Reviews
Part of Memes and virality: methods, tools and useful context
Memes and virality updates 2027: facts and context
Tracing memes and virality back: what a search can and cannot see, what each kind of evidence carries, and how to write an origin sentence you can defend.
Ask where an online joke format came from and you will usually get a confident answer. The confidence is the problem. Nearly every origin claim about internet culture is a reconstruction assembled after the fact, out of whatever copies happened to survive, by someone who had no way of seeing the copies that did not.
This page is about the craft of tracing an item back: what evidence actually exists, what each kind of evidence can carry, and how to write a sentence about origin that will not fall over.
What to take away
- Reverse image search matches pixels, not lineage.
- A clean story is easier to repeat than a hedged one.
- Name what you found and where, rather than what you believe: the earliest located copy, its host, and its date.
- A screenshot is the weakest thing anyone can offer as proof.
The gap between "the earliest I found" and "the first"
Those are two different statements and only one of them is a finding. Search indexes hold what was crawled, from hosts that were reachable, in formats a crawler could read. A thing can be years older than its oldest indexed copy and leave no trace of the gap. When you write "the first", you are claiming knowledge of an absence, which is the hardest claim in research and almost never supported.
Keep the two apart in your notes and they will stay apart in your prose.
What a search can and cannot see
Reverse image search matches pixels, not lineage. It surfaces copies ranked by whatever the index holds, and every generation of re-encoding, cropping, added text and screenshot-of-a-screenshot degrades the match. Two copies of the same thing can look unrelated to the matcher.
Text search finds a caption only if the caption was typed as text somewhere. Words baked into an image are invisible until a human transcribes them, and a phrase that circulated only as pixels can look, to a search engine, as though it never existed.
Searching inside a platform is often worse than searching it from outside. Many services rank recent material and quietly stop returning old material, so the platform's own search will tell you a thing is new when its own servers are still holding the old copy.
What each kind of evidence can carry
| Evidence you have | What it supports | What it does not support |
|---|---|---|
| A dated archive capture of a page | The item existed at that address by that date | That it existed nowhere earlier |
| A sequential post, thread or comment ID | Relative order within that one site | Order across sites, or a wall-clock date if the site ever renumbered |
| File metadata such as creation or camera fields | A candidate creation date, if the file is untouched | Anything at all once the file has passed through a service that strips or rewrites metadata |
| An in-thread reference to the format as already known | That the format was legible to that audience by then | Any date for the format's first use |
| A copy in another language or script | That it crossed a language boundary | Which direction it crossed in |
| Press coverage of the format | That it reached a wider audience by then | The origin, since coverage usually inherits the same contested claim |
What the middle column amounts to is provenance: a documented chain from the item to where it was found, with the gaps marked. The middle column is the one to write from. The right-hand column is the one that gets people into arguments.
Why a wrong origin outlives the correction
A clean story is easier to repeat than a hedged one. "It started on this site in this year" survives being retold at a party; "the oldest surviving copy is in a thread that refers to something older" does not. The tidy version therefore wins on transmission, regardless of whether it is true.
Corrections also travel through a much smaller network than the claim. The claim spreads on the strength of the format itself, carried by everyone who finds the format funny. The correction spreads only among the far smaller group who care about provenance.
The same failure runs through dated claims about services, and the discipline that prevents it is set out under platform timelines. Then there is the citation loop. A reference page cites an article, and the next article cites the reference page. Repetition starts to look like corroboration. Before you count two sources, check whether they are actually independent or the same source arriving twice by different routes.
Finally, whoever writes something down first tends to set the frame permanently. An early write-up becomes the standard answer not because it was well sourced but because it was indexed early and everyone else found it.
Writing an origin sentence you can defend
- Name what you found and where, rather than what you believe: the earliest located copy, its host, and its date.
- Say what you searched and what you could not search. Closed forums, chat logs, private groups and languages you do not read are all part of the result, and why so much of that material was never reachable in the first place is set out under internet communities.
- Keep the hedge inside the sentence. A qualification in the sentence survives being quoted; a qualification in a footnote does not.
- Date the check itself. Provenance decays in both directions: new archives get indexed, dead sites come back, and someone eventually uploads a folder of old screenshots.
- Separate three questions that get merged: earliest known use, earliest use in the form people now recognize, and earliest use in front of a large audience. They usually have three different answers.
Handling the artifacts people send you
A screenshot is the weakest thing anyone can offer as proof. It carries no verifiable timestamp of its own, it can be edited without skill, and it is trivially fabricated. Treat a screenshot as a lead pointing at a record you should now try to find, never as the record.
A personal recollection is a primary source about one person's vantage point, which is a real thing to have. It is stronger than people assume about sequence and weaker than people assume about dates. Someone may reliably remember that one thing came after another while being years out on when either happened. Ask for order, not for years.
An unfindable link is still information. Record the address, the date you tried, and what you saw instead. A dead reference that keeps reappearing in other people's accounts is itself a clue about how the story spread.
Knowing when to stop
Not every question can be answered by the surviving record, and an investigation that ends in "unresolved, and here is exactly what is missing" is a result rather than a failure. A documented gap lets the next person start where you stopped. A confident guess makes them start over, or worse, keeps them from starting at all.
If you are working through the wider memes and virality material, treat this as the method page and the internet culture archives material as the description of what the record physically contains.
Common questions
Is the earliest archived copy good enough to call an origin?
No, but it is good enough to publish, provided you call it what it is. "Earliest capture located" is a checkable statement about your search. "Origin" is a statement about the world.
Two sources disagree on the year. Which do I use?
Neither, until you know what each one is dating. One may be dating a file, the other a news mention, and both may be right about different events. Disagreement about a date is usually disagreement about which event is being dated.
Someone insists they were there. Does that settle it?
It is evidence and it is worth recording, but a participant saw one vantage point in one community. Being present tells you what reached that person, not what existed.
