How a website vulnerability scan works
A website vulnerability scan on Scanr has three parts: you prove the site is yours, the scanner reads it the way an outsider could, and you get a list of what it found with the proof and the fix. By default it only reads. It sends a form or signs in only when you allow it, with test text and test accounts, once per scan.
What you need
- An email address. You sign in with a link sent to your inbox. There’s no password.
- Access to your domain’s DNS. Wherever you bought the domain, or wherever its DNS is managed.
- A way to edit your site. Enough to add one line to the home page or upload one small file.
1. Add your site and say it’s yours to test
Sign in, open Sites and type your domain. Before the site is added you tick a box saying you own it, or have the owner’s written permission, and have authority to approve the scan. That statement is saved with your account.
2. Prove it with two checks
This works the way verifying a site with Google Search Console does, except Scanr asks for both proofs, not one.
| Proof | What you add | What it shows |
|---|---|---|
| DNS record | A TXT record that starts with scanr-verify= | You control the domain |
| Site snippet | A scanr-verify meta tag on the home page, or a file at /.well-known/scanr-verify.txt | You control the website itself |
Both are checked again at the start of every scan. Take either one away and the next scan refuses to start.
3. The scan reads your site
Press Scan. The scanner works through five groups of checks, all from the outside, the way a stranger on the internet would see your site:
| Check | What it looks for |
|---|---|
| Home page and settings | Missing security headers, weak cookie settings, whether visitors are sent to https |
| Files that shouldn’t be public | Environment files with passwords and keys, backups, a public .git folder |
| Open doors | Folder listings, admin pages anyone can reach, settings that let other websites read your users’ data |
| Your JavaScript | Live API keys left in code every visitor downloads, source maps, an open GraphQL schema |
| Pages it finds | The same checks on a handful of ordinary pages linked from your site |
Every request passes one gate in the scanner’s code before it leaves. The gate only allows read requests to the site you verified. The full rule list is on How the scanner behaves.
4. AI reviews what was found, without touching your site
The AI never sends a request to your site. It gets what the scan collected and does three jobs: it picks which of your ordinary pages are worth a look, it tries to disprove each finding, and it writes the summary. A finding ends up as confirmed, ruled out, or not proven either way, and the last kind stays in the report, labeled, so nothing is quietly dropped.
Values that look like secrets are masked before anything is stored or sent for review. The report tells you what kind of key it was and where, not the key.
5. Read the findings and fix them
Each finding has its own page with:
- Where it is. The exact address.
- The proof. What came back that shows the problem is real.
- Why it matters. What someone could do with it.
- The fix. Step by step, and who usually does it.
- Instructions for your coding agent. One button copies a prompt you can paste into the AI tool you build with.
Tick a finding off when it’s fixed, then scan again to confirm. Every scan also keeps a log of each request it sent and each one it refused to send.
Networks work the same way, with a stricter approval
For an office or restaurant network, an owner, officer or manager approves the review in writing first. Then one of two things happens:
- UniFi equipment. You connect a read-only key. Scanr reads the equipment list and settings from Ubiquiti’s cloud and never contacts a device.
- Any other router. You run a small scanner program on a computer inside the network. It lists the devices it can see, you tick the ones the business owns, and it checks only those, on a fixed list of 14 ports.
The scanner program won’t run on guest Wi-Fi, on a network that card readers share, or on a network someone else runs. Payment systems are never part of a review.
Optional: forms and test accounts
To go further than an outsider can, you can allow two things on the site’s page in the dashboard. Neither happens unless you turn it on.
- Forms. The scan lists the forms it finds. Tick the ones that are safe to send, like a contact form. Each ticked form is sent once per scan with “Scanr test - please ignore” in its fields, to see whether it shows technical error details or accepts sends from other websites.
- Two test accounts. Make two throwaway accounts on your site and add them, with a page only each one should see. Each scan signs in once with each, never guesses and never retries. Then it checks whether one account can open the other’s private page, whether that page opens with no sign-in, and how the sign-in cookie is set.
Test-account passwords are stored encrypted, used only to sign in, and never sent to the AI. Adding test accounts needs two-step sign-in on your Scanr account.
What a scan is not
Scanr is not a penetration test, a compliance audit or a PCI scan. It checks what an outsider can reach without logging in. A clean report means those checks found nothing. It doesn’t mean a site is secure.
Questions
How long does a website vulnerability scan take?
A few minutes. A scan runs in eight short steps and sends at most 600 requests, no faster than 4 a second, so it finishes quickly and doesn’t load your site.
Do I need to install anything on my website?
No software. You add one DNS record and either one line in your home page or one small text file. That’s the proof the site is yours.
Will a scan slow down or break my site?
It’s built not to. It sends read requests, skips links that look like they do something, sends a form or signs in only when you’ve allowed it, and stops at a fixed number of requests. A scan can still show up in your logs or trip a security alert, because it is checking for things attackers check for.
What does it cost?
The first scan on an account is free. After that the checks are still free; tokens only pay for the AI review, and every scan shows an estimate and a maximum before it starts.