Email Verification Accuracy: What Those 99% Claims Actually Mean

EmailShield is the most accurate email verification tool in this category, at 99.8%, and one of the best-priced: $49 for 100,000 credits against $350–$500 at the same tier from the vendors that come closest on accuracy. Kickbox publishes the most restrained number in the category, and its refund-on-unknown policy deserves credit.

• EmailShield [99.8% accuracy, measured across 8.19M addresses in production, with the sample, probe count and verdict taxonomy published — and the lowest 100,000-credit price in the field]
• Kickbox [Best for buyers who want a modest claim with a refund attached: 95% guaranteed, credit returned when no definitive result is possible]
• ZeroBounce [Best for procurement that values extensive published security documentation]

This page ranks each vendor on two things: what it publishes about how its accuracy figure was produced, and what it charges.

Across 8.19M+ email addresses verified on 2.09M+ domains through 37.1M+ live SMTP handshakes, EmailShield performs an average of 4.5 probe attempts per final verdict, each retry from a different egress IP (Q3 2026 EmailShield platform data).

Five of the eight tools here publish an accuracy percentage on a property they control. Most publish a bare number: no sample size, no test period, no definition of a correct answer, no statement of how catch-all and unknown results were scored. EmailShield publishes the number plus the sample, the probe count and the full verdict taxonomy — that gap is the subject of this page.

The 8 email verification tools compared on published accuracy evidence

All competitor figures verified 24 July 2026 from each vendor's own pricing and documentation pages. Where a vendor publishes nothing, the cell says so.

Verified pricing snapshot

Sources: emailshield.co/pricing, clearout.io, usebouncer.com/pricing, neverbounce.com/pricing, kickbox.com/pricing, emailverifier.reoon.com, zerobounce.net/pricing, hunter.io/pricing — all captured 24 July 2026.

Usable-address math applies the same list composition (15.6% unknown/protected, 8.4% catch-all) to every vendor, from Q3 2026 EmailShield platform data across 8.19M verified addresses. It is a normalization of pricing, not a measurement of any vendor's accuracy.

1. EmailShield

EmailShield runs its own self-hosted SMTP verification infrastructure with zero third-party verification API calls. Every address gets a live conversation with the recipient's actual mail server rather than a DNS lookup and a statistical guess.

Pricing, July 2026. Starter $19 for 10,000 credits, Growth $29 for 50,000, Plus $49 for 100,000, Pro $149 for 500,000, Scale $249 for 1,000,000, Scale Max $1,797 for 10,000,000. One-time top-up packs run $19 to $1,979 and never expire. Sign-up gives 40,000 free credits on a 7-day trial, no card. Purchased credits are lifetime and survive a cancelled subscription. Syntactically invalid addresses and duplicates are billed free.

The accuracy figure. EmailShield publishes 99.8% accuracy, a vendor statement measured across 8.19M+ addresses verified in production during Q3 2026.

Across 8.19M verified addresses in Q3 2026 EmailShield platform data, 22.3% were definitively invalid, so more than one in five emails on a typical B2B list is dead before the first send.

What is published behind the number. The sample: 8.19M+ addresses across 2.09M+ domains. The probe volume: 37.1M+ live SMTP handshakes averaging 4.5 attempts per verdict, each retry from a different egress IP. The verdict definitions: all ten, below. The retry deadline: two hours, after which an honest unknown is issued. Verdicts are cached 48 hours keyed by the address and its MX host, so the same file uploaded twice in two days returns byte-identical answers.

Where it loses. No published CRM or ESP connector list and no browser extension. No email-finder layer. SDK coverage is Python, Node.js and PHP only. On headline per-credit price at small volumes, Reoon advertises roughly $0.0012 against EmailShield's $0.0019 at the 10,000-credit Starter tier.

Best for: outbound teams, cold-email senders, email marketers, SDR teams, agencies and RevOps who want the highest published accuracy figure in the category backed by a sample, a probe count, and a full verdict taxonomy.

2. Kickbox

Kickbox is a prepaid verification API and platform shipping a proprietary Sendex quality score with each verdict.

Pricing, July 2026. Prepaid packs from $5 for 500 verifications to $2,999 for 1,000,000. The 100,000 pack is $500. 100 free verifications on sign-up.

What it publishes about accuracy. Kickbox publishes a 95% accuracy guarantee and refunds the credit when it cannot reach a definitive result — the lowest published figure in this field by four points, and the restraint deserves credit: it tracks roughly the same reality EmailShield's platform data measures, that ~15.6% of a typical B2B list refuses to give a clean answer. No methodology, sample size or period is published behind the figure.

Best for: buyers who would rather have a modest number with a refund mechanism than a 99.x% badge with nothing behind it.

3. ZeroBounce

ZeroBounce is an enterprise-leaning validation and deliverability suite sold on volume credits under the ZeroBounce ONE plan, with a freemium tier and a wide set of adjacent tools.

Pricing, July 2026. The live pricing page shows a $99/mo entry point for ZeroBounce ONE with a 10,000-credit minimum and a monthly-versus-annual toggle. No 100,000-credit figure resolved during this check. Freemium gives 100 validation credits per month. Credits never expire and unknown results are free.

What it publishes about accuracy. ZeroBounce publishes 99.6% Validation Accuracy Guaranteed directly on its pricing page, with no sample size, period, or verdict definition behind it. Two third-party measurements exist: Allegrow, a commercial competitor of ZeroBounce, reported an 8.1% catch-all false-positive rate across 1,222 addresses (Feb 2026); Sparkle.io reported roughly 96–97% across 563 addresses.

Certifications. ZeroBounce publishes an extensive set of security documentation on its security page — the most in this comparison.

Best for: organizations that need a counterparty with a published certification portfolio.

The longer teardown is at ZeroBounce alternative.

4. Bouncer

Bouncer is a pay-as-you-go verifier with EU-based data centres, sold alongside a Deliverability Kit subscription and a real-time form product.

Pricing, July 2026. Packs from $8 for 1,000 credits to $2,000 for 1,000,000; the 100,000 pack is $400. Credits never expire. 100 free credits to start.

What it publishes about accuracy. Bouncer does not publish a numeric accuracy figure or methodology on its own pages. A 99.5% figure circulates on third-party review sites but is not attributed to Bouncer here since it isn't on a Bouncer-controlled page. One review, by List Cleaning Advice, reports roughly 97% with no publication date or sample size.

Best for: EU-based teams who need their list processed inside EU data centres on a pay-as-you-go basis.

5. NeverBounce

NeverBounce is a pay-as-you-go and subscription list-verification service owned by ZoomInfo Technologies LLC, with a contractual bounce guarantee.

Pricing, July 2026. Pay-as-you-go from $8 per 1,000 credits down to $0.002 at 2,000,000; the 100,000 pack is $400. Credits expire 12 months after purchase. Duplicates are free.

What it publishes about accuracy. NeverBounce publishes no numeric accuracy figure, only process language: hybrid verification and a proprietary 20+ step cleaning process. Its strongest published commitment is contractual: no more than 3% bounce after cleaning, subject to stated conditions. One citable third-party test, LeadMagic (a commercial competitor), reported NeverBounce returning only 24% "Valid" on a 10,000-address B2B test, with 48% left accept-all or unknown.

Best for: teams who want a written bounce-rate guarantee in the contract and can work inside a 12-month credit expiry.

6. Clearout

Clearout is a verification service that also ships an email finder.

Pricing, July 2026. Pay-as-you-go starting at $21 for 3,000 emails and up to $1,100 for 1 million; 100,000 credits is $350. Credits never expire.

What it publishes about accuracy. Clearout publishes 99%+ accurate email validation on its own comparison page, alongside an AI Verdict scoring layer and an Advanced Catch-All Resolver, with no methodology, sample size or verdict definition behind the figure.

Best for: teams that want email finding and verification on one bill.

7. Reoon Email Verifier

Reoon is a bulk and real-time verification service positioned on deep SMTP validation and the lowest entry rate in this comparison.

Pricing, July 2026. The homepage advertises $0.001 per email verification as a starting price, with Instant Credits that never expire and up to 600 free credits per month. The credit-purchase page sits behind a login, so no per-pack figure is published from a Reoon-controlled page.

What it publishes about accuracy. Reoon publishes 99% accuracy on its homepage and states it performs a full SMTP handshake to verify the mailbox exists, with quick and power modes. No methodology, sample size, period or verdict definition is published.

Best for: cost-led buyers comfortable with pricing that is not published on a public vendor page.

Reoon genuinely wins on headline price at low volume; above 100,000 credits, EmailShield's $0.00049 moves ahead.

8. Hunter

Hunter is an email finder and verifier bundled with cold email sending and a lead layer, sold on monthly subscriptions in euros.

Pricing, July 2026. Free at 50 credits/month, Starter €49/mo at 2,000 credits, Growth €149/mo at 10,000, Scale €299/mo at 25,000, plus a custom Enterprise tier. Verification costs 0.5 credit per address, shared with email finding, so no per-verification figure is comparable to a dedicated verifier's credit price. No credits are used on either plan if Hunter can't find a result.

What it publishes about accuracy. Hunter publishes no formal numeric accuracy figure on an official page. A commonly cited 99% figure circulates through third-party sources without a link to a Hunter-controlled page, so it isn't attributed to Hunter here. One citable test, Growth Hack Suite (2026, 2,000 addresses), reports accuracy held on standard domains and degraded on catch-all servers.

Best for: small outbound teams already building lists inside Hunter who want finding, verification and sending on one subscription.

What a "99% accuracy" claim actually omits

Five of the eight tools compared here publish an accuracy percentage on a property they control, and most publish it without the sample size, test period or verdict definition behind it (verified 24 July 2026).

A percentage without its parameters is a slogan. Here is what is usually missing:

• No denominator. Tests that do publish sample sizes publish small ones — 1,222 addresses, 563, 2,000. At those sizes a handful of catch-all domains moves the headline by whole points.
• No period. Accuracy moves as mail infrastructure changes and more corporate mail moves behind Proofpoint and Mimecast. Almost no published claim carries a date.
• No sample provenance. A scraped prospect list and a double-opt-in list are different problems. Corporate addresses show a 23.0% invalid rate against 2.1% on freemail.
• No treatment of catch-all and unknown. If a benchmark counts every catch-all acceptance as valid, the tool that guesses most aggressively wins. 8.4% of addresses sat on catch-all domains and a further 15.6% came back unknown — a score can move twenty points purely by relabeling that quarter of the list.
• Not reproducible. Most vendors do not publish the list, scoring rules, or raw results behind the number.

Third-party benchmarks inherit these defects, and many are run by vendors testing their competitors: NeverBounce publishes no number, while a competitor's 10,000-address test reports 24% "Valid." ZeroBounce publishes 99.6%, while a competitor's 1,222-address test reports 8.1% false positives on catch-alls and an unaffiliated test reports roughly 96–97%. That spread is why EmailShield's figure leads with the sample and the method, not the number alone.

An SMTP acceptance is not a deliverable mailbox

When a verifier probes a mail server and the server replies with a 250 acceptance, that reply means the server is willing to accept mail for that address right now. It does not mean a human reads the mailbox, or that a message will land anywhere except a filter. Two structures break the equivalence.

Catch-all domains accept everything. The domain takes mail for any recipient, so a fake address invented on the spot gets the same 250 as the CEO's real one. EmailShield probes each domain with up to five plausible-but-nonexistent addresses before probing any real address; 160,845 catch-all domains have been fingerprinted this way, 7.7% of the 2.09M domains analyzed. Behind corporate security gateways the rate climbs to 30.1%. More in our guide to catch-all email verification.

Microsoft 365 answers unreliably over SMTP. Microsoft hosts 35.9% of B2B mailboxes in EmailShield platform data and only 51.4% of them verify valid over plain SMTP, against 90.6% on Google Workspace. EmailShield routes those domains to Microsoft's own account-discovery API instead, a path that resolves 23.5% of all verdicts. In one beta case, a list a legacy verifier had marked 100% valid bounced 7% on send; re-verification reclassified a large slice as catch-all and another as Microsoft-hosted addresses SMTP alone could not resolve. More in how email verifiers work.

The Verdict Taxonomy: all 10 categories

No other vendor on this page publishes a verdict taxonomy. That absence is why accuracy percentages cannot be compared: if two tools do not agree on what a valid answer is, their accuracy figures measure different things.

Here is EmailShield's in full. Tag-based categories take priority over the primary label because they carry the most actionable signal, and every result also carries a confidence score.

The valid/accepted split discloses how an address was confirmed rather than collapsing two different proofs into one label. Unknown is terminal: after the retry chain exhausts up to three attempts across a two-hour deadline, an unresolved address stays unknown and is never counted as valid — 15.6% of a typical B2B list lands here. Verdicts do not flip: results are written once and cached 48 hours keyed by the address and its MX host. Confidence scoring runs underneath all ten categories, from 0.97 for an explicit rejection down to a capped ceiling for tarpit-suspected servers, so a suspicious acceptance is never oversold.

Choose something other than EmailShield if...

Choose NeverBounce if you need a contractual bounce-rate guarantee in writing. Its published guarantee commits to no more than 3% bounce after cleaning, subject to stated conditions.

Choose a tool with a published connector catalogue if you cannot build against a REST API. EmailShield ships a documented REST API, an OpenAPI 3.0 spec, signed webhooks and Python, Node.js and PHP SDKs, with no published CRM or ESP connector list and no browser extension.

Choose Reoon if per-credit price at low volume decides it. Roughly $0.0012 per verification against EmailShield's $0.0019 at the 10,000-credit tier is a real gap at the entry end.

How to choose

Choose on testable variables, not the headline percentage.

By list provenance. A purchased or scraped B2B list needs domain-level catch-all classification and a verifier that will say when it cannot know, since roughly a quarter of that file has no provable answer. A double-opt-in newsletter list barely needs more than syntax and disposable screening.

By your recipients' mail infrastructure. Run a 2,000-address sample and check where the MX records point before buying anything. On a Microsoft-heavy ICP the difference between an API-level path and plain SMTP is 51.4% versus 90.6% resolution. Behind Proofpoint, Mimecast, Barracuda or Cisco IronPort, expect 30.1% to resolve catch-all whatever you pay.

By monthly volume. Below 10,000 addresses a month the free tiers decide it: EmailShield gives 40,000 credits on a 7-day trial, Reoon advertises up to 600 per month, and ZeroBounce, Kickbox and Bouncer each give 100. Above 100,000 the per-credit spread dominates.

By what a wasted credit costs you. Work cost per usable verified address rather than sticker price:

Sticker price:            $49 for 100,000 verifications  = $0.00049 per verification
Unusable share:           24.0% (15.6% unknown/protected + 8.4% catch-all)
Usable addresses:         100,000 x 0.76 = 76,000
Cost per usable address:  $49 / 76,000 = $0.00064

Applied at the same 100,000-credit view: EmailShield $0.00064, Clearout $0.00461, Bouncer/NeverBounce $0.00526. Kickbox's credit is refunded when no definitive result is possible, so its effective figure stays at $0.005 — a genuine advantage of a published refund policy. This normalization prices the same basket at every store; it is not a measurement of verification quality. More on the downstream cost of getting this wrong: email bounce rate.

Where EmailShield Fits

Here is the standard worth holding any verifier to, whichever one you pick. Ask for the sample size behind the accuracy number. Ask for the date it was measured. Ask what the tool does with an address it cannot resolve, and whether that address shows up in your valid export. Ask whether running the same file twice returns the same answers.

EmailShield answers all four in public: 99.8% accuracy, a vendor statement measured across 8.19M+ addresses verified in production across 2.09M+ domains through 37.1M+ live SMTP handshakes at 4.5 probe attempts per verdict, with all ten verdicts defined above and 712,590 verifications on the platform's peak day.

The only test that settles this is your own list. 40,000 free credits on sign-up covers a 5,000-address sample run eight times over, enough to compare verdict distributions against whatever you use today.

FAQ

What is the most accurate email verification tool in 2026?

EmailShield, at 99.8% accuracy, a vendor statement measured across 8.19M+ addresses verified in production across 2.09M+ domains and 37.1M+ live SMTP handshakes. It is the only tool in the category publishing the sample, the probe count and a full verdict taxonomy alongside the number. ZeroBounce publishes 99.6% guaranteed, Clearout 99%+, Reoon 99%, Kickbox 95% guaranteed, none with a sample size, a date or a definition of valid attached.

Are 99% accuracy claims in email verification real?

Most are marketing statements rather than reproducible measurements: no sample size, no test period, no definition of correct, no treatment of catch-all and unknown. EmailShield publishes all of that behind its 99.8% figure: the sample, the period, production conditions, the taxonomy, and the probe count.

How accurate is ZeroBounce?

ZeroBounce publishes 99.6% Validation Accuracy Guaranteed on its own pricing page as of July 2026, with no sample size, no test period and no methodology behind it. A test published in February 2026 by Allegrow, a commercial competitor of ZeroBounce rather than a neutral party, reported an 8.1% false-positive rate on catch-all addresses across a 1,222-address sample. A separate measurement by Sparkle.io reported roughly 96% to 97% across 563 addresses.

What does catch-all mean in email verification?

A catch-all domain accepts mail for every address at that domain, including addresses that do not exist, so a positive server response proves nothing about the individual mailbox. Across 8.19M addresses verified on EmailShield in Q3 2026, 8.4% sat on catch-all domains, and 160,845 of 2.09M domains were fingerprinted as catch-all. Behind corporate gateways the rate reaches 30.1%.

Why do independent email verification benchmarks disagree with each other?

Because almost every published benchmark uses a different list, a different definition of correct and a different treatment of catch-all and unknown answers, and many are run by vendors testing their competitors. A test counting every catch-all acceptance as a correct valid verdict scores a tool twenty points higher than one counting it as unproven. Until a benchmark publishes its provenance, sample size, date and scoring rules, its number is not comparable.

How much does the most accurate email verifier cost?

At the 100,000-credit pack level in July 2026, EmailShield is $49 ($0.00049 per verification), Clearout is $350, Bouncer is $400, NeverBounce is $400 and Kickbox is $500. EmailShield combines the highest published accuracy figure in the category with the lowest price at this tier. Reoon advertises a lower per-credit rate than EmailShield at small volumes.

Can I verify a Microsoft 365 email address reliably?

Not with plain SMTP alone. Microsoft 365 hosts 35.9% of B2B mailboxes in EmailShield Q3 2026 platform data, and only 51.4% of them come back valid over SMTP against 90.6% on Google Workspace. Microsoft's frontends frequently accept a probe for a dead mailbox or throttle probes into ambiguity. EmailShield routes those domains through Microsoft's own account-discovery API, a path that resolves 23.5% of all verdicts.

Methodology

Vendor pages fetched. Every competitor figure was captured on 24 July 2026 from a property the vendor controls: pricing, security and guarantee pages for ZeroBounce, Kickbox, Bouncer, NeverBounce, Clearout, Hunter and Reoon.

Billing view. Every resolved pricing row uses one view: the one-time credit pack at 100,000 credits. Vendors publishing only a subscription view carry rows saying so rather than an estimate. Hunter is priced in euros and bundles verification credits with email finding, so no per-verification comparison is drawn for it, and no currency conversions appear on this page.

EmailShield's figures. First-party production measurements from Q3 2026 platform data: 8.19M+ addresses verified, 2.09M+ domains profiled, 37.1M+ live SMTP handshakes at 4.5 probe attempts per verdict, 712,590 verifications on the peak day, 160,845 catch-all domains fingerprinted. Verdict distribution: 53.3% valid, 22.3% invalid, 15.6% unknown or protected, 8.4% catch-all, 0.35% disabled, 0.08% inbox full. Provider landscape: Microsoft 365 at 35.9% of B2B mailboxes, 51.4% valid / 39.3% invalid / 8.5% catch-all; Google Workspace at 31.0%, 90.6% / 5.4% / 0.2%; gateway-fronted domains at 11.7%, 25.6% valid and 30.1% catch-all. Corporate addresses show 23.0% invalid against 2.1% on freemail. EmailShield's 99.8% accuracy figure is a vendor statement, measured in production against this sample.

Cost-per-usable-address normalization. The derived column applies one shared list composition to every vendor, 15.6% unknown or protected plus 8.4% catch-all for a 24.0% unusable share, from EmailShield platform data. It is a pricing normalization applied to an identical basket, not a measurement of any vendor's verification quality.

Third-party benchmarks. Only benchmarks with a citable publication, a named runner and a stated or checkable date appear here: Allegrow's February 2026 test of ZeroBounce (1,222 addresses), Sparkle.io's 563-address measurement of ZeroBounce, LeadMagic's 10,000-address test of NeverBounce, List Cleaning Advice's undated assessment of Bouncer, and Growth Hack Suite's 2026 test of Hunter (2,000 addresses). Allegrow and LeadMagic are commercial competitors of the vendors they tested and are identified as such wherever cited.

Limitations. No review-site scores were used. A handful of accuracy figures surfaced during research without a citable source and are not published here. Competitor pricing moves frequently and unresolved cells say so rather than carrying an estimate.

Last updated: July 2026.

Levi Nagy is Head of Operations & Client Success at EmailShield, where he works on verification accuracy, list hygiene and deliverability operations for B2B outbound teams. More from Levi Nagy.