How ViralWorld documents internet culture

A thing gets a record when people spread it of their own accord and it can be identified as one thing: a video, a photograph, a post, a sound, a phrase, a meme, a trend, or an event that the internet made its own. It needs to be findable under a name or a description, and there needs to be something true to say about it.

It does not need to be a video, and it does not need to be in English. It does need to have spread because people passed it on. Something that was merely published to a large audience — an advertisement nobody shared, a broadcast nobody clipped — is not what this archive is for.

That people chose to pass it on, in numbers, beyond the place it started. It is a statement about how a thing travelled, and a view count does not show that.

A view count says how many times a player started. It cannot tell you whether a video was shared by millions or shown to millions by a recommendation, it is not comparable between platforms, and it can be bought. Plenty of the most-watched videos ever made are children’s songs that nobody ever sent to anybody. So ViralWorld does not rank anything by views, does not publish a “most watched” chart, and where it reports a record — the first video to a billion views — it reports the announcement, with its date and its source, as a piece of history.

Most records begin with something automated noticing that a thing exists: a search, a listing, a reference work that names it. That is discovery, and it proves nothing beyond the fact that the thing was found.

Finding a video under a title does not establish that it is the original, that the account which posted it made it, or that the title is what anybody calls it. Those are separate questions and each needs its own evidence. Until that evidence exists the page says so: “uploaded by” rather than “made by”, “earliest known” rather than “original”.

Four different things are easy to confuse, and the pages keep them apart:

• The original. The first publication, by whoever made it. Claimed only where a source identifies it.
• The earliest known. The oldest copy anybody has pointed to. Often not the original, and labelled as what it is.
• What plays here. The copy this site can show you today. It may be a re-upload made years later, and the page says when it is.
• The version people remember. Frequently none of the above: a crop, a remix, a clip with music somebody added. Where the difference matters the page sets them side by side.

The original is often gone. Accounts are deleted and platforms close. A record is kept anyway, because what a thing was and what it turned into outlasts the file, and a surviving copy is shown as a surviving copy.

A page may carry several dates and they mean different things: when something was made, when it was first posted, when it spread, and when later events happened to it. A home video recorded in 1984, uploaded in 2007 and turned into a meme in 2009 has three dates, and all three are true.

Each is given at the precision it is actually known. If research reached the month, the page says the month and does not invent a day. “This day in internet history” lists only things dated to the day by a source; an event known only to the month never appears on a day.

The date this archive first noticed something is never presented as the date it happened.

In order of weight: the thing itself and what its maker has said about it; reporting from the time it happened; reference works that cite their own sources; and later retrospectives. A page that merely repeats a claim found elsewhere adds nothing to it.

On the researched pages, every dated event, correction and factual claim carries a numbered source you can open. A claim nobody could source is not on the page. The note “sources last read on” gives the day somebody last checked that page’s history against what it cites.

That date is about the history. Whether a video still plays is a different question with a different answer every week: it is checked separately, and a page may correctly say its sources were read last month and its video was removed yesterday.

Software finds candidates, recognises when two uploads are the same thing, checks whether media is still available, and declines things that fail basic tests. It does this continuously and it is why the archive can be large.

Software does not write the histories. The researched sections of a page — the timeline, what people misremember, why a thing mattered — are written by reading sources and are reviewed before they are published. Decisions about real people, about anything sensitive, and about whether something should be public at all are made by a person.

Nothing on these pages is generated to fill space. A section with nothing behind it is left out, which is why some pages are long and most are short.

Some pages carry a correction: a title everybody gets wrong, a year that is really the year of the re-upload, a quote that was never quite said. These are included only where a source supports both halves — that the mistake is common and what is actually the case. Corrections are not invented to make a page more interesting.

Viral media is full of people who did not choose to be in it, and some of them were children. A record can be withheld, kept without media, or published without naming the person in it, and those decisions take priority over completeness. Graphic injury, death and the humiliation of private people are not what this archive collects, however widely they spread.

Nothing is loaded from another platform until you allow it, and the media itself is never copied here. The privacy page says exactly what your browser does and does not send.

Every record that is safe and accurate enough to show has a page, because a short true page is still useful to somebody browsing the archive. Only some of those pages are offered to search engines.

The test is whether the page tells you something you would not get from the source alone: where it came from, what happened to it, what it is related to. A page that is a title, a player and an upload date is a fair entry in a catalogue and a poor answer to a search, so it carries a note asking search engines to leave it out. That is a judgement about the page, not about the thing it describes.

These pages will sometimes be wrong. When one is, tell us: a wrong date, a wrong creator, an account described as the maker of something it only re-posted, a record that should not be public. A person reads every report and records what was decided.

A correction changes the page and the reviewed record behind it together, so a mistake that has been fixed stays fixed. Where sources disagree and it cannot be settled, the page says that they disagree.