The algorithm is not secret. Look up MX records, connect on port 25, send EHLO, MAIL FROM and RCPT TO, read the code. Anyone can implement it in an afternoon, and the result they get depends enormously on where the connection comes from.
Port 25 is blocked outbound by most residential ISPs, because it was the spam vector for two decades. AWS, Google Cloud and Azure all restrict it on new accounts and require a request to open it. That is the first filter on who can do this at all.
Getting the port open is not the same as getting an answer. Gmail and Microsoft 365 refuse or tarpit probes from many cloud IP ranges regardless of what the connection says. A verifier running on default cloud infrastructure gets refused connections on two of the largest mailbox providers in the world.
The result is that identical code produces different coverage depending on the IP behind it. Not different accuracy on the addresses it resolves, different numbers of addresses it can resolve at all, which is why coverage is the figure worth comparing.
What this costs to run properly
Addresses with a history. A newly allocated range is treated as suspicious by receivers who have never seen it, and building a reputation takes weeks of consistent behavior.
Correct forward and reverse DNS on every probe host. Deutsche Telekom is strict enough about this to refuse connections without it, and several other large receivers weigh it heavily.
Enough addresses to spread the load. Microsoft throttles per connecting IP, so probing at any volume from a single address means most of a batch comes back as a rate limit rather than a verdict.
Rate discipline per provider. GMX and Web.de share an operator and a budget. Free is stricter than Orange. Probing everything at the same speed wastes most of the results.
Where this argument costs us something
The short version
- Compare verifiers on coverage, because that is where infrastructure differences show up.
- Be suspicious of a very cheap verifier with very high claimed resolution on Gmail and Microsoft domains.
- Understand that a rise in unknown results can mean a receiver changed its mind about a range rather than anything about your list.
Questions people ask
Why can I not verify addresses myself?
Your ISP almost certainly blocks port 25 outbound, and if you get past that, the large providers will refuse probes from a residential or cloud address with no sending history.
Does this affect accuracy or coverage?
Coverage, mostly. A blocked probe produces no answer rather than a wrong one, which is why we report it as unknown and do not bill it.