, no mark of any kind is made, your browser is not asked anything about itself, and we do not know who you are.
Your voice for the length of the conversation, and the text of what you said.
We do not ask for your name, your email address, your phone number or any other contact detail, and there is no field in which to give them. We do not know who you are. The one exception is anything you choose to say out loud — if you mention your name during the interview, it is in the recording, because the recording is simply what you said.
Running the interview means sending your audio to other companies. This is not optional to the product and you should know it before you speak:
| Who | What they receive | Where |
|---|---|---|
| Deepgram | Your audio, to turn it into text; and our questions, to turn them into speech | United States |
| OpenAI | The text of your answers, to decide what to ask next and to group answers into themes | United States |
| TypeSafe | The text of your answers, to decide whether a question has already been answered so you are not asked it twice. It returns only probabilities and never writes anything | United States |
| Contabo | Hosts the server holding recordings and transcripts | France |
| Neon | Hosts the database | Germany |
Your recording is not used to train anyone's models by us, and we do not sell it, publish it, or hand it to anyone beyond the companies listed above and the organisation that sent you the link.
The organisation that sent you the link. They see the themes across everyone who answered, and they can read individual quotes. They are the ones who decided to ask you; we provide the tool.
The recording of your voice is deleted after the interview ends. It is kept that long only so a technical fault in the conversation can be listened to and fixed. Nothing in the product reads it, and nobody at the organisation that sent you the link ever hears it.
The written transcript — what you said, as text — is kept until the organisation that collected it deletes it, or until you ask us to. That is the part the research is actually made from. There is no automatic expiry on it, and we would rather say so plainly than describe a schedule we do not run.
You can end the interview at any point by saying so or by closing the page — there is no penalty and nobody follows up. If you have already spoken and want the recording removed, email pius@owjar.co. Tell us roughly when you took part and which survey; we hold no name to search by, so that is what lets us find it.
We send you email through Resend (United States): the sign-in link, a note when your first response arrives, and a note when a study finishes. Nothing else.
We count page views with PostHog (European Union), which sees the addresses of the pages you open and nothing that is on them. See Cookies and measurement below for exactly how it is configured — it is not the default one.
The people you interview are your contacts, not ours. You decide who to send the link to and you are responsible for having a lawful basis to approach them. Under UK and EU data protection law you are the controller of that data and we are your processor.
One cookie, named inview_session, set when you sign in so the
site knows it is you. It cannot be read by scripts and lasts 90 days. Signing
out deletes it. It is the only cookie this site sets, which is why you are not
asked to dismiss a banner.
You are not measured at all. The page you speak on carries no analytics of any kind — no page views, no clicks, no recording of your screen. Nothing on this section applies to you.
We count page views on the marketing site and inside the dashboard using PostHog, on their European servers, so that we can tell which parts of the product are used and which are not. It is set up to see as little as possible:
If your browser sends a Do Not Track signal we do not measure you at all. There is no advertising on this site and we sell nothing to anyone.
One thing worth naming: our pages load two typefaces from Google Fonts, which means your browser contacts Google and Google sees your IP address. No cookie is set by it. We mention it because it is the only request this site makes to a company you did not choose.
If you are in the UK or the EU you have the right to ask what we hold about you, to have it corrected, to have it deleted, to object to how it is used, and to receive a copy. You can also complain to your data protection regulator — in the UK that is the ICO.
To exercise any of these, email pius@owjar.co. We answer within 30 days.
If we change this page in a way that affects what happens to your data, we will change the date at the top and email the people who have accounts.
Written by the people who built Probbe, describing what the software actually does. It is not legal advice.