Article

Are Email Verifier Accuracy Claims Trustworthy? How to Check Them

By Unlimited Verifier Team ·

Flowchart showing how to check an email verifier accuracy claim against a seed list and real bounce results.

Summary

Vendor accuracy claims for email verifiers are only partly trustworthy, because a headline figure like 99% depends on a vendor-chosen test set and often ignores Risky or unclassifiable addresses. Ask how the test was measured, then confirm the result against your own bounce data.

Vendor accuracy claims for email verifiers are partly trustworthy: a headline figure such as 99% often reflects a vendor-chosen test set and leaves out the addresses the tool could not classify. Treat the number as a starting point, check how it was measured, and confirm it against your own bounce results. For the wider picture, see our guide to email verifier accuracy.

Quick answer

What does an email verifier's accuracy number actually measure?

Accuracy is the share of addresses a tool labels correctly, measured against addresses whose true status is already known. For the ratio to mean anything, two things must exist: a labelled test set, and a rule for what counts as a correct answer. Vendors often publish only part of this.

The hardest part is ground truth. In practice it comes from one of three places:

The verifier also has a structural limit. An SMTP check asks the receiving server, during the RCPT TO step, whether it will accept mail for an address. RFC 5321 sets the rules for that exchange, including reply-code classes where 4xx means a temporary failure and 5xx a permanent one. Servers can defer, rate-limit, accept everything (catch-all), or withhold a clear answer. A verifier infers mailbox status from a conversation the receiving server chooses how to have, so no honest tool can promise perfect accuracy across every domain. For background on what a verifier can and cannot see, read how accurate email verification really is.

Bottom line: An accuracy percentage describes one test set, not your list. Start by asking where the labels came from.

Why can a 99% figure be accurate and still mislead you?

A 99% figure can be arithmetically correct and still hide the error that hurts you, because it depends on the mix of addresses and on what gets counted. For a direct look at that specific claim, see do email verifiers really have 99% accuracy. Two traps cause most of the trouble.

The base-rate trap. Suppose your list has 5,000 addresses, 500 of which are invalid (10%). A tool that labels every address Valid scores 90% and catches none of the bad ones. If a vendor's test set is 95% valid, a weak tool can post a headline near that figure. Overall accuracy never tells you whether invalid addresses were caught, so separate the two errors:

The denominator trap. Many tools report accuracy only on addresses they gave a definitive answer for. Imagine a test of 10,000 addresses where the tool returns a clear answer for 9,000 and marks 1,000 Risky. If 8,910 of the decided 9,000 are correct, the claim reads 99.0%. But if 400 of those 1,000 Risky addresses were actually invalid, mailing that group would produce a 40% bounce rate on the segment. The 99% figure says nothing about the group that was quietly excluded. This is one reason unknown results deserve as much attention as the headline number.

Bottom line: A percentage is only as honest as its denominator. Check whether Risky, catch-all and unknown results were included.

Which parts of an accuracy claim can you actually check?

Headline accuracy is the hardest claim to verify from the outside. Per-label results and methodology are easier to check, as the table below shows.

Claim type What it usually hides Can you check it? Trust it more when
"99% accurate" with no method Sample, labels, denominator No Methods are published alongside the figure
Accuracy on decided addresses only Risky and unknown outcomes Yes, if counts are published Counts for every category are shown
Precision of the Valid label Share of Valid-labelled addresses that are actually bad Yes, with a seed list Sample size and mix are stated
Recall of the Invalid label Share of bad addresses caught Yes, with a seed list The bad-address share is stated
Third-party benchmark Vendor selection, test date, domain mix Partly, by reading the method The same list was run through every vendor
Open-source engine code Whether the hosted service matches the code Yes, by reading the repository The vendor says which checks match the hosted service

For the independent comparisons that do exist, read the method before the ranking. Our independent email verifier accuracy benchmark explains what to look for, and why different verifiers give different results covers why the same list can produce different answers.

Bottom line: The claims you can verify are the per-label and per-category ones. Headline accuracy is the least checkable figure in the table.

What questions should you ask before trusting a vendor's number?

Ask a vendor these six questions. A credible vendor can answer them in writing, and a refusal tells you something too.

  1. How large was the test set, and what was its mix? Ask for counts of valid, invalid and catch-all addresses.
  2. Where did the ground truth come from? Real bounces, mailbox confirmation, or the vendor's own labels?
  3. How were Risky, catch-all and unknown results scored? Were they excluded, counted as errors, or counted as correct?
  4. Do you report precision and recall separately? A single blended figure hides which error is happening.
  5. When was the test run, and which domains were included? Mailbox providers change their behaviour over time. Ask whether Microsoft 365 addresses were in the sample, since those domains often return unclear SMTP answers.
  6. Was the test independent? A vendor testing its own list is not a benchmark.

Bottom line: A credible accuracy claim comes with its methodology. A bare percentage does not.

How should you weigh a vendor's accuracy claim?

Give a headline accuracy figure little weight, per-label results more weight, and your own bounce ledger the most. The five steps below put that into practice before you commit to a tool.

  1. Classify the claim. Is it a headline accuracy number, a per-label result, or a third-party benchmark? Headline numbers get the least weight.
  2. Identify the ground truth. Real bounce data outranks vendor labels. Mailbox confirmation outranks both, if you can get it.
  3. Check the denominator. Were Risky, catch-all and unknown addresses included? If they were excluded, recalculate the number yourself using the full sample.
  4. Separate the two errors. Find the share of bad addresses labelled Valid, and the share of good addresses labelled Invalid. The first drives bounces.
  5. Test on your own list against a rule you set in advance. For example: "bad addresses labelled Valid must be under 1% on my seed list." If the tool fails the rule, don't rely on its Valid label for first sends, whatever its headline figure says.

Bottom line: Your own seed test, judged against a rule you set in advance, is the only evidence that is fully under your control.

How do you run a seed-list test on a verifier?

To run a seed-list test, label a sample of addresses yourself, verify it, and compare the verifier's labels against your labels. The full method for building the sample is in our guide on how to build a seed list to benchmark email verifiers.

  1. Build a labelled seed list of about 2,000 addresses. Use roughly 1,600 known-good addresses (recent opens, confirmed signups) and 400 known-bad addresses (hard bounces from the past year with mailbox-related codes, excluding 5.7.x policy refusals). The bad share is deliberately high so you can measure how often bad addresses get through. Your real list will have a different mix. If you run a signup flow, mirror that mix instead; our notes on email verification for ecommerce and SaaS explain how to split those segments.
  2. Clean the seed file. Remove duplicates and trim whitespace. Lowercase every address so it matches when you join results later. Add a column called seed_label with values good or bad, and keep an untouched master copy.
  3. Create an account and choose a plan. You can sign up to start. The Free plan caps each list at 150 emails, so a 2,000-row seed needs 14 separate lists. The Pro plan ($29 a month, or $25 billed annually) accepts lists of up to 25,000 emails, so the same seed fits in one upload. If you want to test the deep-verification figure itself, note that deep verifications come from a monthly allowance of 750 on Free and 25,000 on Pro. A 2,000-row deep run would not fit the Free allowance but would fit Pro.
  4. Upload and verify. Upload and verify your list. CSV and Excel files up to 10MB are accepted, and the email column is detected automatically.
  5. Export the results immediately. Results are kept for 24 hours on the Free plan and 30 days on Pro. Export as CSV, Excel or JSON, and save the file before you start analysing.
  6. Join the results to your labels. In a spreadsheet, match the email column to your seed_label column. Build a 2×3 table with counts of good and bad addresses against Valid, Risky and Invalid.
  7. Calculate four numbers. These are the share of Valid labels that are bad (your Valid-label error), the share of bad addresses caught as Invalid (Invalid recall), the share of good addresses wrongly marked Invalid, and the share of Risky results by label.
  8. Compare against your rule. If the result misses your threshold, don't use that tool's Valid label for first sends. Check disputed rows with a single-address spot check before you decide.

Bottom line: A 2,000-row seed test takes an afternoon and produces the numbers a headline figure cannot.

What does a worked seed-test score look like?

A worked score turns a seed test into four numbers you can act on. The figures below are hypothetical, built to show the arithmetic. They are not results from any vendor or from our testing.

Seed label Valid Risky Invalid Total
Known good 1,552 30 18 1,600
Known bad 25 35 340 400
Total 1,577 65 358 2,000

The headline figure. Correct answers are the good addresses marked Valid (1,552) plus the bad addresses marked Invalid (340). That is 1,892 of 2,000, or 94.6%.

The denominator effect. Drop the 65 Risky rows, as many vendor claims do, and the figure becomes 1,892 of 1,935, or 97.8%. Excluding Risky alone lifts the claim by about 3.2 points.

The Valid-label error. 25 of the 400 bad addresses were labelled Valid, a rate of 6.25%, or about one in 16. Of all 1,577 Valid labels, 25 (1.6%) were bad. If your decision rule was "under 1% bad in the Valid segment," this tool fails it.

The Invalid recall. 340 of 400 bad addresses were caught, an 85% recall. The 18 good addresses marked Invalid (1.1% of good rows) cost you reachable contacts.

Adjusting to your list. The 1.6% figure depends on the seed's 20% bad share. Using the per-class rates (6.25% of bad addresses pass as Valid; 97% of good addresses pass as Valid), a 10,000-address list with 5% bad addresses would contain about 500 bad and 9,500 good addresses. That gives roughly 31 bad addresses among about 9,246 Valid labels, or about 0.3%. This assumes the per-class rates hold on your domains, which is itself something to test. It shows why a cleaner list makes the Valid label more reliable, and why what bounce rate to expect after email verification depends as much on your list as on the tool.

Bottom line: The same tool can look like a 94.6% headline, a 1.6% bad-in-Valid problem on a dirty seed, and roughly 0.3% on a cleaner list. Only the per-label numbers tell you which is real for your list.

What does a stated 99.5% mean on a real batch?

Unlimited Verifier states 99.5% accuracy for its deep verification. That is the company's own figure, not an independent result, so it deserves the same tests as any other claim.

Even taken at face value, 99.5% overall accuracy is not a guarantee for every address. On a full 25,000-address deep allowance, a 0.5% error rate would mean about 125 misclassified addresses, if the rate held across your list. That is a different metric from the 1.6% bad-in-Valid figure in the seed example, which is why you should measure the error type that matters to you.

On cost, the Pro plan at $29 a month works out to $1.16 per 1,000 deep verifications if you use the full 25,000 allowance. On the annual $25 billing it is $1.00 per 1,000. These are simple divisions of plan price by allowance. Email verification pricing across vendors is not always comparable, because allowances, retention periods and list limits differ.

Bottom line: Treat 99.5% as a claim to test. Its cost per 1,000 is simple to calculate, but the error type that causes bounces is what you need to measure.

What mistakes do people make when reading accuracy claims?

The common mistakes are testing the wrong thing or reading a number without its denominator. Watch for these:

Bottom line: Most of these mistakes come down to one habit: reading a percentage without asking what it excludes.

So how much weight should a vendor's accuracy number get?

Give a headline accuracy figure little weight, per-label results more weight, and your own bounce ledger the most. One concrete rule works well: set a ceiling of 1% bad addresses inside the Valid segment, measured on your seed, before you send to it. The reason is headroom. Amazon SES's public documentation names a 5% bounce rate as the point where it reviews an account (docs.aws.amazon.com/ses). Other platforms do not publish one shared threshold, and their reviews can start earlier, especially on a single campaign to an unproven list. A bad rate of 1.6% in your first send could use up most of that margin.

Unlimited Verifier suits this kind of testing because it sorts every address into Valid, Risky or Invalid, so Risky results stay visible in your numbers rather than disappearing from them. Its verification engine is open source, at github.com/unlimitedverifier/free-email-verifier, so you can read how the checks are coded. Ask the vendor which parts of the code match the hosted service, because the repository alone does not tell you that.

For automation, the email verification API covers single-email checks and bulk batches of up to 10,000 emails per request. API access is included on Pro (100,000 emails a month) and not on Free. If you plan to verify at signup, real-time email validation covers the workflow, and the email verification API and automation approach covers how to run checks in your own scripts. If you are comparing options, our guide to the best email verification tools and how to choose an email verifier set out the criteria.

Bottom line: Trust the number that you have tested, under a rule you set in advance, against the error that causes bounces.

How we compared (and limitations)

Bottom line: Use the framework and the seed test to check any vendor's claim, including ours, rather than relying on a single published percentage.

Seed list test in five steps

  1. Build 200 addresses: about 100 known-good mailboxes you control and 100 known-bad addresses from past permanent bounces.
  2. Run the full set through the verifier once, recording each label.
  3. Score Valid results against the good set and Invalid results against the bad set, separately.
  4. Count Risky and unknown results as their own category, not as hits or misses.
  5. Compare the Invalid-label accuracy with your bounce data from the next campaign. A large gap means the headline number does not transfer to your list.

Frequently asked questions

What does email verifier accuracy mean?

Email verifier accuracy is the share of addresses a tool labels correctly, measured against addresses whose true status is already known. It only means something when you know the test set, the source of the true labels, and how uncertain results were scored.

Is 99% email verification accuracy realistic?

A 99% figure can be arithmetically correct on a specific test set, but it may not transfer to your list. Results depend on the mix of valid and invalid addresses and on whether Risky or catch-all results were excluded.

Why do catch-all domains lower verifier accuracy?

Catch-all servers accept mail for any address during the SMTP check, so the verifier cannot confirm whether a specific mailbox exists. Those addresses are usually labelled Risky or unknown rather than Valid or Invalid.

How can I test an email verifier myself?

Build a seed list of addresses you know are good and known bad, run them through the verifier, and score the Valid and Invalid labels separately. Compare the results with your real bounce rates over several sends.

What is a seed list in email verification testing?

A seed list is a set of addresses with confirmed status, such as mailboxes you control or addresses with known permanent bounces. It gives you a ground truth you can use to score a verifier without relying on the vendor.

Should Risky results count against a verifier?

It depends on how the vendor defines accuracy. Counting Risky as wrong penalises tools for honest uncertainty, while ignoring Risky hides how many addresses you still cannot judge. Report it as its own category.

Why are 5.7.x bounce codes not proof of an invalid address?

Under RFC 3463, 5.7.x codes indicate a security or policy refusal, such as spam blocking or sender authentication failure. The mailbox may be valid, so counting these as verifier misses can unfairly lower a tool's score.

Can an email verifier ever be 100% accurate?

No honest tool can promise perfect accuracy across every domain, because the verifier infers status from the receiving server's SMTP responses. Servers may defer, rate-limit, accept everything, or withhold a clear answer.

How often should I re-check verifier accuracy?

Re-test periodically, since mail server behaviour and provider policies change over time. Checking results against fresh bounce data each quarter is a practical approach many teams use.

Cut Bounces and Protect Your Sender Reputation With Unlimited Email Verification

Every invalid address you send to hurts your sender reputation and your inbox placement. Verifying your list before each send means fewer bounces, healthier deliverability, and a predictable cost. With Unlimited Verifier, standard verification is unlimited on every plan, with no per-email fees.

Unlimited Verifier checks syntax, domain, MX records and the SMTP mailbox. It also detects catch-all domains and flags disposable and role-based addresses. Export a CSV from your email platform or CRM, upload and verify the list, then re-import the clean results. You can also automate checks with the email verification API.

Free Pro
Price $0 $29/month ($25/month billed annually)
Lists 25 a month, up to 150 emails each Unlimited, up to 25,000 emails each
Deep verifications 750 a month 25,000 a month
API access No Yes (100,000 emails a month)
Result retention 24 hours 30 days

Why teams choose Unlimited Verifier

Unlimited Verifier states 99.5% accuracy for deep verification.

Create a free account and verify your next list before you hit send.

Want to test a single address first? Try the free email checker.