How we know what we say we know

Everything on this site is checkable. This page tells you exactly how the data was built, what the confidence levels mean, and — importantly — what is still wrong with it.

Sourcing

Every record is a real, currently-running programme published by a government, public body or well-known institution. Each one carries an official source_url on the funder's own domain and, where we quoted it, a verbatim source_snippet from that page.

Rules we hold ourselves to, and which you can check us against:

• No invented URLs. If we could not find the exact deep link, we link the official landing page.
• No guessed amounts. If the amount depends on your circumstances, the fields are left empty and a note explains why — we never fabricate a plausible-looking figure.
• No ended programmes. Schemes that closed were deliberately excluded during curation.
• Every record is dated. last_verified_at is when a human last looked.

The two verification states

We show the difference on every card rather than flattening both into one confident-looking badge. A site that claims uniform accuracy across 2,216 programmes in 25 countries is lying to you.

How matching works

Your answers are evaluated against nine published rule types: geography, work or life status, student status, age, household income, dependent children, residency status, length of residence, and housing tenure. Each programme lands in one of three buckets:

• Eligible — you pass every rule the record publishes.
• Needs one more answer — you pass everything we can test, but one rule needs a detail you have not given. We show you the exact question.
• Not eligible — you fail at least one rule, and we name it in plain language.

A record with no published restriction on a given attribute passes that attribute. This is deliberate: it is better to surface a programme you might not get than to silently hide one you would.

What the big number means (and does not)

The headline figure is the sum of published annualised maximums for programmes in your eligible bucket only. Monthly amounts are multiplied by twelve; one-off amounts are counted once.

• It is an upper bound of published ceilings, not a prediction of your payment. Most means-tested schemes taper — you get the maximum only at the bottom of the income range.
• Programmes with no published amount contribute zero. Your real entitlement is almost certainly higher than the number shown, not lower.
• Loan and credit-facility ceilings are excluded from the headline and reported separately, because borrowing capacity is not income.

Privacy

The eligibility check runs entirely in your browser. Your answers are never transmitted anywhere — there is no server to transmit them to. This site is static files. If you use the "copy link" feature, your answers are encoded in the URL you choose to share, and nowhere else. There are no accounts, no cookies for tracking, and no analytics that identify you.

Known limitations

The honest list. If any of this changes, this section changes with it.

1. Income bands are approximations. Band thresholds are fractions of estimated median household income per country, not official poverty lines. Enter an exact income when the wizard offers it and matching gets sharper.
2. 1,462 of 2,216 records have no published amount. They are real programmes; the amount simply depends on circumstances the official body calculates.
3. 1,201 records are not human-verified. Rules may have moved since curation.
4. Snapshot, not a live feed. The data is a curated snapshot dated 2026-08-12. There is no scraper re-checking these pages daily.
5. Regional coverage is uneven. National schemes are well covered everywhere; city-level schemes are covered for major cities only.
6. Not legal or financial advice. We describe published criteria. Only the named authority can decide your case.

Corrections

Every programme page has a "report it" link that opens a public GitHub issue with the programme and page pre-filled. Public tracker, public fix history — no feedback form that disappears into nothing.

The data is yours

The full dataset is open. Use the JSON API or plug it into an AI assistant over MCP, or take the whole repository from GitHub.