This covers two things: this website, and Prospektor Studio — the tool where a sales team researches a target company and drafts the material to approach it. It's written to be read. Where the honest answer is unflattering, it's here anyway.
Prospektor builds Prospektor Studio, an AI go-to-market tool: a client's sales team enters a target company, the studio researches it, scores it as a partner, drafts sales material, and lists the people worth approaching there.
We decide how this website and the studio itself handle personal data, so for that we're the data controller. For the research and drafts produced inside a client's workspace, we're acting on that client's instructions — they choose which companies to research and what to do with the result. We're settling the formal split of responsibility with each client in writing; if you need to know where that stands for a particular workspace, ask us.
There's no data protection officer. privacy@prospektor.ai reaches a person.
You sign in one of two ways, and there's no password to set either way — we never see one. With Google: Google tells us your email address, your full name, your given and family name, and the URL of your profile picture. We link to that picture; we don't copy or re-host it. Or with an emailed link: the studio sends a single-use sign-in link, good for fifteen minutes, to an address already on a workspace's allowlist — so typing a stranger's address into that form sends nothing anywhere. A session made that way knows only your email address; there's no name and no picture, because nobody told us one.
The emailed link leaves two small markers behind: a record that a link has been used, so it cannot be used twice, and a rate-limit marker holding a truncated hash of the requesting address rather than the address itself.
Your email address is your account identifier, and it's written onto the records you create so the workspace can show who did what:
Everything else in a workspace is the work itself: the target company, the website, any notes typed in, a point of contact if someone entered one, the research result, the drafted material, the progress log and later edits.
Support inside the studio is answered by AI, not by a person waiting on the other end. Your question and the conversation around it go to Anthropic like the studio's other model calls (section 08), and each conversation is saved to your workspace so you can pick it up again later.
We read them. The Prospektor team reads support conversations and the feature requests and feedback people send, to work out what to fix and what to build next — and we export the questions asked so we can look at them together. If you would rather ask us something without it being stored in the studio, email privacy@prospektor.ai instead. Support conversations and feedback notes are deleted with the workspace, like every other record in it.
There is a contact form at prospektor.ai/contact. It asks for three things — your name, your email address and your message — and that is all it collects. There is no hidden field, no tracker on the page, and no third-party anti-spam script reading your submission: the only spam control is an invisible decoy field a person never sees.
Your message is handled by Netlify's form service (section 08), which stores it and emails it to us. We read it to answer you, and that is the only thing we do with it. We don't add you to a marketing list from this form, and we don't pass it to anyone else; and writing to us does not create an account or put your company into the studio. Our basis for holding it is our legitimate interest in answering people who contact us.
It isn't deleted on a schedule — messages stay in the inbox and in Netlify's form store until we clear them out. Ask us and we'll delete yours — privacy@prospektor.ai.
That form is closed. Until 18 August 2026 there was an application form at app.prospektor.ai — name, work email, company, then nine questions about your business — from when Prospektor was an agency. It is gone, and that address now brings you here.
If you applied before then, we still hold what you sent: it was handled by Netlify's form service and reached us as an email. We read applications to decide whether we were a fit and to reply — that's all we ever did with them. We didn't add you to a marketing list from an application, and nothing here deletes itself. Ask us and we'll delete yours — privacy@prospektor.ai.
This section is for people who never signed up for anything. If you've been told, or suspect, that your name is in a Prospektor research result, this is the part that applies to you.
When a client researches a target company, the studio produces a list of the people worth approaching there. For each person it may hold:
Names, roles and profiles: publicly available web sources, found by an AI model running up to ten web searches during a single run. We don't take anything from behind a login. The studio is instructed never to invent a person: if it can't attach a credible source to a name, it returns the role with the name left blank — and may then ask a contact-data provider who holds that role, so a blank can be filled by a person that provider names, with their profile as the source.
Work email addresses are bought, one person at a time. For each named person the studio sends the name and the target company's web domain — or the LinkedIn profile URL, where it read one — to Tomba, then to Findymail, and if neither has one, through treg — a routing service that asks up to eight further contact-data providers one at a time, cheapest first, and tells us which one answered (all in section 08) — and gets back a verified work address or nothing. We don't buy lists, and we hold no address neither provider could verify. Personal email addresses and phone numbers are never requested and never come back. Two more questions go through treg the same way: for a role the studio could not put a name to, the role title and the company's web domain, answered with the name, title and LinkedIn profile of whoever holds it; and for a named person whose profile the studio did not find, the name and the web domain, answered with their profile link. A name is kept only when the person is at that company in that role. A user can type free-text notes into a run, though, so anything they choose to type is stored with it.
We rely on legitimate interests — ours and our client's interest in working out who to approach about a business partnership. What we hold is professional information, published in a professional context, about your job rather than your private life, and limited to what's needed to decide whether to contact you. Weighed against that, we think the intrusion is small. You're entitled to disagree, and the balance is yours to challenge: see your rights.
We don't write to people to tell them they've appeared in a run. Since 2 September 2026 we may hold a work address for you, but writing to everyone at the volume the studio produces would mean contacting far more people than any client ever approaches. This page is the notice instead. If you want to know whether you're in there, ask and we'll look.
The list is visible to the client's team inside their own workspace. It isn't sold, published, or shared with anyone else — except that a pitch containing it can be shared with a named person by link, which is covered in section 07; the shared copy carries the name, role and profile, never the email address. It's processed by the providers in section 08, and it isn't deleted on a schedule: see section 09.
If you'd rather not be in a decision-maker list, email privacy@prospektor.ai with your name and employer and we'll remove you. We won't ask you to justify it.
The front page of this site has one field: give us a web address, and we take a quick automated look at that company and say what we think they’d want us to find for them. Most of the time that’s a company looking itself up, deciding whether to buy. But anyone can type any address, so it may have been somebody else looking at you.
The reading itself, filed under the domain:
It’s kept so the same lookup doesn’t cost a second one. Most of it is read off pages you publish: we open your site — the homepage, and at most one more, usually an about or customers page if the homepage left the question open — and we may run a search or two beyond it. The model is instructed to write down nothing it can’t find on a page it actually opened, and to say so plainly rather than guess when a site won’t open.
And one thing that belongs in this list rather than in the small print: a reading can end up naming a person. We don’t go looking for anybody and we take no contact details — but an about page is the page whose job is introducing the people behind a company, so a fragment or an observation can carry a name that was published there. Where the business is one person, the reading is about them.
While the model reads your site, we also ask Clay about you — Clay Runner Inc., in the United States, who sell company data. What we send is your web address, or, if somebody typed a company name into the field instead of an address, that name, so it can be turned into an address we can look up. What comes back is what they hold about the company: headcount, location, industry, founding year.
We ask Clay nothing about any person at your company — the request has no way to carry an email address or a phone number, and none comes back. Nothing about the person who typed your address reaches them either, because we never had it to send.
Anything at all about the person who typed it. No name, no account, no IP address, no referrer. We can’t tell you who looked you up, because we never knew — and we can’t tell a company looking at itself apart from a stranger looking at it, for the same reason.
We rely on legitimate interests. The interest is answering the question somebody asked us — what a prospecting studio should go and look for on that company’s behalf — and not paying to read the same website twice. What we hold is a reading of pages that company publishes, about its business rather than anybody’s private life: no email addresses, no phone numbers, nothing from behind a login, and nothing a person browsing the site couldn’t see.
Here is the uncomfortable part, stated rather than buried. Most scans are companies looking themselves up, and those are unremarkable. Some are not, and then we hold a reading of a business that never asked for one. For most companies that’s information about a company and not about a person — but a one-person business is a person, so where that’s who you are, we treat the reading as being about you and you can have it removed without giving us a reason. You’re entitled to disagree with the balance either way: see your rights.
We don’t write to companies to tell them they’ve been looked up. We hold no way to reach them — the reading carries no address and no phone number — and we couldn’t tell them who did it or why, because we never stored that. This page is the notice instead. If you want to know whether we hold a reading of your site, ask us with the web address and we’ll look: it’s filed under the domain, so that’s the one question we can always answer.
Indefinitely. Nothing here expires on a timer, which is the same answer as section 09 gives for everything else. After about a month a reading counts as stale, and that changes one thing only: the next person to look up that domain still gets the reading we already have, and we take a fresh look behind them and replace it. A domain nobody looks up again keeps whatever we read the first time.
Into the studio, if that company becomes a customer: it’s what saves them entering their own details a second time. Nowhere else. We do not publish it, we don’t list or publish the companies that have been scanned, we don’t sell it, and there’s no page anywhere that can be browsed for it. It’s processed by the providers in section 08.
Then this is about you rather than about the company you looked up, and the answer is short. Typing an address into the field leaves us no IP address, no user agent, no referrer and no cookie — the endpoint reads none of them.
If you go on to a free run, that run holds the domain you typed, the reading of it, whatever you put in the box asking what you’re hunting for, the companies it found, our own token counts, and — if you asked for one — a report on a competitor you named. That report is a reading of that company’s public pages, and holds nothing about you. It’s addressed only by an unguessable link, and there’s no way to list runs or look one up by domain. If you ask us to email you the result, we store the address on that run, use it once for that mail, and never show it to anybody who has the link. That send puts you on no list and starts no sequence. The one box that would is beside it — product news and tips, unticked unless you tick it, and one click in any mail we send stops it.
This is a couple of minutes of automated reading, written to be shown back to a company so they can correct it when they sign up. It isn’t a report about you, we don’t stand behind it as fact, and it may well be out of date — see the paragraph above about how a domain nobody looks up again is never re-read.
If you’d rather we didn’t hold a reading of your site, email privacy@prospektor.ai with the address and we’ll delete it. One honest caveat, because we’d rather say it than have you find out: there’s no suppression list, so if somebody types that address into the field afterwards, a new reading gets made. We’ll say so when we reply.
This section is also for people who never signed up for anything, but you got here a different way from section 04. Nobody researched you. Somebody who already knows you brought their own contacts into their workspace, and you were among them.
A member of a client's team can import who they know. Three doors, and each one is a file or a list that is already theirs:
Connections.csv anyone can download of their own connections. Per connection we hold: name, company, position, the public LinkedIn profile URL, and the date the two of them connected. They can add the messages.csv from that same download, which adds nothing but the counts and dates described below.There's no integration behind any of this. We don't connect to LinkedIn, we don't read anyone's mailbox, and we hold no key to either. A person exports their own data and uploads the file, or types the names.
We don't hold your email address. A LinkedIn export usually contains some; we drop them at import and never store them. An address book export contains far more — addresses, phone numbers, postal addresses, birthdays, whatever private notes the owner kept — and we read the three professional fields out of it and discard the rest at the door. Both rules are covered by automated tests, because they're the sentences here we'd least like to be wrong about.
We hold none of your messages. Not the words, not the subject line, not the thread title — nothing anyone actually wrote. If a member imports their LinkedIn message archive, what survives about you is two counts and two dates: how many messages they sent you, how many came back, and when the first and the last one was. That is how the studio tells someone they really talk to from a name on a list, and it is the whole of what we keep about a conversation. Beyond that: only that they know you, and where you work.
One thing. When that team prepares a pitch to a target company, the studio looks for a warm path — someone in their network who might be able to introduce them there — and shows them the route: you know this person, who could reach that person at the company you're approaching. It uses public web sources to check the connection it's about to suggest is plausible.
An AI model does that matching, so a digest of the network — names, companies, positions, public profile URLs, the member's own notes and, where they imported it, how often and how recently the two of them have spoken — is sent to Anthropic as part of a prompt. Your email address isn't in it, because it was never stored.
Prospektor will never contact you because of this. Nothing we send is triggered by an import — no notification, no marketing, no mail of any kind. If anyone reaches out, it's the person who already knows you, on their own account, the way they could have anyway.
Legitimate interests again: a team working out who could introduce them to a company they want to work with. What we hold is professional — your job, your employer, a public profile — and it's information the person who imported it already had about you. We've written the balancing test down rather than just asserted it, and we'll show it to you if you ask.
This page. We don't write to imported connections to tell them they're in a workspace, and the reason is the fact above: we deliberately don't keep your email address, so notifying you would mean going out and collecting the very thing we refused to store, for thousands of people, in order to announce that we hold less. That's a worse outcome than the one it announces, so this section is the notice instead. If you want to know whether you're in there, ask and we'll look — across every workspace, by name.
That one team, inside their own workspace. It's never shown to other customers, never carried on a shared pitch, never sold, never pooled with anyone else's. The member who imported it can remove that import whenever they like, and removal deletes every entry it brought — only they can do it, since it was theirs to bring. It goes when their workspace goes. Like everything else here, nothing expires on its own: see section 09.
Email privacy@prospektor.ai with your name and employer and we'll search every workspace and take you out. We won't ask you to justify it, and we won't ask the customer's permission — we act, then tell them we did.
A finished pitch can be shared as a link. To read it you sign in — with Google, or with a single-use link the studio emails you — so you end up with an account too. You came for someone else's document, and this is the part you're least likely to expect:
We record your email address, your name, when you first opened the link, when you last opened it, and how many times you've opened it. The team that sent the link can see all of that, listed under their shared links.
So the sender can tell whether you read it once or five times, and when. If you'd rather they couldn't, don't open the link — or email us and we'll remove your row.
A share link shows one pitch and nothing else. Everything else in the studio refuses a share-link account, and an account created this way doesn't get a workspace of its own. A link can be revoked, which stops it working immediately for everyone including people who already read it — but the record of who opened it survives revocation.
No advertising. No session recording. No fingerprinting. No third-party trackers. One thing measures anything at all, it is named below, and it is off until you allow it — the Cookies link in the footer of every page is where you answer, and where you take the answer back.
It sets no cookies at all. It keeps three things in your own browser, and they never leave it:
pps-consent — your answer to the cookie notice, with the date you gave it, so we can show you were asked and so we stop asking. Kept 12 months, then we ask again.prospektor.scan — the scan you asked for, carried from the front page into the checkout form so you don't fill the same thing in twice. Gone when you close the tab.prospektor.goal — the sentence you typed about what you want the studio to find, so a reload doesn't lose it. Gone when you close the tab.Netlify Real User Metrics is the optional one, and the only measurement anywhere on this site. It puts nothing on your device — no cookie, no browser storage of any kind — but as you leave a page it posts that page's load timings, along with your IP address, to Netlify, who host this site. It does not run unless you allow it. The tag is taken out of the page before the page reaches you, and your consent is the only thing that puts it back.
The studio sets three cookies, all of them functional:
pps_session — signed, HttpOnly, Secure, SameSite=Lax, expires after 12 hours. Your name, email and picture URL sit inside the cookie itself; there's no session database behind it.pps_oauth_state — 10 minutes, protects the sign-in against cross-site request forgery.pps_share — 10 minutes, remembers which share link sent you to the sign-in page.The last two are cleared as soon as sign-in finishes. It keeps four more things in your browser, which never leave it:
pps-theme — this device's light/dark choice. Not stored against your account, so nobody, us included, can read how you like to look at the studio. Until you clear site data.pps-library-sort — how this device likes the library sorted. Until you clear site data.pps-consent — your answer to the studio's own cookie notice, with the date. 12 months, then it asks again.pps-example-gone — that you removed the example pitch from your library, so it stays removed. Until you clear site data.That is the whole list, on both properties, and it is the same list the code works from: the studio's own Cookies link shows it, item by item, and is where you change your answer there.
Both this website and the studio serve their own fonts. Google receives nothing when you load a page on either — that was always true here, and it has been true of the studio since 18 August 2026, when it stopped loading fonts from Google's CDN.
| Who | What reaches them | Why |
|---|---|---|
| Anthropic | As prompts to the Claude API: the client's brief, the target company name and website, the user's notes, and any brand-voice examples. The model runs up to 10 web searches per run through Anthropic's search tool. | The research and the drafting |
| Netlify | Hosting, the code that runs the studio, all stored data, DNS for prospektor.ai, and anything you send through the contact form on this site (your name, address and message). Separately, and only if you allow it: the load timings of pages you visit on this website, with your IP address (Real User Metrics). | Infrastructure |
| Sign-in through OAuth, and nothing else. (Google Fonts used to receive the IP address of anyone loading a studio page; since 18 August 2026 the studio serves its own fonts and Google receives nothing from a page load.) | Authentication, typography | |
| Clay | The web address somebody scanned, or the company name they typed — and nothing else. No identifier, no session, no IP address, and nothing about the person asking, because the scan never knew anything about them. What comes back is about the company, never a person: headcount, location, industry, founding year. It is Clay Runner Inc., based in the United States. | Company facts on a scan |
| Tomba | For each named person in a research result: their name and the target company's web domain, or their LinkedIn profile URL — and nothing about the client, the user or the workspace. What comes back is that person's work email address, whether it was verified, and a confidence score; if they have nothing, nothing. It is Tomba Technology Web Service LLC, based in the United States, which says its data is stored and processed in the European Union. | Work email addresses for decision-makers |
| Findymail | The same identifiers as Tomba, only when Tomba had nothing — and, for a role nobody could be named for, the role title and the company's web domain, which comes back as one person's name, job title and LinkedIn profile. It is Peanuts SaaS Studio, trading as Findymail, whose servers are in Finland and which says it stores and processes all its data inside the European Union. | Work email addresses, and who holds a role |
| treg | The same identifiers as Tomba's row — a person's name and the target company's web domain, or their LinkedIn profile URL — only when Tomba and Findymail both had nothing; and, first rather than last, a role title with the company's web domain for a role nobody could be named for, or a name with the web domain for a named person without a profile. treg holds no request or response bodies of its own; it passes the identifiers to one of Hunter, Icypeas, LeadMagic, LeadsForge, Aviato, CompanyEnrich, Fiber AI, Ocean.io, Lusha, Apollo or People Data Labs, one at a time, until one answers, and tells us which one did. What comes back is a work email address and whether it was verified, or a person's name, title and LinkedIn profile. It is operated by Superdesign and hosted in the United States. | Work email addresses the first two providers could not find; who holds a role; a named person's profile |
| Stripe | Your email address, your card and billing details, and the company and goal you gave at checkout. Checkout runs on Stripe's own pages, so your card number never reaches us. | Payment |
| Postmark | Your email address and the message we send you — the welcome mail when your workspace is created, a sign-in link when you ask for one, and internal notices to us. It is an ActiveCampaign company, based in the United States. Transactional only: there is no list, no campaign and nothing to unsubscribe from. |
We do run marketing email. If you have an account, subscribed to our newsletter, or ticked the box when you asked for a free run, we send product news, new features and occasional offers alongside the transactional mail — your receipt, your welcome, your sign-in. That goes through our email provider above and through a newsletter platform, and we can see whether a message was opened and which links were clicked. Every marketing message carries a one-click unsubscribe, and unsubscribing never stops the transactional mail, because that is how having an account works.
Not used, so you can rule them out: no CRM integration, no session recording, and no advertising vendor.
Anthropic, Netlify, Google, Clay, Tomba, treg, Stripe and Postmark all process data outside the European Economic Area, mainly in the United States; Findymail processes only inside it. That's an international transfer, and it needs a documented safeguard — standard contractual clauses, or each provider's own data-processing terms. We haven't completed that documentation. Ask us and we'll tell you where it stands.
Nothing expires on its own. There is no automatic deletion and no retention limit in the product today. A run from January is still there in December unless somebody removed it.
What a user can do from inside the studio:
Two things that catch people out:
A reading of a website from the front-page field is its own case, because it is filed under the domain rather than under an account: it is kept indefinitely like everything else, we will delete one on request, and deleting it does not stop a new one being made if somebody types that address in again. Section 05 is the full answer.
There is no self-service account deletion and no export button. Both exist only as a person doing it by hand, on request. We'd rather say that than imply a button that isn't there.
If you're in the UK or EU, you have the right to see what we hold about you, correct it, have it deleted, object to us holding it at all, get a copy in a portable form, and complain to your national data protection authority. We apply the same to everyone wherever they are, because running two standards is more work than running one.
To use any of them, email privacy@prospektor.ai. Tell us what you want and enough to find you: for section 04 or section 06, your name and employer, since those are the only handles we have on you; for section 07 or an account, the email address you signed in with; for section 05, the web address itself, because a reading is filed under the domain and under nothing else. If we can't identify you from that, we'll ask for more rather than refuse.
What then actually happens: a person goes and does it by hand. We'll confirm when it's done, and we'll come back to you within a month at the outside — normally much sooner. Where the data sits inside a client's workspace, we'll act on it and tell that client.
Every stored record is namespaced by client, and the workspace is resolved fresh from your signed-in email on every single request. One client's data isn't addressable from another client's session, and removing somebody from a client's allowlist takes effect immediately rather than at their next sign-in. That behaviour is covered by automated tests.
Signing in at all needs one of three things: being on a client's email allowlist, an invitation to create a workspace, or a live share link. A share link gets you sight of one pitch — every other part of the studio refuses it. Speaker notes, compliance findings and the internal progress log never leave the workspace they were made in.
No system is perfect, and we'd rather describe what we actually do than promise it can't be broken.
If we change what we collect, we change this page and move the date at the top. Where it's a material change for people with accounts, we'll say so in the studio rather than hope you re-read this.