Two confusion tests, and they are not the same test
Before anything else on this page: the UDRP first element and trademark infringement's likelihood-of-confusion inquiry are different tests, with different inputs, decided by different bodies. An analysis written for one does not answer the other, and an opinion that blurs them will be read as wrong by the attorney who commissioned it.
The UDRP first element is a threshold standing test resolved largely by comparing strings. Infringement analysis, which in the United States turns on whether a use is likely to cause confusion, mistake or deception (15 U.S.C. 1114), is a multifactor inquiry into consumer perception in an actual marketplace, and the factor lists are judge-made and vary between circuits, which makes them counsel's territory rather than an expert's. Between the two sits the ACPA, which asks whether a domain name is identical or confusingly similar to a distinctive mark while separately requiring bad faith intent to profit.
US-specific: the Lanham Act provisions, the ACPA and the Federal Rules of Evidence discussed below are United States law only. The UDRP is contractual and applies across generic extensions worldwide; country-code extensions run their own policies.
What the expert supplies in all three settings is the same thing: the comparison and the record beneath it. The legal standard applied to that comparison is not the expert's to choose.
The first element: a comparison, and what it is for
Paragraph 4(a)(i) of the UDRP requires a complainant to show the domain name is identical or confusingly similar to a trademark or service mark in which it has rights. The WIPO Jurisprudential Overview 3.0 records how panels have approached that, describing the element as a standing requirement involving a relatively straightforward, side-by-side comparison of the domain name against the textual components of the mark, to assess whether the mark is recognizable within the disputed name.
Two things about the Overview belong in any report that cites it. It is a synthesis of panel decisions published by the WIPO Arbitration and Mediation Center — it is not a policy instrument and it binds no panel. And it addresses the UDRP, not national trademark law.
The practical effect for an expert is to define the exhibit narrowly. The first-element comparison works on the mark as registered and the disputed string. It does not need registrant identity, it does not need the site's content, and it can therefore usually begin on day one of an engagement, before any disclosure request or subpoena has been answered.
Added terms, misspellings, and the extension
Three patterns recur often enough that the comparison table should handle each explicitly.
Added terms. Where the mark is recognizable in the domain name, panels have generally not treated the addition of other terms — descriptive, geographical, pejorative, meaningless or otherwise — as preventing a finding of confusing similarity under the first element. That is the combosquatting pattern: the mark plus a word, with no misspelling at all.
Misspellings. A common, obvious or intentional misspelling of a mark has generally been treated as confusingly similar for first-element purposes, with adjacent-key substitution, similar-appearing character substitution and letter inversion among the recorded examples. This is the hinge between this analysis and the variant-set work described under typosquatting evidence.
The extension. The top-level domain is normally disregarded in the first-element comparison as a standard registration requirement, whatever the extension. Two qualifications belong in the report rather than being skipped: the meaning of an extension may be relevant to the panel's assessment of the later elements, and where the second-level string and the extension together spell out the mark, the name is assessed in its entirety. Design and stylized elements of a figurative mark are largely disregarded for comparison purposes, because a domain name cannot reproduce them.
What the comparison exhibit actually contains
Four components, and each is a table or a figure rather than a paragraph of argument.
- The comparison table. The mark as registered — exact character string, class, registration number, status, first-use dates, pulled from the trademark office record and dated on retrieval — against the disputed string, with the second-level label isolated from the extension, the edit operations separating them enumerated, and the resulting character-level rendering shown side by side.
- The rendering exhibit. How the two strings appear in the contexts a user meets them: address bar, link text, sender line.
- The variant landscape. The other registrations in the same family, because both the added-terms point and the ACPA's multiple-registrations factor operate on the set rather than the single name.
- The limits statement. What the comparison does not address — actual consumer perception, marketplace conditions, the strength of the mark, and the legal conclusion.
Registration status is checked and dated rather than recalled. A mark can lapse, be cancelled, be opposed or be limited by disclaimer between its registration date, the domain's creation date and the filing date, and all three dates matter.
Rendering: the exhibit that plain text cannot carry
Some comparisons are unreadable in a written report. A homoglyph case — where a character from one script is visually indistinguishable from a character in another — looks like two identical strings on the page, which is precisely the point being made and precisely what the reader cannot see.
The working method is to render rather than describe. Show the strings as they appear in the environments a user encounters, and record the environment: font, operating system, browser, and whether the browser displays the Unicode form or the ASCII xn-- form. An exhibit that does not state the environment it was produced in is not reproducible, and reproducibility is the whole basis on which a technical exhibit is accepted.
Alongside the rendering, give the underlying data: the codepoint of each character and the script it belongs to. A screenshot with an assertion is not evidence of a confusable string. Codepoints are, and they survive re-typing, copy-paste and font substitution in a way an image does not.
Where a confusion exhibit goes wrong
Five failures account for most of the exhibits I have seen fall apart, and all five are avoidable at drafting stage.
- Comparing the full hostname including the extension, and presenting an extension difference as a distinction, when the first-element comparison normally disregards it.
- Leading with website content. Content is generally not determinative of the first element, though it may support a finding where a site trades off the complainant's reputation. An analysis that opens with archived screenshots has misordered the report and invites the response that it is arguing the wrong element.
- Importing infringement factors into a first-element analysis, or the reverse, and producing an opinion that answers a question the forum did not ask.
- Asserting unregistered or common-law rights without documenting their basis. Whatever the applicable standard requires is a legal question; what the expert can supply is the dated evidence of use — archived captures, resolution history, certificate entries — and it should be supplied rather than assumed.
- Analyzing the disputed name in isolation when the registrant holds a patterned set, which is the point at which this analysis hands off to the variant-set work.
What the comparison cannot establish
It does not measure what any actual internet user thought. A side-by-side comparison tells a panel whether a mark is recognizable within a domain name. It is not survey evidence, and where the forum's question is marketplace confusion under national trademark law, the comparison is an input rather than the answer.
- The Overview is a synthesis, not a rule. It binds no panel, and a decision departing from it is not thereby wrong.
- Trademark scope is jurisdictional and dated. A registration establishes rights in the jurisdiction and classes it names, on the dates it names. The disputed domain is global. What that mismatch means legally is for counsel.
- Registration status changes. Lapse, cancellation, opposition and disclaimers all move between the relevant dates, which is why the office record is pulled and dated rather than relied on from memory.
- A purely figurative right may not support the comparison at all, since design elements cannot be reproduced in a domain name.
- Everything after the comparison needs data the comparison does not need. Registrant identity is redacted from public registration output post-GDPR, and archives may simply hold no capture for the relevant date.
- Rendering is environment-dependent, so an exhibit that omits its environment is not reproducible.
Authentication, and what a retaining attorney should expect
US-specific. Trademark office records and registry or registrar records are machine-generated outputs and records of a regularly conducted activity: authentication under FRE 901, with the self-authentication routes at 902(13) and 902(14) available, and hearsay treatment under FRE 803 where relevant. Retrieved web records are captured with method, timestamp and hash following published online-content acquisition practice. The comparison itself is opinion, governed by FRE 702 as amended effective 1 December 2023 and by FRE 703 on the bases of an opinion. Nothing here states how any court would apply those rules.
What to expect from the engagement: the comparison can start immediately, because it needs no disclosure; the office records and the registration record are pulled and dated at the outset; and the report states the test being applied and its source before applying it, then closes with an express statement of what the comparison does not establish.
Bill Hartzer has testified in domain-related legal cases and has provided expert witness reports in others, and has worked in this field since 1996. The current engagement record is maintained at hartzer.com. Which test governs, and what follows from it, are questions for counsel.