Proofwire vs NumVerify
NumVerify is the cheapest well-known phone validation API, and for format checking it does the job. The gap opens where a number is structurally perfect but not actually reachable — and in how APILayer, its parent, handles billing.
How to read this page. 9 dimensions compared: we come out ahead on 7, behind on 1. Every claim about NumVerify links to their own documentation or to public reviews. We have not run NumVerify through our own benchmark, so you will not find a head-to-head accuracy number here — comparing our measurement to someone else's would not be a comparison.
Choose NumVerify if
- You only need format and country validation, and you know that is all you need.
- Your budget genuinely will not stretch past a few euros a month.
- You are validating a signup form where a wrong line type costs you nothing.
Choose Proofwire if
- You act on the answer — dialling it, texting it, or deciding whether to trust a signup.
- You need to know whether a number is in service, not just well formed.
- You want to be able to cancel without writing a letter thirty days in advance.
Side by side
| Dimension | NumVerify | Proofwire | Source |
|---|---|---|---|
| Independent accuracy measurement | 71% in a 2026 third-party benchmark of ten tools | Not measured by that benchmark. Our own dataset is published with its scope and a command to reproduce it. | Cleanlist, 2026 phone validation benchmark |
| HLR / live network check | None | Tier-2 carrier probe when the cheaper tiers leave the verdict uncertain | Cleanlist, 2026 phone validation benchmark |
| Carrier data freshness | Described as stale, on a months-long refresh cycle | Every signal carries its age, and stale evidence is automatically discounted in the score | Cleanlist, 2026 phone validation benchmark |
| Line type on US numbers | Reported VoIP as mobile in roughly one case out of four | Returns a confidence figure alongside the line type; below 0.7 it is labelled a hint, not a fact | Cleanlist, 2026 phone validation benchmark |
| Regulator-reserved ranges | Not detected — the numbers are structurally valid | Ofcom, NANPA, ACMA and others, caught in the free tier and never billed | — |
| DNC registry | No integration | Checked where a registry exists; returns null rather than false where one does not | Cleanlist, 2026 phone validation benchmark |
| Headline price | $14.99/month for 5,000 lookups, 100/month free | €29/month for 6,000 credits. Most lookups cost 1 credit; only live network probes cost more | Cleanlist, 2026 phone validation benchmark |
| Cancellation | Customers report auto-renewal without warning and a requirement to cancel in writing 30 days ahead | One authenticated API call, effective immediately, with a confirmation reference | Customer reviews on Trustpilot |
| Inconclusive results | Billed like any other lookup | Never billed | — |
Where NumVerify is genuinely the better buy
At roughly $0.003 a lookup it is cheaper than we are per call, and if all you need is "is this string a plausible phone number in this country", it answers that correctly and quickly. Plenty of signup forms need nothing more. Paying us more for evidence you will never look at would be a bad trade, and we would rather say so than pretend otherwise.
Where the price stops being the relevant number
The moment you act on the answer, a wrong verdict costs more than the lookup saved. An SMS to a landline is wasted spend. A dial to a disconnected number is wasted agent time. A call to a number on a Do Not Call registry is a compliance exposure. At that point the question is not what a lookup costs, it is what a wrong lookup costs — and the third-party benchmark that put NumVerify at 71% is measuring exactly that.
The billing complaints are about APILayer, not the data
NumVerify is operated by APILayer, whose Trustpilot reviews contain a recurring pattern: auto-renewal with no advance warning, a requirement to cancel in writing thirty days before renewal, attempts to charge expired cards, and support that customers say went unanswered for weeks. This is worth reading yourself rather than taking from us — the link is above. Our own cancellation is one API call, and the credits you already bought stay spendable afterwards.
Questions
Have you benchmarked NumVerify against your own dataset?
No. Every accuracy figure on this page for NumVerify comes from an independent 2026 benchmark, cited above. We publish our own results on our own dataset, with the scope stated. Comparing our measurement to someone else's would not be a comparison, and we are not going to do it just because the number would flatter us.
Can I migrate without rewriting my integration?
The request shape is similar — a JSON body with the number — but the response is richer, and the important difference is that our verdict has three states rather than two. You have to decide what your code does with unknown.
Check it yourself
Do not take our word for any of this. Run our public benchmark with npm run bench, read the labelled dataset it ships with, or try the free tools without an account.
Think something here is unfair or out of date? Tell us and we will correct it.