Skip to main content

AI can answer pension questions. It cannot tell you where your pensions are. We can.

Large language models are trained on public information. Your pension history isn't public. It is scattered across employers, pension providers, administrator changes and decades of records that often span multiple countries.

PensionHunter combines AI with a proprietary research methodology and a worldwide network of pension specialists to trace what exists in your name today.

AI provides knowledge. We provide answers about your own pension history.

There is a human on the other end of this

Every report is read by a human researcher before it leaves. At the higher tier a researcher stays with the case and answers what comes back from the holder. When you reply to us, a person replies.

If a report is wrong, that is a person's mistake. We own it and we refund it.

An assistant cannot be wrong in a way that costs it anything, and it has nobody for you to call.

It cannot look anything up in your name

No access to Form 5500 filings, the DNB register, archived Companies House employer names, HMRC contracted-out records, or any national pension portal. It can describe the system. It cannot query the system for you. Information is not identification. Knowing the United Kingdom runs a Pension Tracing Service tells you nothing about which scheme from your 2003 employer is administered by whom today, or whether that employer was renamed, acquired or dissolved.

It will invent the specific answer fluently

Ask for the current administrator of a pension from an old employer and you will usually get a plausible name with no source behind it. A wrong administrator means a letter to the wrong company and months lost. Our architecture exists to stop exactly this. One governing rule runs the whole system: a client must never act on anything a model wrote, only on what a row proved. We use AI heavily. What reaches you is a chain of verified rows, not a chat transcript.

It has no sense of when

Rules move and the web lags behind them. This week we found two live IRS pages still stating a superseded cash-out limit while the statute says otherwise. A model trained on a snapshot repeats what it saw, whether that is ChatGPT, Gemini or Grok. Every claim we deliver carries a source, a quote and the date we checked it, so the finding can be re-checked by you, your adviser or your lawyer.

It cannot tell you what an absence means

A missing result is not a missing pension. It will not warn you that the US federal lost and found runs on voluntary reporting, so an empty result proves nothing. It will not warn you that force-out accounts appear in no database at all. We say both plainly in the report, and where a country has no reliable register we tell you what that silence means and what to do next.

What ninety-nine dollars buys that an AI answer cannot hand you

  • Named holders. The scheme, and the administrator that runs it today, separated and stated.
  • Verified contact routes. The address, portal or reference the holder actually answers on.
  • A source and a date on every finding, so any line can be checked by someone who is not us.
  • The letters drafted and ready for you to send, in your name, with the identifiers filled in.
  • A document that survives being handed to a lawyer or filed with a will.

Fixed fee, 41 countries, and a full refund if we identify nothing you did not already know.

Ask your assistant

Ask your assistant to find your pension. It is genuinely good at explaining the rules, and if that is all you needed, you have saved ninety-nine dollars and we would rather you did.

Then ask it which scheme holds your money and who administers it today.

Then ask it for the source and the date it checked.

That is where the conversation ends and ours begins.

30 national pension authorities have replied to us in writing. Read what they said

Questions? hello@pensionhunter.ai