Compare the spelling, not just the shape
Compare Punycode and Lookalike Domain Names
A bookseller preparing a multilingual catalogue may use a domain containing an accented letter. A customer may then see an unfamiliar ASCII spelling beginning with xn-- and ask whether the link changed. That can be an ordinary internationalized domain name, not an attempted deception. The opposite problem also exists: a name can look familiar while using different characters. A useful review keeps these questions separate and compares exact hostnames instead of deciding from their appearance.
Use the Link Inspector to read the ASCII and Unicode views of one supplied HTTP or HTTPS URL. The Unicode view uses tr46 processing to make an internationalized hostname readable. It does not search for impersonated brands or detect every visual lookalike. Begin with a teaching example, then compare the relevant spelling against a reference obtained independently from the message you are reviewing.
Read two representations without inventing two destinations
Punycode is an ASCII encoding used within internationalized domain labels. The prefix xn-- is a useful clue that a label uses this representation, but it is not a warning label meaning dangerous. For a reserved teaching hostname, bücher.example has the ASCII form xn--bcher-kva.example. These spellings can describe the same internationalized hostname. The accented character belongs to the name, not to a separate redirect. This example explains representation only; it is not a working shop address or a claim that the reserved example suffix identifies an operating website.
Paste one complete URL and select Inspect link. The result separates the hostname, its ASCII and Unicode views, the public suffix, the registrable domain where available, scheme, and effective port. Treat each field as an answer to a different question. ASCII helps with exact hostname comparison; Unicode helps a reader recognize the intended writing system. The domain boundary helps locate the meaningful labels within a longer host. No one field replaces the others, and none identifies the person who registered the name or currently controls the content at that address.
Use Load example first when showing this distinction to a colleague. Results remain plain text, never active destination links, so reading the example does not open its host. Clear removes the working example before you inspect a real address. If you receive only a domain spelling without a scheme, ask for the complete link when the full address matters. Adding https:// yourself creates an address for a spelling exercise; it does not establish what the original message linked to or which scheme the sender actually supplied.
Compare exact labels instead of judging the font
Start with a trusted reference that is independent of the questioned message. For the bookseller, that could be the hostname recorded when the shop domain was set up. Compare the ASCII hostname label by label, including hyphens, digits, and the ending. Do not use a screenshot of the same message as the reference: it repeats the original claim. If the source record uses Unicode, inspect a complete reference URL separately and take its ASCII hostname for comparison. Copy the first hostname before switching tabs, because changing tabs clears the working inspection. Also confirm that the reference is for the same purpose. A shop may intentionally use different hosts for its catalogue and support desk; a mismatch needs explanation, not an automatic impersonation accusation.
Visual confusion does not require an internationalized name. The ASCII teaching names catalogue.example, cata1ogue.example, and catalogue-help.example are different spellings even if a small font makes the differences easy to miss. An added word can look reassuring while naming a different domain. Unicode can add further visual similarities: the Cyrillic character а is not the Latin character a. A reader may perceive a familiar word without noticing that substitution. Neither the number of similar letters nor the attractiveness of the displayed name establishes a business connection.
The inspector does not calculate a similarity score, identify a copied logo, or name a brand that a domain might resemble. Its tr46 Unicode view is a representation feature, not a homograph detector. A successful conversion means you obtained a readable representation under that processing, not that an imitation check passed. When you cannot confidently compare unfamiliar writing systems, keep the ASCII spelling available and verify the expected address with the known organization. There is no need to accuse a domain of abuse simply because you need help reading its characters. Ask the organization to provide its expected spelling rather than asking whether a blurred image looks right. A text comparison is easier to repeat and less dependent on the font used in a message preview.
Find the domain boundary before comparing names
A perfect match in one label can still be irrelevant if that label is in the wrong place. Compare https://catalogue.example.com/ with https://catalogue.example.com.help.example.net/. The second hostname contains all the familiar words from the first, but its registrable domain is example.net. The extra labels sit to the left of that boundary. This is a structural difference that exact character reading alone might miss. First identify the full host, then its reported domain boundary, and only then compare the relevant labels with the independent reference.
The boundary comes from tldts using the Public Suffix List, including private entries. A suffix can contain several labels, and a hosted-site platform can have a private suffix rule that separates its customers. This is why simply taking the last two dot-separated parts is unreliable. For an ordinary multi-label suffix example, shop.example.co.uk has co.uk as its public suffix and example.co.uk as its registrable domain. Those labels describe a boundary; their presence is not a statement about registration status, ownership, or the honesty of a shop.
Keep the full hostname in any comparison notes, not just the registrable portion. Two hosts under the same boundary can serve different purposes, and a hosted-site boundary is not proof that the platform endorses an individual page. If the result identifies an IP address instead of a domain, Punycode is not the relevant reading task. If no useful registrable domain is available, do not fill in an owner from the most familiar substring. Record what the parser actually supplied and leave questions about operators, accounts, and commercial relationships unanswered until you have separate evidence.
Two views, one hostname
Compare the exact domain spelling
Read the ASCII and Unicode representations without opening the supplied destination or assigning it a trust score.
Distinguish normal conversion from rejected input
The native URL parser canonicalizes a supplied address without visiting it. Hostname case, encoded host characters, default ports, and path segments can therefore look different in the normalized result. For an internationalized host, the ASCII representation may be less recognizable than what was pasted. That alone does not describe a redirect or a spelling attack. Read any normalization findings and compare the resulting host with your reference. Do not promise that the normalized URL preserves every original visual detail, or confuse a change in presentation with an observed change made by a remote server.
This input accepts one absolute http:// or https:// URL and rejects relative addresses, custom schemes, control characters, bidirectional formatting controls, backslashes, and internal spaces. That is not a blanket rejection of non-Latin domain names. For example, ordinary letters in a right-to-left writing system are different from inserted bidirectional formatting controls. If a copied message contains forbidden formatting, obtain a clean address from a reliable source. Removing characters by guesswork can change the name you are examining and make your conclusion about the original text unreliable.
Outer ASCII spaces are trimmed with a finding, while the address must fit within 8,192 code points, 32 KiB of UTF-8 data, and 200 query fields. These limits are separate from whether an IDN spelling can be converted. An error does not mean the tool detected a malicious homograph, and acceptance does not mean a name is legitimate. Keep any nonsecret error description when asking for clarification. Avoid pasting the full private address into a team discussion merely to explain that an input limit or a forbidden-character rule was reached.
Share the hostname without exposing unrelated secrets
For a spelling review, Copy hostname is usually the smallest useful handoff. The path, query, and fragment rarely help a colleague decide whether two hostname spellings match. They may instead expose a catalogue preview token, customer identifier, private folder, or invitation. User information can also contain credentials. Choose Copy inspected URL only when someone genuinely needs the complete normalized address and is meant to receive those values. Do not assume that an address is public because the hostname is public or because the link arrived in a shared business inbox.
Reveal details displays the normalized URL and its path, ordered query pairs, fragment, and any user information; those details are masked by default. Hide details restores masking without ending the session. If your question really concerns a path label, reveal it privately and describe only the necessary distinction. A non-ASCII word in a path is not the same issue as a non-ASCII hostname label. A company name inside a query value is also not a host. Keeping these locations separate prevents a spelling discussion from turning into a misleading claim about the destination.
Do not edit a signed or token-bearing address to make it easier to circulate, then present the edited version as the original. If the actual task is preparing a known public catalogue link for publication, the separate URL Cleaner can help review attribution parameters. That is a different task from comparing internationalized hosts. Inspect the edited result again if its structure matters, and retain the distinction between an original link and a publishing candidate. The inspector neither removes fields automatically nor decides which parameters a remote application requires.
End with a spelling finding, not a trust certificate
A useful conclusion for the bookseller might be: the Unicode and ASCII views correspond to the expected hostname. For an unexpected customer message, it might instead be: the ASCII hostname differs from the shop record in this label. Those are specific observations another person can check. Neither sentence proves who sent the message or what content the destination would serve. An HTTPS scheme is not proof of safety, and an effective port merely identifies the port implied or supplied by the address. Keep the commercial question separate from the spelling question. If the names match but the request is unexpected, the unanswered question is now about the request, not the spelling. Repeating the conversion will not supply evidence about an invoice, delivery, or account change.
There is no fetch, API call, DNS query, status check, TLS inspection, reputation lookup, or ownership verification. A short link is not expanded, and an embedded destination field is not an observed redirect chain. Findings are descriptive rather than a risk rating. If a message asks for an unexpected sign-in or payment, verify it through an established route instead of following the questioned address to settle the uncertainty. A structurally ordinary URL can still be used in a deceptive request, while an unfamiliar internationalized name can be a legitimate choice.
The working input is isolated in an opaque sandbox rather than an ordinary WordPress input, with no-store and connect-none restrictions. All state clears on hidden visibility, pagehide, pageshow, and freeze events, so copy the needed result before changing tabs. Clear ends the current working inspection too. There is no history, project, PDF, download, or stored report. Clipboard copies are not erased, and browser extensions or a compromised parent interface remain outside the protection. Finish by sharing only the spelling evidence needed for the decision, without creating another copy of private URL details.
Questions about Punycode and Lookalikes
Does an xn-- label mean that a domain is suspicious?
No. That prefix identifies an ASCII representation used for internationalized domain labels. Read the Unicode view for context and compare the exact ASCII hostname with a trusted reference. The prefix alone does not tell you who operates the site or whether a message linking to it is genuine.
Does the Unicode view detect all lookalike domain names?
No. The tr46 view converts the hostname representation; it is not a homograph detector or brand comparison service. Similar-looking ASCII names can also mislead. Compare the full hostname and domain boundary, and use an independently obtained reference when the identity of an organization matters.
Can I compare two private links by leaving one in another tab?
Not as retained inspector sessions. Hiding a page clears its input and derived results, and lifecycle transitions also clear state. Copy only the necessary hostname before switching, then inspect the reference separately. Clipboard copies remain outside that clearing behavior, so avoid copying complete private URLs when the hostname will do.
Continue with another link-inspection guide
Inspect a supplied address in the Link Inspector or choose a related task below. Explore more practical utilities in all free tools.
Keep the conclusion specific
Check the spelling in the next address
Use the free Link Inspector to separate hostname representation, domain boundaries, and the other parts of a URL.