Are email verification benchmarks trustworthy?
Only when you know the denominator. ZeroBounce claims 99.6% accuracy and Clearout 99.73%, counted on the addresses each tool chose to judge. Hunter's benchmark of 3 November 2025 counted every one of 3,000 real business addresses and measured 15 verifiers at 31.2% to 70.0%. Read any accuracy figure beside coverage, the share of your list that got a firm verdict.
The short version
- Vendor accuracy figures of 99% and up count only the addresses the vendor chose to judge.
- Hunter's benchmark counted all 3,000 real business addresses and put 15 verifiers between 31.2% and 70.0%, with the top ten from 57.6% to 70.0%.
- Catch-all, Yahoo and greylisted addresses stay unsettled at every vendor, so a figure that keeps them in the count lands far below 99%.
- Ask for coverage first. It tells you how much of your list a verifier actually settled.
Where these numbers come from
- Hunter's benchmark page, dated 3 November 2025 and read on 2026-09-24: 15 verifiers tested on 3,000 real business addresses, with accuracy counted across the whole set.
- Vendor accuracy claims read on each vendor's own site on 2026-09-18: ZeroBounce, Clearout and Reoon. Bouncer and Emailable were read the same day and publish no accuracy percentage.
- The protocol behaviour below (catch-all domains, Yahoo, greylisting) is how SMTP works for every verifier. You can see each step in how verification works.
Two figures that count different things
A vendor's accuracy figure answers one question. Of the addresses it gave a firm verdict on, how many were right? Unknowns and catch-alls leave the count before the sum is done, and that's where the hard cases live.
Hunter asked a wider question. It kept every address in the count, including the ones a tool couldn't settle, and scored 15 verifiers on 3,000 real business addresses. Its own tool came top at 70.0%, and it published the method beside the result.
Both kinds of figure can be true of one product at the same time. You can't put them side by side, though, and you can't compare two vendors' headline figures either until you know what each one left out.
| Figure | Published by | What it counts | Date |
|---|---|---|---|
| 99.6% | ZeroBounce, about its own tool | Addresses it gave a firm verdict on | Read 2026-09-18 |
| 99.73% | Clearout, about its own tool | Addresses it gave a firm verdict on | Read 2026-09-18 |
| Over 99% | Reoon, about its own tool | Addresses it gave a firm verdict on | Read 2026-09-18 |
| 31.2% to 70.0% | Hunter, across 15 verifiers | Every one of 3,000 real business addresses | Dated 3 Nov 2025, read 2026-09-24 |
| 57.6% to 70.0% | Hunter, the top ten of those 15 | Every one of 3,000 real business addresses | Dated 3 Nov 2025, read 2026-09-24 |
Vendor figures read from each vendor's own site on the date shown. Bouncer and Emailable publish no accuracy percentage. Figures move, so confirm them on each site before you rely on one.
How a 99% claim is built, in round numbers
Take a list of 1,000 addresses. Say a verifier gives a firm verdict on 700 of them and marks the other 300 unknown or catch-all. If 693 of those 700 verdicts are right, it can print 99% accuracy.
Now count all 1,000 addresses. The same run scores 69.3%. Nothing in either sum is false, because only the denominator moved.
These are example numbers to show the arithmetic. The lesson is the question you ask a vendor: how much of my list reached a verdict? This market calls that coverage, and it's the number a single accuracy figure hides.
Who ran the test changes how you read it
A benchmark run by a vendor puts that vendor's own tool in the field it scores. Hunter's did, and Hunter came first. That doesn't make the test useless. Hunter named its sample, its date and its method, and that's what lets you weigh it.
When you read any benchmark, look for the sample size, where the addresses came from, the date, and whether unknowns stayed in the count. Hunter's page gives you all four. A headline percentage on a pricing page gives you none of them.
You'll find the same split when two tools grade one list differently. Our answer on why verifiers disagree walks through where the gaps come from.
Why every real list has addresses nobody can settle
Some mail servers answer the same way to every address. A catch-all domain returns 250 for real and made-up mailboxes alike, so an SMTP check can't tell them apart.
Yahoo and AOL accept every RCPT TO and bounce later, which is why Yahoo addresses so often come back unknown. Gmail and Microsoft 365 block probes from cloud IPs. A server using greylisting returns a 4xx on the first try by design.
Those addresses end up unknown or catch-all at every vendor. A figure that drops them can look close to perfect. A figure that keeps them can't, which is why Hunter's range sits so far below the claims.
How ZapBounce reports its own results
ZapBounce reports two numbers together. Coverage is the share of your list that got a firm verdict. Accuracy is how right those verdicts are. You always see both, so you can read our results the way Hunter's benchmark reads everyone's.
Every address comes back as valid, invalid, catch_all or unknown. ZapBounce runs its own SMTP check on port 25 and closes the connection before DATA, so no email is ever sent. A catch-all domain is found with one probe to a random address at the same domain, then labelled as its own result and never folded into valid.
You pay for firm results only. Unknowns and duplicates are never billed, credits never expire, and pricing starts at $5 for 1,000 addresses with no minimum.
Run the benchmark that matters: your own list
The fairest test is the one you run yourself. Pick a few hundred addresses you already know about, some that bounced and some that reply. Run the same file through two verifiers.
Compare how many each one left unknown before you compare anything else. Then check the verdicts against what you know. It takes an afternoon and tells you more than any vendor's headline.
ZapBounce gives you 100 free checks a month with no card, so your side of the test costs nothing. If you want a shortlist first, start with our ranked email verification services.
Sources
Questions people ask
Is a 99% email verification accuracy claim false?
Usually not, but it answers a narrow question. It is the share of right answers among the addresses the verifier chose to judge, after unknowns and catch-alls leave the count. Counted across every address, Hunter's November 2025 test put 15 verifiers between 31.2% and 70.0%.
Is Hunter's email verifier benchmark biased?
Hunter ran the test and its own tool came first at 70.0%, so read it as a vendor's test. It still published the sample of 3,000 real business addresses, the date and the method. That makes it far easier to check than a headline figure with no method behind it.
What should I compare instead of an accuracy percentage?
Compare coverage first: how much of your list got a firm verdict of valid, invalid or catch-all. Then compare accuracy on those verdicts and whether unknowns are billed. ZapBounce reports coverage beside accuracy and never bills an unknown.
Why do catch-all addresses drag benchmark scores down?
A catch-all server answers 250 to every address, real or made up, so no SMTP check can settle one at mailbox level. A test that keeps them in the count scores every verifier lower. ZapBounce labels them catch-all after one probe to a random address at the same domain.
Check your own list's coverage
100 free checks a month, no card. Every result comes back as valid, invalid, catch-all or unknown, and unknowns are never billed.