Reading a result
For name checks. Wallet, email and IP checks are explained on Methodology.
What does a hit mean?
A hit is a possible match: a name on a government list that looks like the name you checked. It's a reason to look closer, not a finding that your customer is sanctioned. Most hits on common names turn out to be different people.
"No match" means no match on the lists named in the result, nothing broader.
What does the match percentage mean?
How alike the two names are written, from 30% to 100%. It says nothing about whether they're the same person: "John Smith" against a listed "John Smith" is 100%, and there are thousands of John Smiths.
To decide whether it's the same person, compare the date of birth, nationality and type (person or company) shown under each hit.
Which match setting should I use?
- Recommended (40%): suits most onboarding and payouts. Start here.
- Broad (30%): catches more spelling and transliteration variants, but gives you more to review.
- Strict (70%): only very close names. Fewer alerts, but you may miss a respelled name.
- Exact only: fuzzy matching off. Same words in any order, case or punctuation.
What do match, differs, partial and "can't compare" mean under a hit?
- Match: what you entered agrees with the list record.
- Differs: it disagrees, for example a different birth year. Strong sign it's someone else.
- Partial: agrees as far as the record goes, for example the list only has a birth year, or a range of years.
- Can't compare: you didn't enter it, or the list doesn't have it.
Entering the date of birth and nationality when you run a check is the single best way to resolve hits quickly.
Does PayPerScreen make the compliance decision?
No. It's a data service: it tells you what matched and how closely. Whether to onboard, pay out or report stays with you and your own policy.
How up to date are the lists?
Every list is downloaded from the authority that publishes it once a day. Lists we check shows each list's record count and last update. If a download fails, the previous copy stays in use.
Reading the AI data comparison
The optional comparison of a possible match's details, in your case queue. It shows how closely the data lines up. The decision is always yours.
What is the AI data comparison?
On a name alert, an AI model lines up each possible match with what you entered: date of birth, nationality, person or company, and the other names the listed person is known by. It shows how closely those details agree as a traffic light, with the facts behind it.
It describes data. It doesn't say whether your customer is the listed person, and it never clears or escalates an alert. It doesn't change the hits or the match percentage either. Your reviewer looks at the data and decides.
What do the colours mean?
The details that could be compared line up with the listed record, for example the full date of birth and the nationality are the same.
What it doesn't tell you: that it's the same person. Data can line up by coincidence, and the list record may be incomplete. Your reviewer weighs it under your policy.
There's too little to compare, or some details line up and others don't. Most often a detail is missing, such as the customer's date of birth.
What helps: the comparison lists the missing data that would allow a fuller comparison. Adding it and screening again usually gives a clearer picture.
One or more details clearly don't line up with the listed record, for example a different birth year, or your customer is a person and the listed record is a company or a ship.
What it doesn't tell you: that it's safe to clear. Check the differences against the comparison table, for example a typo in the date you entered, before your reviewer decides.
An alert has several possible matches. How is its overall colour set?
Each possible match gets its own colour. The alert shows the closest one:
- If any match is red, the alert is red.
- The alert is green only if every match is green.
- Anything else is amber.
This rule is fixed in our software, not left to the AI, so one record whose details line up can never be hidden by several that don't.
What do strong, some and limited evidence mean?
Each match also shows how much data could be compared:
- Strong evidence: a decisive detail could be compared, such as a full date of birth, or a person compared with a company.
- Some evidence: some details could be compared, but not all.
- Limited evidence: little beyond the name could be compared. Read the colour with that in mind.
How do I read the observations and missing data?
Each match comes with one to four short observations, each naming the detail it's based on. Check each one against the comparison table for that match: if an observation says the birth years differ, the table should show differs on date of birth.
The summary line describes how the details line up across the whole alert. The missing data list names information that would allow a fuller comparison. Neither tells you what to decide.
Why do I see amber so often, and how do I get a clearer picture?
The AI is told to show amber rather than guess. The usual causes:
- No date of birth or nationality entered. Collecting it and screening again resolves most ambers.
- The list record is thin, for example only a birth year, or none. The authority's own site may have more detail; use the record ID shown with the hit.
- The match is on another name the person is known by, not their main name.
- A very common name with nothing else to compare. An ID document gives more to compare.
Can the AI comparison be wrong?
Yes. It only sees what you entered and what the list contains, and it can misread them. Typical traps:
- A typo in the date of birth you entered can make aligned details look different.
- People can hold more than one nationality, so a different nationality alone isn't decisive.
- Lists sometimes record an approximate or wrong date of birth.
That's why it describes and your reviewer decides. If a comparison looks wrong, tell us: we use those cases to improve it.
Is the comparison recorded for my auditors?
Yes. It's saved with the check in your audit trail: the colours, observations, which AI model made it, when, and who asked for it. Your reviewer's own clear or escalate decision is recorded separately, next to it, so an auditor can see what data was shown and who made the decision.
What does it cost?
10¢ per alert, from your prepaid balance, however many possible matches the alert has. You're only charged if a comparison comes back. Opening the same comparison again is free.
What does the AI see?
Only the details you screened (name, date of birth, nationality, type) and the matching list records. Nothing else about your customer. It's processed by Google's Gemini service under terms that don't allow Google to train on it. See privacy.
Why can't I compare some older alerts?
The comparison needs the full details of each possible match, which we've saved with name checks since October 31, 2026. Alerts from before then don't have them. Run the check again to compare it.
Red: the details line up. Amber: mixed, or something's missing. Green: the details differ. The colours describe the data, nothing more. Your reviewer makes the decision, and it's recorded as theirs.
Paying
Do I need an account or a subscription?
No. Each check is 2¢, paid in USDC on the Base network. You can pay per check from your wallet, or load a prepaid balance once so checks come straight off it. Visitors also get 3 free checks a day on Try a check.
Am I charged if a check fails?
No. You're only charged when a result comes back. A check that errors costs nothing.
Can an AI assistant or my own software run checks?
Yes. Software can call the API and pay per request; see API docs. AI assistants like Claude can use our MCP server; see Use it with Claude.
Your data
What do you keep, and for how long?
A record of each check, so you or your auditor can verify it later with its reference. We keep check records for 5 years, the usual record-keeping period for anti-money-laundering checks. The details are on privacy.