EmailCheck / Questions问题页 / How accurate are email verification APIs, really?
How accurate are email verification APIs, really?邮箱验证 API 到底有多准？
Less accurate than the marketing says. In Hunter's public benchmark — 3,000 real business emails run through 15 verification services — the best service scored 70%. Treat any “99% accurate” claim as a sales line: judge a verifier by what it checks and what evidence each result carries, not by a percentage it awards itself.
比营销话术说的低得多。在 Hunter 的公开基准里——3,000 封真实商务邮箱、跑 15 家验证服务——最好的一家也只有 70% 准确率。任何「99% 准确」的说法都当销售话术看：判断一个验证工具，要看它查什么、每个结果带什么证据，而不是看它给自己打的百分比。
The benchmark, and why it matters那份基准，以及它为什么重要
Accuracy claims in this category are usually self-graded: each vendor picks its own test set and its own definition of a correct answer. Hunter's test is the reference point because it was run by a third party, on real business addresses, across 15 competing services at once — and the ceiling was 70%. If the best-measured service in the category gets three in ten wrong (by that test's definitions), a homepage that promises 99% is telling you about its copywriting, not its engine.
Part of the gap is structural, not incompetence. Roughly 30–40% of business addresses sit on catch-all domains, where a definitive answer does not exist without sending mail. Any accuracy figure that quietly includes or excludes that third of the list can be made to say almost anything.
这个品类的准确率宣称通常是自己给自己打分：每家挑自己的测试集、自己定什么叫答对。Hunter 的测试之所以是参照点，是因为它由第三方执行、用真实商务地址、一次横评 15 家互相竞争的服务——天花板是 70%。如果品类里被测得最好的一家也有三成「错」（按那份测试的定义），某个首页写着 99%，那说的是它的文案水平，不是引擎水平。
差距有一部分是结构性的，不是能力问题。商务地址里约 30–40% 在 catch-all 域名上，不真发信就没有确定答案。任何准确率数字，只要悄悄把这三分之一算进或排除，几乎能说成任何样子。
What to evaluate instead of a percentage不看百分比，看什么
- The check list. Syntax, domain, MX, disposable, role accounts, typos, domain age — are they enumerated, and does each result show them individually?
- Evidence per result. A score with its working shown (which checks passed, which failed, what was suggested) can be audited; a bare verdict cannot.
- Stated scope. Does the vendor say, in the response itself, that mailbox existence is not probed? EmailCheck puts
scope: "domain-level"and a plain note in every response. - Behaviour on the hard cases. Run the same awkward addresses — a typo domain, a disposable domain, a catch-all domain — through any tool you are evaluating and compare what comes back. This page's check box is a good place to start.
- 检查清单。格式、域名、MX、一次性、角色账号、拼写、域名年龄——有没有逐项列明，每个结果里有没有逐项展示？
- 每个结果的证据。带过程的分数（哪些检查过了、哪些没过、建议是什么）可以审计；光秃秃的裁决不行。
- 写明的范围。厂商有没有在响应本身里写清「不探邮箱存在」？EmailCheck 每个响应都带
scope: "domain-level"和大白话说明。 - 难例上的表现。拿同一批别扭地址——拼错域名、一次性域名、catch-all 域名——跑任何你在评估的工具，比一比返回什么。本页的检查框就是个好起点。
Try it on a real address拿真实地址试一下
Live call to the production API — the same check described on this page.
实时调生产 API——就是本页说的这个检查。
Scope, stated plainly: EmailCheck checks domains, not mailboxes — no SMTP probing, no claim that a specific mailbox exists. Every API response carries scope: "domain-level" and a note saying exactly what was checked.
范围说死：EmailCheck 查的是域名，不是邮箱——不做 SMTP 探测，不声称某个具体邮箱存在。每个 API 响应都带 scope: "domain-level" 和一段说明，把查过的范围写清。