How we check the facts on every page
We draft our pages with AI help. A separate AI fact-check pass then opens the sources the draft cites and removes anything it can't support, so every number and date left on a page has a source and the day we read it. If we still get something wrong, email [email protected]: we aim to fix a confirmed mistake within 2 working days.
How we fact-check AI-drafted pages
Every guide, hub and trend explainer on the site goes through the same five steps.
- Gather the facts, with sources
Before anything is written, each fact for the page is noted with where it came from: the post itself, a Google Trends read, a news report or a reference site. A fact with nothing to point to doesn't go into the draft.
- AI writes the draft
An AI model drafts the page from those notes, plainly and with dates. We use AI for speed: across the six AI video remakes of 2026 we timed, Google search peaked a median 4.5 days after the first AI remake. A page that took a week to write would arrive after the peak.
- A separate AI pass checks it
A second AI pass, run apart from the one that wrote the draft, opens the sources the page cites and compares every claim with what they say. Whatever it can't support is removed.
- A script checks what's left
Our page check stops a page from shipping if a link goes nowhere, a required part is missing or it repeats another of our pages too closely. It also flags pages that cite no sources or mention lyrics, so they get another look.
- Numbers keep their read date
Play counts change by the hour, so each one shows the day it was read. Guides and hubs pull trend numbers from the trend's own page as you load them, so they're never older than that page. On 16 Oct 2026, the numbers on all 8 trend pages were re-read that day.
Every number on these is dated: try one free
All trends →Add your selfie once. See yourself in every trend here, free.
Adults only. Deleted after 30 days and never used for training.
Make yours
Make yours
↗ Rising
Make yours
↗ Rising
Make yours
- Free preview first
- Pay only when it's ready
- Automatic refund if it fails
What an AI fact-check can and can't catch
Our checking pass is AI too, so we don't call it an independent fact-checker. It can tell whether a source says what our sentence says. It can't tell whether the source itself is right, and it can miss things.
So we make each source easy to judge. A claim says who it comes from, the post itself beats anything written about it, and when sources give different numbers we'd rather show both than pick one. For the 2022 studio performance behind Hotel Lobby, Know Your Meme counts more than 121M views, while YouTube showed about 131.7M when we read it on 16 Oct 2026. The trend page gives both numbers, each with its source.
What a person does decide is which trends you can make. None goes live here until the person who runs the site approves it; our about page says who that is.
Why spell this out? Google's guidance on helpful content (updated 5 Oct 2026) says AI or automation disclosures are useful wherever readers might ask how a page was made. This page is our disclosure.
What our pages cite: 415 sources, counted 16 Oct 2026
We counted the sources listed on our live guides, hubs and trend explainers on 16 Oct 2026: 415 across 30 pages. Every one of them shows the day we read it.
| Kind of source | How many | What we use it for |
|---|---|---|
| Posts on TikTok, YouTube and Instagram | 199 (48%) | Who posted first, when, and how many plays. The post itself comes before any write-up of it. |
| News, fact-checks, studies, company posts, laws and domain records | 61 (15%) | Dates and context the posts can't show, credited to whoever published them. |
| Other AI trend sites, templates and tools | 50 (12%) | That a template exists and when it appeared. Their numbers are credited to them by name, never passed off as ours. |
| Reference sites: Wikipedia, Know Your Meme, song and chart databases | 48 (12%) | Origins, release dates and song credits, with the site named. |
| Google search data: Trends and autocomplete | 47 (11%) | When search started, peaked and faded, and the questions people type. Trends is relative interest, never a count of searches. |
| Our own pages: legal pages, trend lists and our search-timing data | 10 (2%) | Our rules, our trend pages and the trend lifespans we measured. |
Read dates: 293 on 16 Oct 2026, 116 on 15 Oct, 4 across both days and 2 on 14 Oct. The site is new: our domain was registered on 13 Oct 2026.
Three trend pages that show these rules
All three are live, and each shows one rule from this page. Add a photo and the free preview still usually takes under a minute.
- Hotel Lobby ≈137M views · 2 people · 15s The stage, told straight. On 16 Oct 2026, its 32nd day, the page called it slowing: the biggest remakes had barely moved since 14 Oct, though people were still searching 'hotel lobby ai'. It's for two, so bring a friend to the hanging mic. See yourself in it
- Raindance ≈200M views · 2 people · 28s What we leave out. The music video had about 285M YouTube views when we read it on 16 Oct 2026, and none of them are in the trend's estimate, because watching the video isn't joining the trend. 28 seconds on the pier at golden hour, for you and your person. See yourself in it
- Smalltown Boy Cats↗ Rising ≈46M views · 2 pets · 21s A hidden number, said out loud. Instagram hides the original's views from logged-out readers, so its timeline gives the likes we could see, about 1.31M on 15 Oct 2026, and says why. It explains the 1984 song without quoting it. Your two cats at the red fence. See yourself in it
How we label what we can't confirm
Some facts can't be pinned down. When that happens, the page says so with one of these labels:
- Estimate. A total we work out from the posts we can read, given with a likely range. Hotel Lobby: about 137M views, likely 87M to 247M, read on 16 Oct 2026.
- As read on a date. If a post won't load when we go back to it, we keep its last count with the day we read it. On 16 Oct 2026 we couldn't read Rumpelstiltskin's Vampire Diaries recast, so its page kept the 1.8M plays from 15 Oct with that date. When it loaded again on 17 Oct, we read it fresh: 1.9M.
- The earliest we found. An origin date is the first post we could find, and an older one may exist, so the page says so. Rumpelstiltskin's page dates the earliest TikTok edit we found to 27 Sep 2026, two days before the edit people often credit.
- By their count. A number we take from another trend site is credited to that site by name, so you know it isn't ours. Some pages add 'not verified by us' when we had no way to check it.
Why we never print lyrics
Song lyrics are copyrighted, and we have no licence to print them. So when a trend runs on a song, its page names the song and the artist, explains what the song is about and why it suits the clip, and stops there.
Alors on Danse shows how that reads: its page says the title means 'so we dance' and that the song is about dancing as a way out when work and money pile up. If you came looking for the words, you'll find the meaning here and the song itself in your music app.
Found a mistake? Here's how we fix it
Email [email protected], the same address as on our terms and privacy policy. Send the page link, what's wrong and, if you have one, a source that shows it. Then:
- We check it against the sources, the page's and yours.
- We fix a confirmed mistake, aiming for 2 working days, the same target as our takedowns. If the source itself was wrong, the claim comes out or gets a better source.
- The page's Updated date moves. It changes only when the words on the page change, never because a number was re-read, so it shows when we last corrected or rewrote something.
- We reply to say what we changed, or why we didn't.
- Not a page error? A video of you made without your consent goes through report content, and a request to correct your own data goes through our privacy policy.
Questions people ask
How do you fact check AI content?
Check each claim against the source it came from, not against another AI answer. That's what our second pass does: it opens the sources a draft cites and removes what it can't find there. It can't make a wrong source right, which is why every claim on our pages names where it came from.
Can AI fact check itself?
We don't let it mark its own work. The pass that checks a page is separate from the one that drafted it, and it has to find each claim in a source it has opened, not in what it already 'knows'. It's still AI, though, so we don't call it an independent fact-checker.
Why does AI make up sources?
Because language models sometimes guess when they're unsure, and a guess can look just like a real citation. A 2025 research paper, Why Language Models Hallucinate, argues that the way models are trained and scored rewards guessing over admitting uncertainty. That's why our check opens the sources instead of trusting how a citation looks.
How do you fact check AI videos?
Start with a fact-checker's rating, then look for details that change between frames. Lead Stories rated the lion reunion reel AI-generated on 26 Aug 2026 after spotting a watch that disappears from the woman's wrist. Our Lion Friend page, about a trend we track but don't make, cites that rating rather than our own eye.
Is Google Trends accurate?
Good for timing, not for counting. Google says Trends works from a sample of searches, scales results from 0 to 100, adds small random noise that shows most on low-volume terms, and should be one data point among several. We use it to date when search for a trend started, peaked and faded, never as a number of searches.
Is Know Your Meme reliable?
It's a useful starting point. Its about page says any registered member can submit an entry, which its editors and moderators then research to confirm or reject. We name it whenever we use it, and where we can we also read the post it points to: for Hotel Lobby's original we give its count and YouTube's own.
Are TikTok views accurate?
They're TikTok's own count, and we have no way to check them independently. Big counts are also shown rounded: the Bulin 47 original read 8.2M plays on 16 Oct 2026. So we copy each count as shown, with the day we read it, and anything we add up ourselves is labelled an estimate.
Sources
- Google Search Central, Creating helpful, reliable, people-first content: the "How" section on AI and automation disclosures (updated 5 Oct 2026) · read 16 Oct 2026
- Jump on the Trend sitemaps: the 30 live guides, hubs, roundups and trend explainers with a sources list (415 sources, all with read dates) · read 16 Oct 2026
- Jump on the Trend, how long TikTok trends last: first AI remake and search-peak dates for the 2026 video remakes (median gap 4.5 days, our Google Trends dataset of 15 Oct 2026) · read 16 Oct 2026
- Hotel Lobby trend spec: Know Your Meme's 121M+ and YouTube's 131.7M for the 2022 original, the 137M estimate (87M to 247M), slowing on day 32 · read 16 Oct 2026
- Raindance trend spec: the music video's 285M views, not counted in the estimate; the 28-second clip · read 16 Oct 2026
- Smalltown Boy Cats trend spec: 1.31M likes on the Instagram original (15 Oct), views hidden when logged out · read 16 Oct 2026
- Rumpelstiltskin trend spec: the earliest TikTok edit we found (27 Sep), the often-credited 29 Sep edit, the Vampire Diaries recast kept as read on 15 Oct and re-read on 17 Oct · read 18 Oct 2026
- Alors on Danse trend spec: what the song and its title mean · read 16 Oct 2026
- Bulin 47 trend spec: the original freestyle at 8.2M plays and TikTok's rounded counts · read 16 Oct 2026
- Google Trends Help, FAQ about Google Trends data: sampling, the 0 to 100 scale, added noise, not a poll · read 16 Oct 2026
- Know Your Meme, About: member submissions, then research by editors and moderators to confirm or reject · read 16 Oct 2026
- Lead Stories: the lion reunion video is AI-generated, and the watch that disappears (26 Aug 2026) · read 16 Oct 2026
- Kalai, Nachum, Vempala and Zhang, Why Language Models Hallucinate (arXiv, 4 Sep 2025) · read 16 Oct 2026
- Jump on the Trend, Terms of service and contact address (last updated 15 Oct 2026) · read 16 Oct 2026
- Jump on the Trend, Report content: takedowns usually within 2 working days · read 16 Oct 2026
- Google Registry RDAP record for jumponthetrend.app: registered 13 Oct 2026 · read 16 Oct 2026
Trend numbers on this page come from each trend's own page, where you'll find the posts they were read from. We never invent numbers; estimates are marked.
JUMP on the TREND
before it's over.
Pick a trend
- Free preview first
- Pay only when it's ready
- Automatic refund if it fails