How to Investigate a Suspicious Link Safely
Read a URL properly, check a domain's age and registration, expand redirects and sandbox the page, without putting your own accounts or device at risk.

A colleague forwards you a message and asks whether it is real. The link reads https://login.microsoftonline.com.auth-session-verify.co/validate?id=8831f. Half the people who see that will decide it is Microsoft, because the first thing their eye lands on is a domain they trust. It is not Microsoft. It is a host called auth-session-verify.co, which has been given a subdomain designed to read like one.
Working that out took five seconds and no tools. Most link investigation is like this: a short sequence of cheap checks, in a specific order, with the expensive and risky steps saved for last. What follows is that sequence, along with the things not to do, which matter more than the techniques.
Rule zero: what not to do
Each of these mistakes is common and each of them makes the situation worse.
- Do not click it on your working device. Your browser holds authenticated sessions, saved passwords and cookies. Those are exactly what a malicious page wants, and some of them can be taken without you typing anything.
- Never sign in to "test" whether a login page is fake. There is no version of this that produces useful information. If you want to know whether credentials are accepted, the answer is that fake pages accept everything.
- Do not forward it onward without a warning, and do not forward a live link into a group chat. Rename the scheme, insert spaces, or send a screenshot with the URL as text.
- Do not submit URLs containing secrets to public scanners. Password reset links, invoice portals, calendar links and anything with a token or an email address in the query string become publicly searchable once scanned. This single habit has caused real breaches.
- Do not treat https or a padlock as evidence. Certificates are free, instant and automated. Essentially all phishing is served over https now.
- Do not rely on a clean scanner result. Brand new infrastructure has no reputation, which is why attackers use brand new infrastructure. Absence of detections means the page has not been seen before, which is slightly suspicious in itself.
- Do not visit it from the office network and then declare it safe. Many malicious pages serve harmless content to datacenter addresses, scanners and repeat visitors, and show the real page only once, to a residential address, on a mobile device.
Step 1: Get the URL as text, without following it
In a desktop email client or browser, right-click the link and copy the link address, then paste it somewhere plain. Do not read the displayed text, which the sender controls completely and which routinely shows one address while pointing at another.
Where that is not enough:
- On a phone, press and hold the link to see the destination rather than tapping it.
- In HTML mail, view the message source and read the href attributes directly. This also reveals tracking pixels and any additional hosts the message loads from.
- In an attachment, extract the link without opening the document in its normal application where possible. A PDF or an HTML attachment is a delivery mechanism in its own right, and opening it to find the link is the thing you are trying to avoid.
- From a QR code, decode it to a string instead of letting your camera navigate. Most phone cameras preview the URL first, so read it there. See QR code phishing for why this route is popular.
Step 2: Read the URL properly
This step resolves most cases, and almost nobody does it carefully. A URL has a fixed structure: scheme, then an optional userinfo section, then the host, then an optional port, then the path, then the query, then the fragment.
The host is the only part that tells you who you are talking to. It runs from after the double slash to the first single slash. Everything after that first single slash is chosen by whoever controls the host and can say anything at all.
Inside the host, read from the right:
- Find the public suffix, the registry-controlled ending such as .com, .co.uk, .io or .com.au. The Public Suffix List is the authoritative record of which endings are suffixes rather than registrable names.
- Take one label to the left of it. That pair is the registrable domain, and it is the thing somebody paid for. In secure.paypal.com.account-review.info, the registrable domain is account-review.info.
- Treat everything further left as decoration. Subdomains are free and unlimited, so brand names appearing there mean nothing.
Then look for the specific tricks:
- The userinfo trick. Anything between the double slash and an @ sign is credentials, not a host. In https://www.yourbank.com@192.0.2.10/login the destination is the address after the @.
- Lookalike characters. Internationalized domains can use characters that render like Latin ones. Browsers often display these as punycode beginning with xn--, and seeing that prefix in a domain claiming to be a major brand is close to conclusive.
- Hyphenated brand names, such as microsoft-support-login.com. Large organizations do not usually authenticate users on hyphenated variants of their name.
- Percent-encoding used to obscure, where parts of the host or path are written as %xx sequences. Decode it before judging it.
- Open redirects on a legitimate domain, where a genuine host carries a parameter containing another URL. The visible domain is real and the destination is not. Check every parameter that looks like a URL.
- Raw IP addresses in place of a hostname, which almost never appear in legitimate consumer-facing links.
- Credential-shaped query strings, such as a parameter containing the recipient's own email address, which pre-fills the fake login form and makes it more convincing.
Our shorter companion piece, how to check a link without clicking it, covers this at a quicker pace for non-technical readers.
Step 3: Expand shorteners and redirect chains
A shortened link tells you nothing by itself, so resolve it without loading the page. From a command line, request headers only and follow redirects, for example curl -sIL "https://example-shortener/abc" and read each Location header in the chain. Use a disposable environment for this, because even a header request reveals your address to the server.
What to note as you go: how many hops there are, whether the chain passes through a legitimate marketing or tracking domain before landing somewhere else, and whether the final host differs from the one the message implied. Long chains that end on a domain unrelated to the brand are a strong signal.
Two caveats. Servers can detect non-browser clients and return a different destination, so a benign expansion is not an all-clear. And some chains only complete with the right referrer, user agent or cookie, meaning you may see a decoy. Our article on shortened links and redirects goes into the evasion techniques.
Step 4: Check the domain's registration and age
Now you have a registrable domain, look it up. WHOIS and its structured successor RDAP return the registration record, and ICANN's own lookup service at lookup.icann.org is a reliable place to query it.
Read these fields:
- Creation date. The most useful single data point in link analysis. A domain created yesterday or last week that presents itself as an established bank, courier or supplier is almost certainly malicious. Phishing infrastructure is usually days old because it gets blocked quickly.
- Registrar and registrant country, which may be inconsistent with the brand being imitated.
- Privacy or proxy registration. Common and legitimate for individuals, much less so for a company inviting you to log in to a financial account.
- Name servers, which indicate the hosting provider and sometimes group a whole campaign together when several lookalike domains share one.
- Expiry date, where a one-year registration on a supposedly long-established brand is telling.
Also compare it with the real thing. Look up the brand's genuine domain and note the registrar, creation date and name servers, then see how different the suspect record looks. Certificate transparency logs are another way to see when a certificate for that host first appeared, which often matches the domain's age closely.
Step 5: Check prior sightings and reputation
Before you generate new traffic, see what others have already recorded. Search urlscan.io for existing scans of the domain, which may show you a screenshot of the page without you touching it. Check the URL and the domain on VirusTotal, and check Google's Safe Browsing site status page.
Interpret the results carefully. Several detections on a well-aged domain usually means a compromised legitimate site, which changes your response because the owner needs telling. Zero detections on a domain registered three days ago means nothing other than that you may be early. Look at the first-seen dates and the associated files or domains, as those relationships often reveal the broader campaign.
And remember the privacy rule from the start of this guide. If the URL carries a token, an identifier or an address, search by domain rather than submitting the full URL, or use a private scan.
Step 6: Look at the page in something disposable
Only now, and only if you still need to know. Options in rough order of convenience: a private scan on a scanning service, a remote browser isolation session if your organization has one, or a throwaway virtual machine on a network segment that touches nothing you care about, with no credentials, no saved sessions and a snapshot you revert afterwards.
When you look, capture rather than browse. Record the final URL after all redirects, a screenshot, the page title and any brand assets used, where the form actually submits to, which fields it asks for, whether it asks for an MFA code as well as a password, any file it offers to download, and the hash of that file. A page that requests a one-time code is running a real-time proxy attack rather than simple credential collection, and that distinction matters for your response.
Close the session, revert the machine, and do not revisit from a normal device to check something you forgot.
Want a quick second opinion first? Paste the message or link into CyberWatch AI Scan for a free check.
Step 7: Decide, then act on the whole blast radius
The investigation is not the point. What you do next is.
- Block the domain at your DNS filter, web proxy or email gateway, and block the sibling domains on the same name servers if you found a cluster.
- Search your mail logs for the domain and the sender to find every other recipient. One forwarded message almost always means several delivered ones.
- Find out who clicked. Proxy logs, DNS query logs or your email security product will tell you. Clicks matter less than submissions, but you cannot tell them apart from logs alone, so ask the people directly.
- If anyone entered credentials, reset the password, revoke all active sessions and tokens, check for new mailbox rules and forwarding, and review recent sign-ins. Do this immediately rather than after confirming misuse. Our guide on what to do when someone has entered a password into a phishing page sets out the order.
- Report it. Use your internal reporting button so the detection improves, and report externally to your national reporting service and to the brand being impersonated, which is how takedowns start.
- Write down what the signal was. The specific tell that resolved this case is training material, and it is more useful than any generic advice.
If the clicking has already happened, what to do after clicking a phishing link is the next thing to read, and our complete guide to phishing covers the wider pattern these links belong to.
Frequently asked questions
Where is the real domain in a long URL?
Read the host portion from the right, not the left. Take the registrable domain, which is the label immediately before the public suffix, so in login.microsoft.account-verify.example.ru the real domain is example.ru and everything before it is a subdomain the owner chose. Anything after the first single slash is path and tells you nothing about ownership.
Is it safe to just click the link to see where it goes?
Not on a device holding your work sessions, saved passwords and cookies, and never while signed in. Visiting a page can fingerprint your browser, confirm your address is live, start a download or serve an exploit. Use a disposable environment or a scanning service instead.
Can I paste the URL into a public scanner?
Usually yes, with one important exception. Public scans are visible to others, so a URL containing a token, a password reset code, an invoice reference, an email address or an account identifier should never be submitted publicly. Use a private scan or strip the parameters if the result will still be meaningful.
How do I check how old a domain is?
Look up its registration record through a WHOIS or RDAP lookup, such as ICANN's lookup service, and read the creation date. A domain registered days ago that claims to be an established bank or supplier is strong evidence on its own, though age alone does not make a domain safe.
Does a padlock or https mean a site is legitimate?
No. Certificates are free and automated, so almost every phishing page has one. The padlock means the connection is encrypted, not that the other end is who it claims to be.
How do I expand a shortened link without visiting it?
Request the headers only and follow redirects without downloading the body, for example with curl -sIL, or use a link expansion or scanning service. Be aware that servers can return a different destination to a scanner than to a real browser, so a clean expansion is not proof.
What if the link was in a QR code?
Decode the code to text rather than letting your phone open it. Most phone cameras show the URL before navigating, so read it there, or use a decoder that returns the string so you can analyze it like any other URL.
What should I do once I have confirmed it is malicious?
Report it, block the domain at your DNS or web filter, search mail logs for everyone else who received it, and find out whether anyone clicked or submitted credentials. If credentials were entered, reset the password and revoke active sessions immediately rather than waiting for evidence of misuse.
Sources
- URIs, MDN Web Docs
- RFC 3986: Uniform Resource Identifier (URI): Generic Syntax, RFC Editor
- Public Suffix List, Mozilla Foundation
- Registration Data Access Protocol (RDAP), ICANN
- URL and website scanner - urlscan.io, urlscan.io
- VirusTotal, VirusTotal
- Google Transparency Report: Safe Browsing site status, Google
- Recognize and Report Phishing, CISA
- Phishing scams: how to spot and report them, UK National Cyber Security Centre


