Web2Disk Review: Download Websites for Offline Browsing

Inspyder still takes money for Web2Disk. Read on 13 August 2026, the order page prices a full licence at $39.95 as a one-time cost, with a free trial beside it, over HTTPS. Then you go looking for a date, and there is none. Not against the current version, not against any of the seventy-odd versions listed before it. The reference manual the vendor still serves is copyright 2015, printed in October of that year, and its requirements line reads "Web2Disk requires Windows XP/Vista/7 or Windows Server 2003/2008."

So the question is not whether Web2Disk works. It does, and on the right kind of site it works well. The question is what a link-following downloader built for the static web still captures in 2026, what it drops without telling you, and what the same $39.95 buys elsewhere. Everything below was read from vendor pages, manuals, changelogs and release feeds on 13 August 2026.

Still on sale, standing very still

The Web2Disk product page is live and taking orders. System requirements are Windows XP SP2 or higher, 1 GB of RAM and 1 GB of free disk. The bullet list you get for the money includes up to 25 crawler threads, a command-line interface, scheduled downloads, a login wizard, no page limits, free minor updates, and the ability to distribute the result on CD or USB drive. The CD is not a typo. It is the period detail that tells you when the feature list was written.

Then the revision history. Roughly seventy entries, from 1.0.0 to 5.0.1, and not one carries a date. The newest reads, in full: TLS 1.1 and 1.2 are now supported if .NET 4.5 or later is installed; the crawler auto-detects when the Windows-1252 codepage is mislabelled as ISO-8859-1; "Minor internal tweaks for Windows 10"; soft 404 detection is now case insensitive; JavaScript errors can be hidden on the Login Wizard.

Two of those five lines date the entry without meaning to. TLS 1.2 was the scramble of 2016. And Windows 10 went out of support in October 2025, 22H2 included, yet it is still the newest operating system the changelog mentions.

The build that ships is newer than the changelog admits. Download mirrors carry 5.1.4.11487 at about 18 MB, and one lists 5.1.5. No first-party page numbers it, dates it, or says what changed. Web2Disk 5.1.x exists, and nobody outside Inspyder knows when it was made.

None of this is the loud kind of dead. No archived repository, no farewell post, no parking page. The order form works and the download is there. What the evidence supports is more awkward than death: a product sold in 2026, documented in 2015, by a vendor that has never published a release date for anything. The sibling product tells the same story, down to the same 5.0.1 entry and the same October 2015 manual, in our Inspyder Power Search review.

What $39.95 actually gets you

Point it at a start URL, set a depth or page-count limit, and it walks the links, pulls HTML, images, CSS and other assets, and rewrites references so the folder browses locally. The main window keeps the common settings in front of you and hides the rest, which is the whole reason to pay for this instead of learning a command line.

Four things are worth knowing first, all of them from the manual.

• The scheduler is Windows Task Scheduler. Web2Disk does not need to stay open, but the machine has to be awake. A laptop that sleeps at 18:00 does not run a 02:00 re-crawl.
• The login wizard is a real mini-browser. You log in by hand and the credentials are captured for the crawl. Whether it copes with a single-page login form built in 2026 is another question, and probably not.
• The command line is more complete than the GUI suggests. The manual documents -project=, -rooturl=, -savepath=, -filelimit=, -depthlimit=, -email=, -checkchanges and -rename. Enough to drive it from a batch file, and -checkchanges is the mechanism behind the change-monitoring feature the marketing leads with.
• JavaScript handling is link-scraping, not rendering. The option looks for URLs inside script text so the crawler can follow them, with a more aggressive heuristic mode alongside it. Nothing executes. No browser engine, no DOM after hydration, no waiting for a fetch to land.

What a link-following mirror drops without telling you

Web2Disk, HTTrack and wget all work the same way: fetch HTML, parse it, follow what is in it. HTTP Archive's 2024 Web Almanac put the median mobile page at 558 KB of JavaScript and the median desktop page at 613 KB, the mobile figure up 14% year on year. A parser that reads markup and never runs any of that is looking at a smaller page than you are.

The failure is rarely total. Partial is worse: the copy opens and looks roughly right.

Lazy-loaded images. A JavaScript lazy-loader puts a placeholder in src and the real URL in data-src. The crawler grabs the placeholder. You get a folder of grey rectangles and a page that looks fine until you scroll.

Responsive images. srcset and <picture> list several candidate files. A downloader that reads only src takes one and leaves the rest, so the copy breaks on a screen other than the one you tested on.

Assets on a second origin. Fonts, icon sets and bundled JavaScript usually sit on a CDN hostname. Unless you deliberately allow that host, the mirror stops at your domain boundary and the copy renders in Times New Roman.

Video, and anything that arrives after load. An HLS stream is a playlist plus a few thousand segment files a player requests at runtime. Infinite-scroll feeds and filtered listings are assembled from API calls. Point a traditional downloader at a single-page app and you get the shell.

And then the part that surprises people who did everything right. A perfect capture of a modern site still will not run from a folder. Double-click the saved index.html and the browser gives the page a file:// origin, where module scripts are blocked. Mozilla's documentation is blunt: "if you try to load the HTML file locally (i.e., with a file:// URL), you'll run into CORS errors due to JavaScript module security requirements. You need to do your testing through a server." Service workers need a secure context too. The modern way to read an offline copy is to serve the folder over http://localhost, one command that Web2Disk's burn-it-to-a-CD world never anticipated.

The folder is not the site.

Where the copy breaks before the site does

Path length. A mirror turns URLs into directories, and Windows caps a path at 260 characters. Microsoft is specific that the registry switch alone is not enough: the application also has to declare longPathAware in its manifest before the limit lifts. Inspyder documents neither, so assume 260. Deep category trees with long slugs reach it, and the result is a partial copy with no obvious error.

Query strings. Faceted navigation is the classic trap. A listing page with three filters of ten values each is a thousand URLs of the same page, all one hop apart. Depth limits do not help, because the trap is wide rather than deep. The defence is the pattern filter and the page-count cap, so the settings people skip are the ones that matter. Session IDs in URLs do it smaller: every page is new again on the second visit.

Disk. Assume 2 MB per page with assets, conservative for anything media-heavy. A 20,000-page mirror is about 40 GB before the near-duplicates. That is arithmetic from a stated assumption, not a measurement.

Politeness. Twenty-five parallel connections against shared hosting is a small denial-of-service attack with your name on the IP. Turn the thread count down before you point the crawl at production, and start from the sitemap if the site publishes one.

Repetition. One mirror is an afternoon. The same four sites every week, landing somewhere other than a folder on one Windows desktop, is a pipeline. The question stops being which downloader and becomes where the output lives, and a scheduled desktop task and a recurring data feed solve different problems.

The alternatives, checked on 13 August 2026

HTTrack is the free like-for-like replacement, and it is less dead than the roundups claim. The Windows stable installer really is 3.49.2 from 20 May 2017, which is where most reviewers stop reading. The same page also serves a Windows beta, 3.50-beta-3, dated August 2026, and a Linux source tarball, httrack-3.49.20.tar.gz, from the same month. The full picture is in our free website backup guide: the binary most people download is nine years old, and the source line is still moving.

GNU wget is the scriptable workhorse, at 1.25.0 since 10 November 2024. One command mirrors a site and rewrites links for local viewing:

wget --mirror \
     --convert-links \
     --adjust-extension \
     --page-requisites \
     --no-parent \
     --wait=1 --random-wait \
     https://example.com/

Two things about that recipe matter, and the earlier version of this article said neither. First, --mirror is documented as equivalent to -r -N -l inf --no-remove-listing, so recursion is already infinite and --no-parent plus the host restriction is all that bounds it. Second, and this is the one that bites: --convert-links only rewrites links to files wget actually downloaded. Anything it skipped is rewritten to "its full Internet address rather than presenting a broken link", in the manual's words. Your offline copy quietly keeps one foot on the live web, and you will not notice while you still have a connection. Conversion also happens only at the end, because "only at the end of the download can Wget know which links have been downloaded". Interrupt a long crawl and nothing is converted. wget2 (2.2.1, 30 December 2025) is a separate program, not a drop-in upgrade.

Cyotek WebCopy is the correction this article owed you. The previous version called it a modern GUI copier. Its current release is 1.9.1.872 from August 2023, it wants .NET Framework 4.6 and lists Windows 7 through 10, and Cyotek's own page states the limitation without flinching: "WebCopy does not include a virtual DOM or any form of JavaScript parsing. If a website makes heavy use of JavaScript to operate, it is unlikely WebCopy will be able to make a true copy." Free, pleasant, three years since a release, and honest about what it cannot do.

SiteSucker is the price comparison nobody makes. On the Mac App Store it is $5.99, version 6.2, macOS 12 or later, and the listing showed an update the day before this review was written. A separately sold Pro edition adds embedded video and Tor support; the product page publishes no price for it, so we cannot quote one. Against $39.95 for a Windows binary of unknown vintage, $5.99 for a maintained one is the sharpest number here.

If you are going to pay anyway

Two paid competitors sit either side of Web2Disk, and both change the arithmetic.

A1 Website Download from Microsys undercuts it: a free tier and a $39 professional licence, version 12.0.1, on Windows and macOS. The vendor publishes no date for that version, which puts it in Web2Disk's evidential bracket.

MetaProducts Offline Explorer is the one that matters. The standard edition is $101.95 and Pro is $252.95, and alone on this page it can use a Chromium engine as its internal browser. It is also demonstrably maintained: the vendor's news page dates 8.8 SR1 to 16 February 2026 across the whole product line. Two and a half times the price of Web2Disk, and the difference is exactly the capability that decides whether a modern site copies at all.

That is the shape of the paid market in 2026. Cheap and static, or expensive and rendering. Web2Disk is cheap and static, at a price set when cheap and static was the only option there was.

Two formats built for this job

Mirroring rewrites your copy so it browses locally, which means it is no longer what the server sent. For anything you might have to defend later, that matters.

WARC stores the HTTP transactions themselves, request and response headers included, in the format standardised as ISO 28500 and maintained as WARC 1.1 by the International Internet Preservation Consortium. It is what the large web archives run on. wget writes one directly, and Browsertrix Crawler (1.14.1, 30 July 2026) drives a real browser and packages the result, so dynamic pages survive the trip. Self-hosting is free; Webrecorder's hosted Browsertrix starts at $30 a month for 180 execution minutes, 100 GB of storage and 2,000 pages per crawl. Reading a WARC back used to be the hard part, and ReplayWeb.page now replays one entirely in the browser with no server behind it.

ZIM is the other answer, and almost no comparison of website downloaders mentions it. zimit (3.1.2, 3 February 2026) runs Browsertrix Crawler and warc2zim inside Docker and hands back a single ZIM file that opens in Kiwix. One file, no folder, no local web server, offline reading on a phone. It is the format the offline editions of Wikipedia ship in. We have not benchmarked zimit against a large commercial site, so treat it as the right shape of tool rather than a guaranteed result.

ArchiveBox sits between the two: a self-hosted archive that saves each URL several ways at once. Stable release 0.7.4, 18 May 2026, updating the container's Chromium to v147 through Playwright 1.59.0, while the 0.9 line stays in release candidates.

SingleFile remains the right answer for one page: the rendered DOM and every asset bundled into one self-contained HTML file, executed JavaScript included. Version 1.22.98 on Firefox, February 2026, 4.8 stars from over a thousand reviews.

For a whole dynamic site the two routes are unchanged. Save it with a real browser, using Browsertrix or an automated headless browser, which runs the JavaScript first and captures the result. Or go after the data instead of the pages: call the JSON endpoint the page already uses, or run a proper web scraper that renders and extracts structured records. Mirroring gives you files. Scraping gives you rows you can query.

Downloading a site responsibly

Whatever you pick, the target's server is somebody's bill. Aggressive crawling also triggers anti-scraping protection, so restraint is self-interest as much as manners.

• Throttle. A delay between requests, and fewer parallel connections than the tool allows.
• Bound the scope. Depth limits, --no-parent and pattern filters, so one stray link does not become the rest of the internet.
• Read robots.txt, and know what it is. A standard since RFC 9309 in September 2022, and that document is explicit that its rules "are not a form of access authorization". Ignoring it is a choice you make knowingly, not a technicality.
• A mirror is a reproduction. A personal offline copy and a republished one are different acts under copyright. If the pages carry user profiles, comments or contact details, you have also made a copy of personal data, with everything GDPR attaches to that.
• Mind logins. The form wizard reaches gated content. Point it only at accounts you are authorised to use.

Who should still pay $39.95

A narrow group, and it is real. You are on Windows, the site you need is static HTML, you want scheduling and change alerts in one paid box, and you would rather pay once than learn a command line. Web2Disk does that job, and the GUI is genuinely the least demanding of any tool here.

Everyone else has a better option, and which one depends on a test that takes a minute. Open a page on the target site, view source, and search for text you can see on screen. If it is there, HTTrack or wget will mirror the site for nothing, and SiteSucker will do it on a Mac for $5.99. If it is not there, no link-following downloader will find it and no setting will change that: you want Browsertrix, zimit, SingleFile, or Offline Explorer's Chromium engine if a Windows GUI is non-negotiable. And if you need the data behind the site rather than a browsable copy of it, mirroring is the wrong tool, and a managed extraction service returns structured records instead of a folder of rewritten HTML.

We have no relationship with Inspyder Software and bought nothing to write this. Every price, version and date above came from a vendor page, manual, changelog or release feed opened on 13 August 2026. Where a source publishes no date, this review says so rather than guessing.

View source first. That answer decides the rest.