Frequently asked questions

Call your bank on the number printed on your card and ask for the fraud department. Say the words "fraud" and "recall request" and ask for the reference number. For a card payment, ask about a chargeback. For a bank transfer, ask about recall. For cryptocurrency, contact the exchange the funds went through — they can sometimes freeze a receiving address. Then file a report: IC3.gov in the US, Action Fraud in the UK, your national cybercrime unit elsewhere. Do all of this today. Recovery windows close in hours, not days, and 94% of victims recover nothing — the ones who do are overwhelmingly the ones who called the same day.

Almost certainly not, and the timing is the tell. Recovery scams target people who have just been defrauded, because those people are hopeful, ashamed and already primed to believe an authority figure. The caller typically claims to be the FBI, Interpol, a government agency or a specialist recovery firm, knows details of your loss, and asks for an upfront fee, a tax payment or wallet access. No legitimate law-enforcement agency ever contacts a victim to offer paid recovery. It is frequently the same criminals, or resellers working from the same victim list.

No. You lose the evidence trail, and if money has moved you lose any chance of it being returned through their cooperation. Take screenshots of everything first — the profile, the number, the photos, the links, the wallet addresses — because accounts like this are deleted once they stop producing. Then stop replying. There is no version of this where an extra message helps you.

No, and no tool honestly can. PrivateCue reads the shape of a conversation, not the identity behind it. Anyone claiming to trace a scammer, recover funds for a fee or reveal a person's real identity is running the second stage of the same fraud.

It is analysed inside your own browser and never sent to us. There is no upload, no API call carrying your text, and no copy on a server, because the analysis is performed by JavaScript running on your device. Close the tab and the text exists nowhere but where you copied it from. This is not a policy promise — it is the architecture, and it is why the tool still works with your network disconnected.

Metadata about the fact that an analysis happened: a timestamp, a risk band, a numeric score, a word count and a country. Not the conversation, not an excerpt of it, not the names in it. Your account also holds your email, a salted password hash, your plan and your payment records. We deliberately cannot show you your past conversations, because we never had them.

There is nothing to train on. No conversation text reaches us, so none can be used for a model, for tuning, for research or for resale. The rule library is written by hand from published fraud guidance and survivor reports, and you can read all of it on the taxonomy page.

Three first-party cookies and no third-party ones. A session cookie keeps you signed in and a CSRF token stops a form on another site acting as you; both are 30 days and both are strictly necessary. The third is an anonymous visitor id — a random string that identifies a browser and nobody — which counts the pages you visit and, if you are not signed in, how many of your free analyses you have used. It lasts a year, and clearing it resets that count. For pageviews we record the path, a coarse device type and the referring hostname only, never the full referring URL. We do not store IP addresses, user-agent strings, query parameters or any advertising identifier, and we use no ad network. Pageviews and anonymous usage rows are deleted after 90 days, security events after 30.

Yes, by email to support@privatecue.com, and deletion removes the account, its sessions, its usage rows and its payment records. Under GDPR you can also request a copy of what we hold, which will be the metadata described above — we cannot return conversation text we never received.

The honest answer is that it is accurate at the thing it does and silent about everything else. It matches a conversation against a documented library of fraud scripts, and every point it awards traces to a rule you can read. It is not a classifier trained to a published precision figure, and we will not invent one. Its value is that it surfaces the specific pattern that fired and explains why that pattern is dangerous, so you can judge it yourself.

Yes, and the design assumes it. A genuine investment adviser discussing a platform will trip some of the same indicators as a pig-butchering script, because they use overlapping vocabulary. That is why the report shows the matched phrases rather than only a verdict, why protective patterns reduce the score, and why a high score is presented as "this pattern is dangerous — verify through a second channel" rather than "this person is guilty".

No. A low score means no known indicator fired, which is not evidence of honesty. The most damaging scripts in this space deliberately stay clean for months — building a relationship with no mention of money at all — before the first request appears. Re-run the analysis as the conversation develops, and especially the moment money, investment, crypto or an emergency is first mentioned.

Not reliably, and it does not claim to. It can flag the behavioural patterns that accompany an AI-operated romance scam — impossible scheduling, instant intimacy, scripted deflection of a video call, a sudden platform migration — but it cannot authenticate a voice, a face or an image. For a video call, the only meaningful test is an unpredictable live challenge that you invent in the moment.

The rule library is written for English. A conversation in another language will still be analysed, and the link checker works on URLs regardless of language, but the phrase patterns will not reliably match translated text and the score will understate the risk. If the conversation is not in English, treat a low score as "not measured" rather than "not dangerous".

No. It analyses an interaction you are party to, and the terms prohibit using it to surveil, harass or build a dossier on a private individual. There is no person-search and no bulk lookup, by design.

Not for the first one. Paste a conversation and read the full report — score, script, every line that triggered it — without signing up or paying. We ask for a free account from the second analysis onward, and that is the whole gate. It used to sit in front of the first one as well, and that was a mistake: the analysis runs in your browser and costs us nothing to give you, so the only thing the signup form was protecting was a count.

Because a free account is what lets the free limit be counted per person rather than per browser, and it is what makes the two paid capabilities possible — exporting a report so you can show it to someone, and keeping history. The current free allowance is shown on the pricing page, and no card is required. Signing in does not upload your conversation: the result is rendered locally and then the score is recorded, and that is true whether you have an account or not.

With a cookie on that device, and it counts one thing only: how many analyses this browser has run. Clearing your cookies resets it, and blocking cookies means we cannot count at all, so we let the analysis run rather than punish a privacy-strict browser for a nudge. That is a soft count and we are not going to pretend otherwise, or start fingerprinting visitors to defend it. The analysis itself is in your browser either way, so anyone determined to bypass the gate could already do it by running the engine offline.

Because the person who most needs this is the one who cannot afford it, and the person in immediate danger should never be shown a signup form or a paywall before an answer. The first analysis needs no account at all, and a free account then receives the current free allowance shown on the pricing page. The paid tiers exist for people who need to act on the answer repeatedly, or on someone else's behalf.

One click of Analyse conversation, on a conversation of at least a few exchanges. Re-running the same text after you have edited it counts again, because it is a different input. Loading a sample counts, so you can try the product before you trust it.

Upgrading applies the paid plan allowance immediately; analyses already used under the free plan do not consume that new allowance. Downgrading or cancelling takes effect at the end of the period you have already paid for.

Yes. Family supports five adults with separate accounts and private histories. The organiser sends invitations and manages membership, but cannot read another adult’s reports or history unless that person explicitly shares them.

If money has already gone. It produces an evidence package built for your bank and for law enforcement, a country-specific filing roadmap, and the wording that gets a recall request taken seriously. It is free and open to anyone, including people with no account, because the recovery window closes in hours and we are not going to put a signup form or a price in front of that.

Subscriptions: within 14 days, by email, for any reason or none. The Recovery Pack costs nothing, so there is nothing to refund — and nothing that can ever be taken from you.

The Recovery Pack is free, forever, so the worst week of your life never meets a price here. For subscriptions, email support@privatecue.com and say so — we would rather you used the tool than not.