Article
Seamless.AI Email Accuracy Review: How to Measure It Yourself
By Unlimited Verifier Team ·

Summary
You can measure Seamless.AI email accuracy by verifying a random sample of at least 385 contacts and calculating the Invalid and Risky shares. That gives roughly a ±5 point margin of error at 95% confidence, so you can judge how much cleaning an export needs before sending.Review Seamless.AI's email accuracy by verifying a random sample of at least 385 of its contacts and measuring the Invalid and Risky shares, which gives a margin of error of about ±5 points at 95% confidence. Use that measured rate, not a vendor's accuracy claim, to decide how much cleaning an export needs before you send. This guide gives you a method rather than a verdict, because accuracy depends on the segment you pull and no figure here has been independently verified. For the wider process, see our pillar on email verification for sales prospecting.
Quick answer
- Treat vendor accuracy claims as unverified. Measure the deliverable share on your own random sample before you rely on any export.
- Sample size sets your margin of error. 385 random contacts gives about ±5 points at 95% confidence. About 1,068 gives ±3 points.
- Uncleaned exports can bounce heavily. In the hypothetical worked example below, a 12% Invalid rate on 12,000 contacts would mean about 1,440 bounces if sent unchecked.
- A Valid result proves deliverability, not identity. Check on a smaller manual sample whether the person still holds that title at that company.
- Send to Valid contacts first. Hold Risky addresses, especially catch-all ones, until the Valid list has performed.
What does "email accuracy" actually measure?
"Accuracy" bundles three separate questions, and most disputes come from blurring them together.
- Deliverability: Does the address accept mail? An SMTP mailbox check can answer this, though only to the extent the receiving server reports it honestly.
- Identity match: Is this the right person, at the right company, with the right title? No email verifier can confirm this. It needs manual checking.
- Currency: Is the person still in that role? B2B contact data decays as people change jobs, so this score falls over time.
Unlimited Verifier checks syntax, domain, MX records and the SMTP mailbox, so it speaks to the first question. It also flags catch-all, disposable and role-based addresses. It does not confirm who is behind an address. When a team says a list is "inaccurate" after a verifier has passed it, the usual explanation is that the complaint is about identity or currency, and the check was about deliverability.
Bottom line: Score deliverability, identity and currency separately, because no single number covers all three.
Why can't you rely on a vendor's accuracy percentage?
A vendor accuracy percentage is only as useful as its definition, and public claims often leave the terms unstated. Ask four questions before you accept any figure:
- What counts as accurate? A deliverable address, a correct title and a confirmed current employer are three different standards.
- Which population was tested? Accuracy often varies by country, industry and company size. A figure from one segment rarely transfers to another.
- How recent is the data? Contact data goes stale, so a figure is only as current as its collection date.
- Who ran the test? A vendor testing its own data is making a claim, not running an audit.
Read the exact wording of Seamless.AI's accuracy statements on its own website, and treat them as claims until your sample confirms them. Status labels in other tools carry the same risk. Our explainer on Apollo's verified versus guessed email status shows how a label can mean something narrower than it sounds.
Bottom line: A vendor percentage is a hypothesis. Your random sample is the test.
Where do prospect emails usually fail?
Prospect emails tend to fail in a few predictable ways. Knowing the pattern tells you what a result means and what to do with it.
| Failure type | Why it happens | Usual result | What to do |
|---|---|---|---|
| Pattern-generated address (first.last@company.com) | The tool infers the address from a name and domain. The pattern can suit the company but not the person, for example because of an old alias or a different naming convention. | Invalid, often with a 5.1.1 "bad destination mailbox" code defined in RFC 3463 | Suppress it. Don't retry the same guessed address. |
| Catch-all domain | The mail server accepts mail for any address, so the verifier can't confirm the individual mailbox. | Typically Risky | Hold back. Test a small segment only after the Valid list has performed. |
| Role address (info@, sales@) | The address is real but shared, and it rarely reaches a decision-maker. | Often Risky | Replace it with a named contact or drop it. |
| Changed job | The address still works, but the person has moved on. | Valid, yet wrong | Only a manual or engagement check will catch this. |
| Domain change after an acquisition | The old domain may stop accepting mail. | Invalid | Suppress it, then look for the new domain. |
The catch-all row deserves the most attention. A server that accepts everything can make an address look deliverable while the message lands in a mailbox nobody reads. That shows up as lower engagement rather than as bounces, so teams often miss it.
Bottom line: Each failure type has a different fix, and a single "accuracy" score hides all of them.
How do you run an email accuracy review, step by step?
Follow these steps in order. Each one produces a number you can defend to a colleague or a client.
- Define the population. Name the exact export and segment, for example "US SaaS VPs of Sales at 50–500-employee companies." Sample each segment you plan to contact, because accuracy often varies between them.
- Set your margin of error and draw a random sample. Use the table below. In a spreadsheet, add a column with
=RAND(), sort on it, and keep the top rows. Sorting by name, company or export order introduces bias, because the top of an export is often the most recently added or most complete data. - Verify the sample with the same settings you'll use on the full list. Changing settings partway through makes the results incomparable.
- Tally the buckets and calculate rates. For a share p from sample size n, the 95% margin of error is approximately 1.96 × √(p(1−p)/n).
- Break down the Risky bucket. Count catch-all, role and disposable addresses separately. This shows whether the problem is the people data or the domains.
- Run a manual identity check on about 100 Valid contacts. Score each for correct company, correct title and evidence they still work there, such as a current profile or company page. Use a fixed rubric so two reviewers reach similar scores.
- Decide what to do. Apply the thresholds in the decision section below.
Sample size depends on the margin you need. These figures use the standard formula at 95% confidence with the worst case (p = 50%):
| Margin of error (95% confidence) | Minimum sample |
|---|---|
| ±10 points | 97 |
| ±5 points | 385 |
| ±3 points | 1,068 |
| ±2 points | 2,401 |
If your population is small, you need fewer contacts. A 2,000-contact export needs about 323 contacts for ±5 points once the finite population correction is applied. The formula assumes simple random sampling. If many contacts share one company, they are clustered, and the true margin is wider than the table suggests.
Bottom line: Random sampling, identical settings and a stated margin turn "the data looks bad" into a number you can defend.
Which review method should you use?
No single method answers all three accuracy questions. The table shows what each one can and can't tell you.
| Method | What it measures | What it misses | Time and cost | Use it when |
|---|---|---|---|---|
| Vendor accuracy claim | Whatever the vendor defines | Definition, population and recency are usually unstated | No effort | As a starting hypothesis only |
| Self-sample verification | Deliverability by bucket, plus catch-all and role flags | Identity and job currency | Hours, plus any verification cost | Before every purchase or large import |
| Manual identity check | Company, title and evidence of current role | Scale: it's slow, so samples stay small | Staff time, which rises with sample size | To estimate identity match on Valid contacts |
| Pilot send with bounce tracking | Real bounces and replies | The damage is done once mail goes out | Reputation risk if the list is unverified | Only after verification, and only to a small segment |
For most teams, the best combination is a self-sample verification plus a manual identity check. Run a pilot send last, and only after the list has been cleaned.
If you're shortlisting verifiers for this audit, our guide to the best email verification tools covers the broader market. For this review, look for these features:
- Three-bucket output (Valid, Risky, Invalid), so uncertain results are separated from confirmed failures.
- Catch-all and disposable detection, which explains most Risky results.
- Export formats that match your spreadsheet or CRM workflow.
- Result retention windows, so you can download results before they expire.
Our how to choose an email verifier guide goes deeper on these criteria.
Unlimited Verifier sorts addresses into Valid, Risky and Invalid, flags catch-all, disposable and role-based addresses, and exports results as CSV, Excel or JSON. Standard verification is unlimited on every plan, so re-checking a sample costs nothing per email. Its verification engine is also published as open source on GitHub, which lets technical reviewers inspect the checks themselves. Before you start, open a results file and confirm how catch-all addresses are labelled. The worked example below assumes they land in Risky.
Bottom line: Use the self-sample to measure deliverability and the manual check to measure identity. Neither replaces the other.
How do you verify a Seamless.AI export, step by step?
This walkthrough covers one audit cycle. Set aside about an hour for a 385-contact sample.
- Export the contacts you're evaluating as a CSV. Include the email column, the company domain, and the title and company-size fields you'll segment on. If the export is large, add a
=RAND()column in your spreadsheet and keep the top 385 rows. - Create an account. Sign up for free. The Free plan allows 25 lists a month of up to 150 emails each. Your 385-contact sample splits into three lists: 150, 150 and 85.
- Upload each file. On the Upload and verify a list page, CSV and Excel files up to 10MB are accepted, and the email column is detected automatically.
- Run verification. Each address is checked for syntax, domain, MX records and the SMTP mailbox, and catch-all, disposable and role-based addresses are flagged.
- Read the three buckets. Treat Valid addresses as candidates for sending, suppress Invalid addresses, and examine Risky addresses individually.
- Break down the Risky bucket. Use the catch-all, disposable and role flags to group the results. For a single address that needs a second opinion, the free email checker runs a one-address check.
- Export the results immediately. Results are kept for 24 hours on the Free plan and 30 days on Pro. Download the CSV, Excel or JSON file before the window closes.
- Calculate your rates. In your spreadsheet, use
=COUNTIFon the bucket column for each bucket, then divide by the sample size. - Feed the Valid contacts back into your workflow. There are no native ESP, CRM or Zapier integrations, so re-import the Valid CSV into your sequencer or CRM. For repeat audits, a script-driven verification workflow can call the API keys instead. Pro includes 100,000 API emails a month, and each bulk request accepts up to 10,000 emails.
Bottom line: A 385-contact sample costs nothing beyond a free account and gives you a measured Invalid rate to decide with.
What does a 12,000-contact accuracy audit look like?
The numbers below are hypothetical. The arithmetic and plan limits are real.
Setup: A sales team exports 12,000 prospects from a Seamless.AI search and draws a random sample of 400, which clears the 385 minimum for ±5 points.
Sample results:
| Bucket | Count | Share of sample | Breakdown |
|---|---|---|---|
| Valid | 268 | 67.0% | — |
| Risky | 84 | 21.0% | 66 catch-all, 14 role, 4 disposable |
| Invalid | 48 | 12.0% | — |
Check: 268 + 84 + 48 = 400.
Margins of error:
- Invalid: 12.0% ± 3.2 points, so 8.8% to 15.2%.
- Risky: 21.0% ± 4.0 points, so 17.0% to 25.0%.
- Catch-all share of Risky: 66 of 84 (78.6%). The Risky problem here is mostly a domain problem, not a people-data problem.
Projection to the full 12,000:
- Expected Invalid: 12,000 × 12% = 1,440 (range about 1,056 to 1,824).
- Expected Valid: 12,000 × 67% = 8,040.
- Expected Risky: 12,000 × 21% = 2,520.
Why an unchecked send would fail: Sending all 12,000 unchanged would produce about 1,440 bounces, a 12% bounce rate. Amazon SES's documentation describes review at a bounce rate around 5%, with pausing possible at 10%. A 12% rate would sit beyond both lines.
Identity check: The team manually scores 100 Valid contacts and finds 82 correct on company, title and current role. That's 82% ± 7.5 points, or about 74.5% to 89.5%. A 100-contact check is too small for a tight estimate, so report the range rather than a single figure.
Cleaning plan:
- Suppress the 1,440 Invalid addresses and keep a copy of the suppression list.
- Send to the 8,040 Valid contacts first, starting with the most relevant segment.
- Hold the 2,520 Risky contacts. Test a small catch-all segment only after the Valid send has shown stable results. Catch-all email verification explains how these addresses are detected.
Cost: The 12,000-contact pull fits within one Pro list, which allows up to 25,000 emails. Pro is $29 a month, or $25 a month billed annually. Deep verification of the full pull would use 12,000 of the 25,000 deep verifications in a Pro month. The Free plan wouldn't work here: 12,000 contacts at 150 per list needs 80 lists, and the Free plan allows 25 a month, which covers at most 3,750 contacts.
Bottom line: The audit turns "the data might be bad" into a bounded estimate, a cleaning plan and a dollar cost.
How do you decide whether the data is good enough?
Use the measured Invalid rate and the identity check together. The thresholds below are our recommendation for prospecting lists, not an industry standard.
| Measured Invalid rate (sample) | What it means | Recommended action |
|---|---|---|
| 5% or less | Most of the export is deliverable | Verify the full pull, send to Valid contacts and hold Risky contacts |
| 5% to 15% | Meaningful cleanup needed | Verify everything, then compare the cost per usable contact with other sources |
| Above 15% | The export is mostly unverified | Treat it as a discovery list, not a sending list, and verify before any send |
Price the export by its output, not its input:
Cost per usable contact = total spend ÷ number of Valid contacts
Suppose a hypothetical export costs $1,200 for 12,000 contacts. The naive price is $1,200 ÷ 12,000 = $0.10 per contact. But if only 8,040 are Valid, the real figure is $1,200 ÷ 8,040 = about $0.15 per usable contact, 50% higher. If that gap is too wide, compare against the alternatives in our guide to the best email finders for B2B prospecting. For the cost of verification itself, see email verification pricing.
Bottom line: A list is only as cheap as its Valid contacts, so price it on that basis.
How do you keep prospect emails from hurting deliverability?
Verification before sending protects your sending domain and reputation. Three thresholds matter:
- Keep bounces under 2%. This is our recommended ceiling. It leaves room below the roughly 5% level where Amazon SES's documentation describes review, and below the point where many ESPs act sooner. Our guide to what bounce rate is safe for cold email explains the reasoning.
- Keep spam complaints under 0.1%. Sources differ here. Google's sender guidelines recommend keeping user-reported spam below 0.10% and warn against reaching 0.30%, and some older guidance quotes 0.3% as the line. We recommend 0.1% as the target, because it gives a safety margin. Bounces and complaints are separate signals, so control both.
- Re-verify on a schedule. Re-check before each campaign. For active prospecting lists, we suggest a full re-verification at least quarterly, because job changes make a list decay faster than most teams expect. The email list maintenance best practices guide covers the cadence.
When contacts enter your system, validate them at the point of entry as well. Real-time email validation prevents bad addresses from reaching the list in the first place.
If your sequencer has already started warning you, these walkthroughs help: Smartlead's bounce rate fix, verifying Instantly leads before launch and verifying Apollo lists before sending. For a full pre-launch sequence, see the cold email list verification workflow.
Verification doesn't establish consent. In the US, the FTC's CAN-SPAM compliance guide sets out requirements for commercial email, including opt-out mechanisms and accurate sender information. Clean data supports email verification compliance and hygiene, but it doesn't replace those obligations.
Bottom line: Verify before each batch, keep bounces under 2%, keep complaints near 0.1%, and treat verification as one part of compliance, not all of it.
What are the most common mistakes in an accuracy review?
- Quoting the vendor's figure as your result. Your sample is the evidence. The vendor's number is only a claim.
- Sampling the top of the export. The first rows are often the most recently added or most complete. Use a random draw.
- Reading a catch-all Valid as confirmed. A catch-all server accepts everything, so its results don't prove the mailbox exists.
- Calling a Valid result "verified identity." Deliverability and identity are different measurements.
- Sampling too few contacts. Twenty contacts can't support a precise percentage. Report the margin of error with every figure.
- Changing settings between the sample and the full run. The results won't be comparable.
- Sending to Risky contacts first. They belong at the end of the ramp, after the Valid list has performed.
- Missing the download window. Free-plan results are kept for 24 hours only, so export immediately.
- Assuming one segment represents the rest. A figure for US SaaS may not hold for other regions or industries.
- Verifying once and assuming the list stays clean. Contact data decays, so re-check on a schedule.
Bottom line: Most weak reviews come from unrandomised samples, unstated margins and reading "deliverable" as "correct."
How we compared (and limitations)
This article draws on public documentation: the RFC 3463 status-code definitions, Google's sender guidelines, Amazon SES's published bounce thresholds, the FTC's CAN-SPAM guide and Unlimited Verifier's published plan details. We haven't tested Seamless.AI's dataset, so this article gives no verified accuracy figure for Seamless.AI.
Keep these limitations in mind:
- The 99.5% accuracy figure for Unlimited Verifier's deep verification is the company's stated figure, not an independent benchmark.
- All worked-example numbers are hypothetical. The arithmetic and plan limits are real.
- Sample-size and margin-of-error figures assume simple random sampling. Clustered contacts widen the true margin.
- Verification measures deliverability. It doesn't confirm identity, currency or consent.
- Vendor thresholds and sender policies change without notice, so check current documentation before relying on any figure here.
Bottom line: Use this method to measure your own list, and treat every external number as something to confirm.
Sample size and bounce estimate
| Item | Value |
|---|---|
| Random sample size | 385 contacts |
| Margin of error (95% confidence) | about ±5 points |
| Larger sample for ±3 points | about 1,068 contacts |
| Export size (hypothetical) | 12,000 contacts |
| Measured Invalid rate (hypothetical) | 12% |
| Estimated bounces if sent unchecked | about 1,440 |
Frequently asked questions
Is Seamless.AI email accurate?
The honest answer is that it depends on the segment you pull, and no independently verified figure is available here. Measure accuracy on your own random sample of at least 385 contacts, then use the Invalid and Risky shares to judge the export before sending.
How many contacts should I test to check Seamless.AI accuracy?
At least 385 random contacts gives a margin of error of about ±5 points at 95% confidence. If you need tighter results, about 1,068 contacts gives roughly ±3 points. Pick contacts randomly rather than the best-looking rows, or the result will be biased.
What is the difference between deliverable and accurate email data?
A deliverable email is one that accepts mail, which a verifier can check. An accurate contact also needs the right person, company and title. A Valid result proves deliverability only, so identity and currency need a separate manual check.
Why does a verified email still bounce?
Verification checks whether a mailbox appears to exist, but receiving servers may not report honestly, and catch-all domains accept mail for any address. Bounces can also come from addresses that were valid when verified but have since been closed.
What is a catch-all email and should I send to it?
A catch-all address belongs to a domain that accepts mail for any local part, so a verifier cannot confirm the specific mailbox. Hold Risky and catch-all addresses until your Valid list has performed well, then send to them in small batches.
Should I trust a vendor's accuracy percentage?
Treat it as an unverified claim until you test it. Ask what counts as accurate, which population was tested, how recent the data is, and who ran the test. A vendor auditing its own data is making a claim, not running an independent audit.
How do I calculate the Invalid share from my sample?
Divide the number of contacts marked Invalid by the total number of contacts in your random sample, then multiply by 100. Calculate the Risky share the same way. Use those percentages to estimate how many bounces the full export would produce.
How many bounces should I expect from an uncleaned export?
Multiply the export size by the measured Invalid rate. For example, a 12% Invalid rate on 12,000 contacts implies about 1,440 bounces if you send unchecked. Use your own measured rate rather than a vendor figure.
How often should I re-check Seamless.AI contacts?
Re-check whenever you pull a new export or before a major campaign, because B2B contact data decays as people change jobs. A contact verified months ago may no longer hold the same role or employer.
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 standard verification on all plans, including Free, with no per-email fees.
- Catch-all detection so you know which domains accept every address.
- Clear Valid / Risky / Invalid results, exportable as CSV, Excel or JSON.
- An open-source verification engine you can inspect on GitHub.
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.