Pressfront / Trace / How fast does a rumour reach the news?
Explainer · Updated September 2026
The honest answer is that we have one clean measurement of that specific gap, not a benchmark. On the story where we could measure it, it took 15 hours from the earliest public post we could verify to the first fact-check article in our corpus. One story is not a distribution, so this page shows you the numbers we have and stops there.
On one traced story, 14.9 hours derived from the earliest public appearance we could verify to the first fact-check article in the same corpus, and 31.0 hours to the Snopes entry derived. That is a single data point.
We also measured the gap from earliest public post to biggest amplifier on three traces, which is a related but different thing: 130.2 hours, 56.4 hours and 56 days derived. Three numbers spanning from nine hours to eight weeks is not a median, it is a range so wide that quoting an average from it would be dishonest. So we are not quoting one.
Every lead time we have actually measured
Four traces, run 2026-09-15 and 2026-09-16Every figure on this page is computed at build time from the JSON the tracer wrote. If a trace file changes, this page changes with it.
The private-account rule. We never log in, follow anyone, request access to a locked account, or use scraped private data. A private account is invisible to us and stays that way. So what you see below is the earliest public appearance we could verify, never a proven origin. If a story started in a group chat, a DM or a locked profile, the first public post is the node we can see, and we say so rather than dressing it up as the source.
Run one yourself: Trace a story. How the sweep and the confidence number are built: the methodology card.
You wanted "rumours reach the news in about X hours". Every content marketer in this category will hand you that number. We have one measurement of it.
And the three fuse lengths we did measure argue against averaging anyway. A satire share hit its biggest amplifier in 130.2 hours. An earnings post took 56.4. A pseudo-scientific claim sat for 56 days and then moved in four. Those are not noisy measurements of one underlying speed; they are three different phenomena. A mean over them would describe none of them.
This is the part most tools in this space get wrong on purpose, because a benchmark number is more quotable than a sample size. It bugs me. The benchmark gets repeated for years and nobody ever goes back to check what it was computed from.
Read as shapes rather than as an average, they are useful.
The practical read: the useful question is almost never "how fast do rumours travel". It is "how fast is this one travelling right now", and that one has an answer you can measure.
Every trace writes its lead times into the same graph. When there are enough of them, the distribution of earliest-public-post to mainstream pickup becomes a real research output, with the sample size, the window, and the platform mix printed next to it, the same way our AI visibility study reports its own.
Until then this page says four and one, because that is what four and one looks like.
On the one traced story where we could measure this cleanly, 14.9 hours from the earliest public post we could verify to the first fact-check article in our corpus, and 31.0 hours to the Snopes entry. That is a single measurement, not a benchmark. We have not published an average because one data point does not have one, and we would rather say that than invent a figure people will quote.
We have measured it on three stories: 130.2 hours, 56.4 hours, and 56 days. The range is wide enough that an average over it would describe none of the three. What the spread shows is that fuse length depends on what kind of story it is, which is why a trace reports the earliest node and its text rather than only a speed.
Because we do not have one yet, and a benchmark computed from four traces would be a number people repeat for years without checking what produced it. Every trace deposits its lead times into the same graph. When the sample is large enough to support a distribution, we will publish it with the sample size, the window and the platform mix attached, and it will be measured rather than estimated.
No, and Trace does not claim to know which stories will break out. It reports what has already happened: the earliest public appearance it could verify, who amplified it, how fast it has moved so far, and what the sweep could not reach. Breakout is shown as High, Medium or Low with the rule that produced it printed next to it, and it will stay a label rather than a percentage until we have enough recorded outcomes to back a percentage.
A free trace shows the earliest public post we can verify and how fast it has moved since.
Honest fine print. Trace reports the earliest public appearance we could verify inside the window we searched, which is not the same thing as a proven origin. We only read public posts and public metrics, so deleted posts, screenshots laundered out of private accounts, and anything that started in a DM or a locked group are invisible to us, and every result names the sources that came back empty. Confidence is a number we publish together with the gaps behind it, never on its own. Trace describes what has already happened and how fast it is moving right now; it does not predict what will go viral, and no honest tool does.