The website at a disputed domain name is often the best evidence of bad faith, and the easiest to lose. Recent WIPO decisions show what happens when the screenshots in the annex are the wrong ones, too few, or the only record left, and a sample of 301 cases from 2025 shows how often the Internet Archive has nothing to fall back on.

In December 2024, Meta sent a cease-and-desist letter to the operator of <snap-insta.app>, a site offering an “Instagram Downloader”. After that, the site changed: the Instagram tool disappeared, and a TikTok downloader took its place. When Instagram filed its UDRP complaint, the annex meant to show the website contained, in the panel’s words, “only screen captures of the Respondent’s current website” (Instagram, LLC v. Akash Rakshit, WIPO Case No. D2025-1906).

The historic captures existed. They had been left out by mistake, and the complainant asked to add them in an unsolicited supplemental filing. Panelist Matthew Kennedy admitted the filing, noting that the captures had been “described in detail in the Complaint” and that the respondent had been able to comment on them. The domain name was transferred.

That case ended well, but it depended on a panel’s discretion to accept a late filing. The more useful lesson is about timing: the content that proved bad faith existed only before the letter went out.

What the Overview Actually Asks For

The WIPO Overview 3.1, published in February 2026, is explicit about the kind of evidence a bad-faith claim needs. Section 3.1 says that complaints alleging paragraph 4(b) conduct “should be supported by arguments and available evidence such as dated screenshots of the website to which the disputed domain name resolves or correspondence between the parties.” It adds that “wholly unsupported conclusory allegations may not be sufficient to support a complainant’s case”, even when the respondent defaults.

Section 3.2.1 lists, among the factors panels weigh, “the content of any website to which the domain name directs, including any changes in such content and the timing thereof.” The change itself, and when it happened, can be part of the case. That only helps if someone recorded the before and the after.

Two other sections matter for anyone preparing an annex.

Section 4.8 confirms that a panel may do limited research of its own, which “may include visiting the website linked to the disputed domain name” and “consulting historical resources such as the Internet Archive”. In practice, panels do this. In a 2025 case brought by Fenix International Limited (WIPO Case No. D2025-1852), the panel recorded that, “Having checked the Disputed Domain Name”, there was “no change on the website after filing of the Complaint.”

Section 3.11 deals with respondents who block archived copies: the use of robots.txt or similar tools to prevent access to historical content on archive.org, “particularly where employed only after notice of potential trademark abuse, may support an inference” that the respondent tried to hide incriminating content. An inference is helpful, but a capture of the content itself is better.

The Other Side Can Read the Archive Too

The archive is public. A complainant who picks one convenient snapshot takes a risk: the respondent can produce the rest, and under section 4.8 the panel can look for itself. Euronet Worldwide, Inc. v. Euronet Internet (WIPO Case No. D2024-0293) is the clearest recent example.

Euronet sought <euronet.com>, registered in 1994. Archive screenshots in the record, from 2001 to 2022, showed decades of use for Dutch hosting and telecommunications services under a series of corporate owners. The complaint relied on a 2021 screenshot of a later corporate information page, and the respondent accused the complainant of having “selectively picked from the publicly-available evidence”. Panelist Adam Taylor denied the complaint and found Reverse Domain Name Hijacking. Among the reasons, he listed a lack of candour: the complainant, “which has plainly reviewed the archive history of the website of the disputed domain name, failed to produce any evidence of the former extensive active use”, and “has simply selected a single screenshot showing the later corporate information use of the disputed domain name in 2021, and indeed has omitted the archive screenshots indicating that this usage dated back to at least 2014.”

The lesson cuts both ways. Website evidence is about the history of the site as well as a single moment, and an annex that shows only the helpful moment can be turned against the complainant.

When the Archive Has Nothing

Archives are not complete. Many pages are never crawled, and short-lived sites built to impersonate a brand are exactly the kind that come and go between crawls.

Boston Carriage, Inc. v. Jason Dib / Jason Dibson and Name Redacted (WIPO Case No. D2025-3256) involved two domain names. For the second one, <bostoncarriagema.com>, the panel noted that at the time of the decision, the domain name “does not resolve to an active website, and the Wayback Machine does not have any archived screenshots associated with it.” What remained were the complainant’s own screenshots of the site as it looked in June 2025, a page presenting itself as a partnership of “Boston Carriage, Inc.” with a local charity. Panelist W. Scott Blackmer ordered that domain name transferred. (The complaint about the other domain name, <bostoncarriage.com>, was denied on the history of its use.)

Had the complainant not captured that page while it was live, the record would have shown a dead domain and an empty archive. The same decision shows the other limit of archives. The respondent argued that a period of inactivity is “not proven simply by the fact that the Wayback Machine did not capture any screenshots during that period”. A gap in the archive does not prove that the site was offline.

How Often the Archive Has Nothing

To see how common this is, I drew a random sample of 400 WIPO case numbers between D2025-0001 and D2025-5480, close to the highest 2025 number with a published decision. 354 of them had a published decision, 342 in English, and 301 of those stated both the filing date and the disputed domain name. For each of the 301, I asked the Wayback Machine for the latest capture of the domain name’s homepage made before the complaint was filed.

• For 117 of the 301 domain names (39%), the archive held no successful capture of the homepage from any time before filing.
• For 143 (48%), it held none from the year before filing. For 210 (70%), it held none from the 90 days before filing.
• New registrations fare worst. Of the 133 domain names that, according to the decision, were registered less than a year before the complaint, 62 (47%) had never been successfully captured. Of the 53 registered earlier, 5 (9%) had never been successfully captured.
• In 104 of the 301 cases, the decision records that the domain name did not resolve to an active website when the complaint was filed or decided. For 59 of those (57%), the archive had no successful capture of the homepage before filing either.

These figures are generous to the archive. They count any successful capture of the homepage, whether it shows the infringing content, a parking page or a previous owner’s site, and they ignore subpages. Even so, for about half of the disputed domain names the archive held nothing from the year before the complaint. In those cases, what the site showed exists only if someone captured it.

Whaley Products, Inc. v. Rahul Kumar (WIPO Case No. D2025-0996) shows what the complainant’s own record is worth when a site goes dark. The complainant sent a cease-and-desist letter on October 28, 2024. As of November 6, 2024, the domain name resolved to “a clone of the Complainant’s website” that purported to sell the complainant’s products. By the time the complaint was filed, it did not resolve to an active website. The panel ordered the transfer of the domain name on the basis of the clone, noting that the fact that it “does not currently resolve to an active website does not prevent a finding of bad faith.”

A Change After Notice Is Evidence, if You Can Show It

The snap-insta decision also shows how a post-notice change is treated on the merits. The respondent argued that any resemblance to Instagram “has long since been corrected”. The panel did not accept that the change cured the problem. It found that “while the Respondent has removed references to INSTAGRAM on his website, he still uses the INSTA mark in the disputed domain name itself to attract Internet users and direct them to his site where he displays advertising”, and added that “there is still a risk that downloading may be reauthorized at any time while the site is under the Respondent’s control.”

That reasoning was only available because the record contained both versions of the site, the one before the letter and the one after it. Section 3.2.1 makes “changes in such content and the timing thereof” a relevant factor. To use it, a complainant needs captures from both sides of the letter.

A Practical Checklist for Complainants’ Counsel

None of this requires a particular tool. It requires doing things in the right order.

1. Capture before any contact. The cease-and-desist letter, a purchase inquiry through a broker, and the filing itself can all prompt the registrant to change the site. Capture first, then write.
2. Capture again after each step. After the letter, after any reply, and after filing. If the content changes, the sequence of dated captures is the evidence that section 3.2.1 refers to.
3. Capture the whole page, and the pages users actually reach. Include product or login pages linked from the homepage, and follow redirects. Some sites show different content depending on the visitor’s country or device. If that is the case, record it and say how each capture was made.
4. Record the domain data at the same moment. A WHOIS or RDAP lookup and the DNS records taken at the same time tie the page to the domain name. If the domain is used only for email, the deceptive emails and the domain’s DNS records are the evidence, as section 3.4 of the Overview describes. David Kreider’s July 2026 CircleID post on MX records as bad-faith evidence covers that in depth.
5. Make every capture checkable. Each annex should show the full URL and the date and time of capture. Keep the original files, not only the pasted images. A record of how the capture was made (who, when, which method) answers authenticity questions before they are asked. The Overview 3.1 now notes, in section 4.2, that the credibility of evidence created with AI tools may depend on disclosing how it was made. Panels will be more alert to manipulated images, not less.
6. Ask a neutral third party to keep a copy. The Internet Archive’s Save Page Now creates an independent public copy of a page at no cost. It is not always possible (some sites block it), and it does not replace your own capture, but it is a copy the respondent cannot say you made.
7. Review the full archive history before filing, and disclose what is relevant. If the panel can find it, so can the respondent, and Euronet shows what an incomplete picture can cost.

Conclusion

UDRP is a paper proceeding. Panels decide on what is in the annex, plus what they can see for themselves on the live site and in the archive. When the decision is written, the live site may have changed, and the archive may never have captured it. The complainant’s own dated captures, taken before the registrant knew a dispute was coming, are often the only record of what the site showed.

Method: case numbers D2025-0001 to D2025-5480 were sampled at random (400 numbers, seed 20251112). Decision texts were taken from the PDFs on wipo.int, and archive data from the Wayback Machine CDX API in November 2026, limited to the homepage of each domain name and to the first three domain names in multi-domain cases (369 domain names in total; the figures above use the first domain name in each case). The case list, the measurements and the method are published at https://getproofsnap.com/posts/udrp-wayback-machine-coverage-study-2026.html (CSV, CC BY 4.0).