We Scanned 13 Web Agencies' Own Domains. Not One Scored Above 70.
2026-10-29 · 7 min read · By the DomainGuard team
Web agencies build everyone else's digital front door. They register the domains, ship the websites, set up the email, and hand over the keys. So we asked a simple question: who checks the agency's own domain?
We ran DomainGuard's free impersonation report — the same one anyone can run, no account, no credit card — on 13 agency domains taken from public agency directories. The report looks at three things an attacker looks at first: the certificates publicly logged for your name, the lookalike domains strangers have already registered, and your CAA records, the DNS setting that says which companies are allowed to issue certificates for you.
One honest note before the findings: this is a sample of 13 completed reports from a 51-domain candidate list, run on October 28–29, 2026. The free tool deliberately limits how many reports one visitor can run per day — each report does real work — and we worked within the tool's own limits plus its public API's free tier rather than around either. The remaining 38 domains are still being scanned at that pace. But the first 13 already tell a remarkably consistent story.
Finding 1: The lookalikes are already registered — 269 of them
Across 13 agency domains, the report found 269 lookalike domains already registered — between 12 and 44 per agency, out of 7,599 name permutations checked. Every single domain had at least a dozen. These are the classic impersonation spellings: swapped letters, missing letters, hyphenated versions, the .com/.co/.net twins of the real name.
A registered lookalike is not a theoretical risk. It means a stranger already owns the version of the agency's name that a tired client is most likely to type at 11 p.m. And most of them are not parked: 186 of the 232 lookalikes the reports could display carry MX records, which means they are set up to send and receive email. Whoever owns one can write to the agency's clients from an address that looks, at a glance, exactly right.
What does that cost? The standard play is a fake invoice or a "our bank details have changed" email, sent from the lookalike, timed to a real project the attacker can see on the agency's own portfolio page. The client pays the wrong account. The money is usually gone. And the phone call the client makes next is not to the stranger — it's to the agency. Agencies sell trust. A lookalike they never knew about spends it for them.
Finding 2: 10 of the 13 publish no CAA record at all
CAA is a small DNS record with an outsized job: it lists the certificate authorities allowed to issue certificates for your domain. Publish it, and every other authority on Earth must refuse. Don't publish it, and any authority can issue a certificate for your name — including, in practice, for a lookalike operation that passes a routine domain check.
10 of the 13 agencies publish no CAA record. Not a weak one, not a partial one — none.
Put the two findings together and the attack builds itself: register the lookalike (done — 269 times), put mail on it (done — 186 times), get a valid certificate for it (nothing in the real domain's DNS posture stands in the way), and the clone of the agency's login or payment page arrives with a working padlock. The padlock was never proof of identity. It just feels like it.
Finding 3: One DNS record is the entire difference between 45 and 70
The report scores each domain from 100, deducting 25 when there is no CAA protection and up to 30 for registered lookalikes. The 13 agencies landed on exactly two scores: ten scored 45, and three scored 70. Nobody scored above 70.
That split is not a mystery, and it isn't about size, budget, or sophistication. All 13 maxed out the lookalike penalty — every one has more registered lookalikes than the penalty counts. The three that scored 70 did exactly one thing the other ten didn't: they publish a CAA record. That single DNS record — minutes of work, published once, effective permanently — is worth 25 points and is the entire spread of this study.
The fix, when an agency gets around to it, takes minutes: one DNS record naming your real certificate authority. It is possibly the highest-value five minutes in domain security, and in this sample, 10 of 13 hadn't taken them.
Finding 4: A quiet blind spot in certificate logs
One more observation, because it changes how monitoring should work: for 7 of the 13 domains, the report found zero certificates in the public CT logs under the exact domain name. That does not mean these businesses don't use HTTPS — it means their certificates live under names this exact-name view doesn't capture. An agency that watches only its exact name in the certificate logs is watching a door the attacker was never going to use. The variants — the lookalikes, the subdomains, the campaign names — are where a rogue certificate would actually appear.
A note on names
We are not naming the 13 agencies, and we won't publish problem examples with names attached — that's not the point of this study. Our standing rule for this series is that agencies scoring 80 or above get named as positive examples. In this sample, none did — the ceiling was 70, set by the lookalike penalty nobody escaped. If your agency would score 80+, we would genuinely like to hear how you keep your lookalike count that low.
What "doing it right" looks like
Nothing in this study requires a security department. The checklist is short enough to steal:
- Publish a CAA record naming your certificate authority. Minutes of work, permanent effect — and, in this sample, the single biggest score difference there is.
- Inventory your lookalikes. Run your own name through a checker and read the list like an attacker: which of these would fool your busiest client?
- Watch the mail-enabled ones. A lookalike with MX records is a loaded weapon; it's the one to monitor, dispute, or take down first.
- Monitor certificates for your name and its variants, so a rogue cert shows up in your inbox before it shows up in a client's browser.
- Have the takedown path ready before you need it. Registrars move fastest when the evidence pack is already assembled.
Agencies do versions of this for their clients every week. The gap this study keeps finding is that nobody does it for the agency.
Check your own domain
Your score takes about a minute, costs nothing, and shows you your certificates, your registered lookalikes, and your CAA posture in one page — the same report this study is built on.
If the report finds lookalikes already registered against your name, DomainGuard's paid plans exist for exactly the next step: continuous monitoring of the names around yours, and takedown packets assembled for filing. But start with the free report. What you do with the list is your call — seeing it is not optional anymore.
Method: DomainGuard's free impersonation report (domainguard.getingeniousai.com/report) and the same scanning engine consumed through DomainGuard's public API on RapidAPI, run sequentially from a single connection on October 28–29, 2026, on agency domains sourced from public agency-directory listings (The Manifest, Clay, DesignRush and publisher lists). Sample: 13 completed reports of a 51-domain candidate list; the free report's own limit (3 per visitor per day) and the API free tier's monthly request quota set the pace, and both were respected rather than routed around — the remaining 38 domains continue at the free report's pace. Scores follow the report's published arithmetic: 100, minus 25 for a missing CAA restriction, minus 5 per registered lookalike up to 30. No agency with a problem finding is named; agencies scoring 80+ will be named as positive examples, with their permission sought first.