How to See a Competitor's Old Website: Web Archives, Ranked

A web archive stores what a crawler received when it asked for an address on a particular day: the HTML, and whatever assets it managed to fetch alongside it. The largest of them has been doing this since 1996 and passed one trillion web pages in October 2025, marking the milestone publicly on 22 October that year. For competitive work that turns a competitor’s website from a description of the present into a dated series, which is the only way to see a decision rather than a result.

What an archive does not contain is a record. It holds captures, taken irregularly, weighted toward addresses that were well linked, missing anything behind a login or a form, and frequently unable to reproduce a page that assembles itself in the browser. Every useful thing on this page follows from holding those two facts at once: the evidence is unusually strong, and the coverage is unusually uneven.

1. 1Start from the exact address, not the home page. Archives index addresses rather than sites, so the coverage of a pricing page and of the home page it links from can be completely different. Paste the deep link to the page whose history you want. If it returns nothing, try the address it used to have, since a redesign usually moved it.
2. 2Read the calendar before reading any snapshot. The calendar shows how densely the address was captured, and that shape is itself information. Dense years mean the page was popular or heavily linked, and a gap means the archive simply was not there, not that the page went unchanged. Pick your comparison dates from where the captures actually are.
3. 3Compare two captures rather than reading one. A single old page is a curiosity. The value is the difference, and the archive will compute it for you: choose two dates and it marks what was added and what was removed. Start with roughly a year apart, then narrow around whichever interval shows the change you care about.
4. 4Check whether the page actually replayed. Archived pages frequently render incompletely, and an empty pricing table is far more often a broken replay than a removed price. View the source of the snapshot as well as the rendered version, because the text is usually still in the file even when the layout collapses.
5. 5Try a second archive when the first one fails. The archives crawl differently and hold different things. A request-based archive that renders with a full browser often has the page a crawler-based one could not capture, a national library may hold a local site the global archives skipped, and a curated institutional collection can be searched by what a page said rather than by its address.
6. 6Capture today what you will want to cite next year. The one part of this source you control is the future. Any page you may later need to prove said something can be saved on request in seconds, which puts a public dated copy at a permanent address. Do this for a competitor's pricing, claims and positioning pages the moment you first read them.
7. 7Record the snapshot address, not a screenshot of it. The archived address contains its own timestamp and is the whole reason this source is different from every other one here. Paste that address into whatever you write, so anybody who doubts the finding can open the same page rather than take your word for what a picture shows.

Almost everything a competitive team collects has the same weakness: it is an assertion by the person who collected it. A screenshot in a slide is a picture you made. A note saying their pricing changed in March is your memory. Both are fine internally and both collapse the moment somebody with an interest in the answer decides to push back, which is exactly when this work matters most.

Archives are different in one structural way. A capture lives at a public address that contains its own timestamp, hosted by an organisation with no stake in your argument, and anybody who doubts you can open it themselves in five seconds. That is provenance rather than assertion, and nothing else in this cluster produces it. Archived pages have been accepted as evidence in courts and by regulators in several jurisdictions, usually alongside testimony about how the archive works, which is a far higher bar than a competitive brief needs to clear.

This matters most in three places: an advertising or comparison claim a rival has since quietly revised, a capability they used to advertise and no longer do, and a price that moved. In all three the argument is about what was said, not about what is true, and a dated third-party copy settles it in a way nothing else will.

Capture frequency broadly follows how prominent and how heavily linked an address is. A large company’s home page can be captured many times a day; a mid-sized competitor’s feature page might be captured twice a year; a page nobody ever linked to may have no captures at all. Nothing about that is a judgement on the page’s importance to you.

The consequence is worth writing on the wall, because it is the error that turns this source from useful into misleading. An archive can show you that a page said something on a given date. It can never show you that the page did not say something else in between two captures. A claim added in April and removed in June, with captures in March and July, leaves no trace whatsoever, and the archive will look complete while hiding it.

Two habits follow. First, always report the interval rather than the date: the pricing page changed between 12 March and 8 July, not on 8 July. Second, read the calendar as data in its own right. A competitor whose site is captured densely has a lot of inbound links and a lot of attention, and one whose captures are sparse does not, which is a rough but free reading of web presence that you got on the way to something else.

Archived pages frequently render badly, and people conclude the content is gone. It usually is not. A capture stores the files the crawler could fetch, and a modern page needs considerably more than that, so the failure is almost always in the reassembly rather than in the archive.

One more failure has no workaround and needs stating: a crawler that arrived during an experiment was served one arm of it, and nothing in the capture tells you that. So a single snapshot of a heavily tested page, a home page or a pricing page above all, is one variant that existed rather than the page everybody saw. Where the conclusion matters, check two or three captures from the same period before treating any of them as canonical.

1. The Wayback Machine calendar

The default and the deepest. Paste the full address of the page you want, not the domain, and read the calendar before opening anything. The density of the captures tells you how precisely you can date any change, which is the ceiling on every conclusion you will draw. If the address returns nothing, try the address the page had before their last redesign, since a redesign usually moved it rather than deleting it.

2. The Wayback Machine changes comparison

The most under-used feature in this whole area. From the calendar view, open the changes option, select two captures, and it renders them side by side with the added and removed text marked. It converts an hour of squinting at two tabs into a minute, and it is particularly good on pages where the change is a single clause: a dropped guarantee, a removed integration, a softened claim. Start about a year apart to locate the interval, then narrow.

3. Save Page Now

The only part of this source you control, and the reason to bookmark it. Paste an address and the archive creates a capture immediately, at a permanent public location with today’s timestamp. Make it a habit for competitor pricing pages, homepage claims and any comparison page that names you, because those are precisely the pages that get quietly revised and the ones you will later wish you could prove.

4. archive.today

A request-based archive that renders each page with a full browser and stores the visual result. That makes it markedly better than a crawler on script-heavy pages and on anything where what you need is how the page looked rather than what was served. Its coverage is entirely dependent on somebody having asked for that page, so it is thin as history and strong as a second attempt when the first archive returns a broken shell.

5. Search engine caches

Gone, and worth knowing so you stop looking. The largest search engine removed its cached-page links in early 2024 and retired the cache operator shortly afterwards, with its search liaison explaining that the feature had been built for an era when pages routinely failed to load. In September 2024 it added a link to the Internet Archive inside its about-this-page panel, which is now the nearest equivalent workflow. The practical effect is that one archive carries far more of this job than it did two years ago, which is the argument for having a second.

6. The site’s own surviving URLs and redirects

Before reaching for an archive, try the old address on the live site. Companies routinely leave retired pages up because nobody removed them, and where they did tidy up, the redirect tells you which page replaced which. That mapping is genuinely useful on its own: a set of feature pages all redirecting to one platform page is a consolidation, and it dates a repositioning without any archive involved.

7. Their changelog, release notes and press archive

The company’s own history, maintained by them, usually more complete and always more precise than a crawl of it. A changelog gives you dated capability changes without any inference, and a press archive gives you the announcements in their own framing. Read these first for anything product-related, and use the archive for the pages they would rather you did not compare.

8. App store version history

For any competitor with a mobile product, the store holds a dated list of every release with its notes. That is a product history nobody had to crawl, it goes back years, and the cadence alone tells you how much engineering attention the mobile product actually gets. Release notes are also written for users rather than for search engines, which makes them unusually direct.

9. National library web archives

Several countries archive their own domain under legal deposit rules, and the coverage of local sites is frequently deeper than the global archives manage. This is the right place to look for a competitor whose main site is on a country domain, or for a regional business the global crawlers only visited occasionally. Access rules vary, and some collections are readable only from a library reading room, which is a real constraint rather than a formality.

10. Curated institutional collections

Libraries, universities and public bodies build themed web collections, and the largest of these programmes exposes its holdings publicly across well over a thousand collecting organisations. The reason to care is the search: these collections can be queried by the full text of the captured pages rather than only by address, which is the one thing the ordinary archive lookup cannot do. If you need to find who was saying a particular phrase in 2018 rather than what one address said, this is where to start.

11. Their public repository history

Where a company publishes its website or documentation in the open, the actual edit history is available with authors, dates and the exact wording before and after. That is better than any reconstruction, because it records the change rather than sampling the result. Common for developer-facing companies and for documentation sites generally, so check it first when the competitor sells to engineers.

• Reading and saving are both free, with no account. The largest archive is run by a non-profit library funded by donations rather than by subscriptions, which is why nothing here has a paywall and also why nobody owes you an uptime commitment on the day you need it.
• Robots files are no longer treated as an archiving instruction. The archive said publicly in 2017 that rules written for search engines work badly for archives, having already stopped honouring them on government and military sites some months before. Removal is handled as a request to the archive instead, so a site currently blocking crawlers may still have years of history in it.
• Nothing behind a login or a form is in there. Product interfaces, gated content, customer portals and anything requiring a search query were never reachable by a crawler. This is the single largest gap and it does not close.
• Coverage is by address, not by site. A domain being well captured tells you nothing about the specific page you want. Always start from the deep link, and check the calendar for that address rather than assuming the site’s general coverage applies.
• Work at reading pace, one page at a time. These are free public services run on donations. Open the captures you need, save the ones you care about, and treat the archive the way you would treat a library rather than a database.

The material was published to the public web deliberately, and an archive is a library holding a copy of it. Reading that copy is unremarkable. The care needed here is almost entirely about how you present what you found, because an old page is a true statement about the past and a false one about the present.

• Never present a withdrawn claim as a current one. A competitor who advertised something in 2023 and stopped is not making that claim today. Repeating it to a market as though they were is misleading comparative advertising, and the fact that you sourced it honestly does not help you.
• Date every quotation at the point of use. Put the capture date in the sentence, not in a footnote. A battlecard line reading “as of March 2024 their pricing page stated” is usable indefinitely; the same line without a date becomes wrong the moment they change it and nobody notices.
• Copyright survives archiving. An archived page carries the same rights as the original. Short extracts with attribution are ordinary practice; reproducing a competitor’s page wholesale into your own materials is not, and no research purpose changes that.
• Old pages can contain personal data. Team pages, testimonials and author bylines name people who may have left, and privacy rules attach to that information regardless of where you found it. Keep what has a business purpose in a professional context, and drop the rest.
• Do not go looking for what was never meant to be public. Archives occasionally hold a page that was published by mistake and removed quickly. Deliberately hunting for material a company took down for confidentiality reasons is a different activity from researching their public positioning, and it is the version that gets a company into trouble.

• Anything behind a login. The product itself, the customer portal and gated content were never reachable by a crawler. Proxy: their documentation and release notes, which describe the product publicly and are dated.
• What happened between two captures. A change that was reversed leaves no trace at all. Proxy: save the page yourself on a fixed rhythm, which converts a sample into something closer to a series.
• Why anything changed. The archive records wording, never reasoning. Proxy: what else moved in the same fortnight, since a pricing change alongside a new tier and three sales adverts is a strategy rather than an edit.
• Which version real visitors were shown. Experiments, personalisation and geography all mean the crawler saw one page among several. Proxy: several captures from the same period, and your own visits from different regions.
• Small competitors in any depth. Capture density follows links, so the private rival you most want to track is often the one with four captures a year. Proxy: their own changelog and press archive, plus saving their key pages yourself from today onward.

The rhythm here is backwards compared with every other source in this cluster. There is no point checking the archive monthly, because it will not have added much and nothing you already read will have changed. What is worth doing monthly is the opposite: save the five or six pages per competitor that decide arguments, their pricing page, their comparison page, their homepage and their customer list, so that a year from now you own a dense series rather than whatever a crawler happened to take.

Do the actual reading quarterly, and only against a question. Comparing two captures a quarter apart with no hypothesis produces a list of cosmetic edits. Comparing them because you want to know whether their entry tier moved, or whether a named integration disappeared, produces an answer in ten minutes. Record findings as intervals with both timestamps, and file them where the rest of the price history lives, which for most teams is a pricing teardown rather than a separate archive document.

The structural problem with this source is that it is somebody else’s crawler on somebody else’s schedule. A competitor changes their entry price on a Tuesday; the next capture lands in nine weeks; you notice in the following quarter’s review, by which point three deals have been lost to a number nobody on your side knew about. The archive did its job perfectly and still told you months late, because it was never designed to alert anybody. It is a library, and libraries are for afterwards.

Watching a handful of pages across several competitors and knowing the day they change is a different function entirely, and it is what competitive intelligence software exists to do. Flares keeps its own dated record of what moved on the surfaces a competitor controls, so the history belongs to you and arrives as it happens rather than whenever a crawler next passes. What that does not replace is the archive itself. No monitoring product can show you a page from 2019, and the third-party timestamp that makes an archived capture citable to somebody who does not trust you is something only a neutral archive can provide.