What can actually be recovered from an archived website?

When a website has gone and its domain has lapsed, a public web archive such as the Internet Archive's Wayback Machine is often the only record of it left. How much can be brought back from that record depends on two things: what kind of material you want, and how recent a copy you need.

On the first, a web archive records what a visitor's browser was sent, so the visible front of public pages often survives and the machinery behind it does not. On the second, even a well-archived site may have no working copy from its last year or two.

Below is a recovery matrix, asset by asset, grounded in the Internet Archive's own help pages, followed by an original DropOps study of the archive records for 7,613 .uk domains that reached their drop time between 16 August and 30 September 2026. Those domains are not a random sample. Our archive sweep favours names with strong link evidence, so the figures describe well-linked names, not every dropped domain.

The matrix uses four labels. Likely means a typical archived site will have it, provided the page itself was captured. Possible means it depends on whether a crawler fetched that particular file. Often missing means it regularly fails, or survives only as a shell that no longer works. Not recoverable from a public archive alone means the archive never sees it, so it has to come from somewhere else: the old owner, a backup or the original host.

Where the Internet Archive's help pages say something directly, the "Why" column quotes them. Where they do not, the reason given is an explanation of how archiving works, not a measurement.

Pages from different years, stitched together

An archived site is not one snapshot. The Internet Archive says "Not every date for every site archived is 100% complete", and fills the gaps with the closest date it has. A page you are looking at can combine a homepage from one year with a stylesheet or picture from another. If you restore what the replay shows, you may be restoring a combination that never existed.

The live web leaking in

The same help page adds: "In the event that we do not have the link archived at all, the Wayback Machine will look for the link on the live web and grab it if available." When you browse an old site in the Wayback Machine, part of what you see may be today's web. Anything you plan to reuse should be checked against the capture record itself, not only against how the replay looks.

An archive is not a backup

The Internet Archive is direct about this: "Our terms of use do not cover backups for the general public." It adds that you may use the Wayback Machine to find archived versions "of a site to which you own the rights", and that "We can't guarantee that your site has been or will be archived." Its Save Page Now feature saves "a single page, not the whole site". What exists for any particular site is whatever crawls happened to reach it.

To see what is there, the Internet Archive suggests the pattern http://web.archive.org/*/www.yoursite.com/* to list every archived file for a site. The CDX index behind the Wayback Machine returns the same information in a form a script can read. Our technical guide to Wayback restoration walks through it, and how to restore a website from the Wayback Machine covers the practical steps.

One point belongs here, once. Archive presence is not permission. The Internet Archive "does not make guarantees as to the copyright status of items" and asks users to make sure their use "is non-infringing". Whether you may republish recovered words and pictures is a separate question from whether they exist. This article is general information, not legal advice.

The question. For .uk domains that have just dropped, what does the public web archive hold for the homepage? Is there any working capture at all, how far back does it go, how many versions are there, and how old is the most recent working copy by the time the name drops?

The population. 210,484 domains whose scheduled drop time passed while they were still on Nominet's published drop list, between 16 August and 30 September 2026 (UTC). DropOps imports that list daily. "Dropped" here is a list status: the drop time passed with the name still listed. It does not prove the name was released or registered again afterwards.

The dataset. Homepage timelines that DropOps collected from the Wayback Machine's CDX index, plus the recorded answers where the index held no working homepage capture. Archive records were extracted on 7 October 2026.

The sample. 7,613 of those domains, 3.6% of the population, have an archive answer: 7,310 with a homepage history and 303 for which the archive held no working homepage capture.

The suffix mix shows the bias. Compared with the drop list as a whole, the sample has far fewer names registered directly under .uk and proportionally more .org.uk names. The population column below comes from our study of the drop list for 17 August to 30 September 2026 (UK dates, 205,988 domains). It covers almost the same period but is not the identical set of domains; the suffix mix of the 210,484-domain population was not computed separately.

We have not tested why names directly under .uk are under-represented. One plausible factor is age: second-level .uk registrations only opened in June 2014, whereas .co.uk and .org.uk names have had decades to build a site and collect links.

For a made-up domain, the request has this shape: https://web.archive.org/cdx/search/cdx?url=example-domain.test&collapse=digest. Each line of the answer is one capture, and among the index's seven public fields are its timestamp, its HTTP status code and a digest of its content. The CDX documentation describes collapsing as filtering out "captures after the first one that are duplicate" on adjacent lines, and warns that "only adjacent digest are collapsed". A homepage that changes and later changes back therefore has its earlier version counted again. The version figures below count changes the archive recorded, not distinct designs.

Exclusions. The other 202,871 domains on the drop list in this window have no recorded archive answer in DropOps, so they are not in the sample. A domain counts as having no captures only where the archive answered and held no working homepage capture; a lookup that failed is not an answer and is not counted. Inner pages, images, stylesheets and files are not measured at all. For the final-year analysis, 309 of the 7,310 timelines carry no year-by-year status breakdown and are left out, so that table covers 7,001 domains. Every figure comes from one read-only query file run against DropOps' database on 7 October 2026.

In this sample, 7,310 of 7,613 domains (96.0%) had at least one working homepage capture, and 303 (4.0%) had none. That high figure is largely a product of how the sample was chosen. A domain with 75 or more referring domains has usually had a real website for some time, so it is no surprise that the archive saw one. It does not mean 96% of dropped .uk domains are archived.

These are long-lived names. 2,603 domains (35.6%) were first captured as working sites before 2010, including 108 before 2000. Only 214 (2.9%) were first seen in 2025 or 2026. For half of the domains with history, working captures span 11.2 years or more.

A first capture date is a lower bound on a site's age, not a measure of it. A site can run for years before a crawler reaches it, and a first capture may be a holding page rather than the business that came later.

Two-thirds of the domains with history (66.2%) have more than 20 recorded homepage versions; only 206 (2.8%) have a single one. For anyone planning a restore, that is both good and awkward news. There is plenty of material, and there is a real choice to make about which version counts as "the" site. The latest version is not automatically the best one, as the next result shows.

By the time these domains dropped, the archive's most recent working copy of the homepage was typically well over a year old. The median gap was 519 days, about 17 months. 4,650 domains (63.6%) had no working homepage capture in the final year before the drop, and 839 (11.5%) had none in the final five years.

Part of that gap is built into the .uk lifecycle. Under Nominet's current process, in place since 13 September 2022, a domain that is not renewed is suspended 30 days after expiry, from which point Nominet describes it as "Does not resolve", and it drops 95 days after expiry. A website on a lapsing domain therefore stops resolving at least 65 days before the drop time, and no crawler can capture a working page in that window. (Nominet is moving .uk to a standardised registry on 9 February 2027, and the lifecycle changes then.) That floor explains why the "within 3 months" band is small. It does not explain the rest: for most of these domains, the last working copy pre-dates the suspension by many months or years. A site may have been replaced by a holding page, pointed somewhere else, or simply not visited by a crawler again. This study does not tell those cases apart.

Nine domains (0.12%) have a working capture dated after their drop time. We have not examined them individually; a quick re-registration by someone else is one possible explanation.

Looking at the latest calendar year in which the archive captured each homepage at all, 56.8% of domains showed only working captures and a further 25.6% showed working captures mixed with redirects or errors. For 17.6% (1,230 domains), the final year held no working homepage at all. A final year of nothing but redirects usually means the domain was pointing somewhere else by then, which matters when you are trying to work out what the site was in its last useful period.

Put the matrix and the study together and a realistic expectation emerges for a well-linked .uk domain that has just dropped. You will probably find a site: almost every domain in this sample had a working homepage capture, and most had dozens of versions spread over a decade or more. It will probably be old. The newest working copy of the homepage was typically more than a year older than the drop, so a restore brings back the site as the archive last saw it working, with whatever prices, contact details, staff and claims it carried then. And there will be gaps below the homepage, which this study did not measure: inner pages, images and files are captured unevenly, and the "not recoverable" rows of the matrix will not be there at all.

The number of versions also forces a choice. With a median of 33, the last one before the site went quiet is not automatically the right one, and a final year of redirects or a holding page is a reason to look further back.

That age gap is also why a faithful restore is not always the right outcome. If the archive's last good copy is several years old, keeping the useful pages and addresses but bringing the content up to date may serve visitors better than reviving the site exactly as it was. Our comparison of the four options, restore, modernise, repurpose or start fresh, sets out how to choose.

This is how DropOps Sites treats archive history for a domain you own. It reads the old addresses and a sample of old pages from public web archives such as the Internet Archive's Wayback Machine, to learn the site's structure, topics and active years. Scripts, trackers and embeds are stripped from recovered pages, and nothing is hotlinked from the archive. Old details such as phone numbers, prices, opening hours and testimonials are held until the owner confirms, replaces or removes them. A faithful restore is offered only after the owner states that the content is theirs to restore; DropOps records that statement and makes no legal determination.

If you are looking at a name before it drops, our explainer on expired and dropping .uk domains covers how the drop list works and what it does and does not tell you.

• Homepage only. One request per domain, for the homepage URL. Nothing here measures inner pages, images, stylesheets or files, and a well-archived homepage does not imply a well-archived site.
• A biased sample. Domains were prioritised for strong link evidence, with a name-led cohort and newly available names added. The results describe those 7,613 domains, not the 210,484 on the list, and the suffix mix differs from the list's.
• CDX coverage is not completeness. The index lists the captures the archive holds and will serve. It says nothing about pages that were never crawled, were blocked by robots.txt or were excluded at the owner's request, all of which the Internet Archive gives as reasons a site may be missing.
• A 200 is not meaningful content. HTTP 200 means the server returned a page, not that the page was the business's website. Parking pages, holding pages and "coming soon" pages can return 200, so it is possible that some "working" captures, particularly late ones, are parked pages. We did not measure how many.
• Versions are approximate. Collapsing folds only consecutive identical captures, and any change to the captured page, however small, can count as a new version. A version count is a count of recorded changes, not of redesigns.
• A moving index. Archive records were extracted on 7 October 2026. The archive keeps crawling and sometimes removes material, so a re-run could give different figures.
• "Dropped" is a list status. It means the drop time passed while the domain was still on Nominet's list. It is not proof that the name was released or registered again.

For the well-linked .uk domains in this sample, the public web archive usually holds plenty: a working homepage capture for 96.0% of them, with a median of 33 versions over a median span of 11.2 years. What it often lacks is a recent copy. The last working homepage was typically captured about 17 months before the drop, and for nearly two-thirds of domains (63.6%) it was more than a year old.

A restore from the archive is therefore a restore of the past: the visible front of public pages, as they stood some time ago, without the forms, search, logins, databases or traffic history that made the site work. Check what exists before promising anything to anyone, choose the version deliberately, and treat every old fact on a recovered page as something to verify rather than republish.