It usually starts the same way. A 40-second clip lands on your timeline. Grainy studio lighting, a familiar journalist’s voice, a club crest in the corner, and a sentence that makes your stomach drop: your club has agreed personal terms with a player you never thought you’d get. Fifteen thousand shares later, someone finally points out that the journalist never said it, the studio doesn’t exist, and the whole thing was generated in an afternoon.
Football has always run on rumours. What changed is that rumours now come with a face, a voice, and a timestamp.
Why football is the perfect target
If you wanted to design a subject that AI-generated content could exploit, you’d end up with something very close to football. The audience is enormous and emotionally invested. There’s a permanent stream of unverifiable claims, because transfer talks genuinely do happen behind closed doors. And there’s a built-in excuse for low production quality: half of all real transfer news breaks as a screenshot of a tweet or a shaky phone recording outside a training ground.
That last point matters more than people realise. In most fields, a badly lit, oddly compressed video is a red flag. In football, it looks like a scoop.
Add the money side and the picture completes itself. Fake transfer content drives clicks, clicks drive ad revenue, and a channel that publishes twenty AI-generated “BREAKING” videos a week only needs one of them to go viral to be profitable. Nobody involved has any incentive to slow down.
The gaming side is just as bad
Football gaming communities have their own version of this problem, and it arguably arrived earlier.
Every year, in the months before a new FC or Football Manager release, the same cycle plays out. Fake gameplay footage appears, showing menus that were never in the build. Ratings lists circulate as screenshots with plausible-looking fonts and a couple of deliberately controversial numbers to guarantee arguments in the comments. Entire YouTube channels exist purely to publish “leaked” content that nobody leaked.
The generated stuff has become genuinely difficult to dismiss at a glance. A few years ago, a fake gameplay clip fell apart the moment you looked at the crowd or the sideline. Now the giveaways are subtler: a scoreboard that shifts by a pixel between frames, a stadium that changes shape in the background, player faces that are almost right but sit slightly wrong on the neck, or ball physics that look convincing for two seconds and then stop obeying gravity.
The frustrating part is that real leaks also exist. Beta builds do get shared. Screenshots do escape. So “this looks like a leak” is not evidence either way, and the community ends up arguing about authenticity instead of the actual game.
What to actually do with a suspicious clip
The honest advice is that your eyes are no longer a reliable instrument here, and pretending otherwise is how people get caught out. The old checklist — count the fingers, look at the ears, check for warping around the jaw — worked on a generation of models that has already been replaced.
What still works is a bit more boring:
Find the origin, not the copy. The version you saw is almost never the first one. Search a distinctive phrase from the clip, or reverse-search a frame. If the earliest instance is a two-week-old account with no other content, you have your answer.
Check the silence. Real transfer news is competitive. If a genuine agreement had been reached, the established reporters would be on it within minutes. A “scoop” that no credible outlet has touched after several hours is not a scoop.
Listen to the audio separately. Voice cloning has improved faster than video, but cloned speech still tends to be too even. Real people stumble, breathe audibly, change pace mid-sentence, and step on their own words. A perfectly metered delivery over stock footage is a warning sign.
Run it through a detection tool. When the clip actually matters — you’re about to publish it, or a whole fan forum is about to spend the evening arguing about it — pushing the file through an AI video detector takes seconds and gives you a probability score based on artefacts you can’t see. It won’t be right every single time, and it shouldn’t be the only thing you rely on, but combined with the checks above it turns a gut feeling into something you can defend when someone challenges you in the replies.
The other side of the problem: people making content
Detection is only half the story. The other half is that a large share of football coverage is now produced with AI assistance, and most of it is produced badly.
You’ve read these articles. Match previews that describe both teams as “formidable opponents in what promises to be a fascinating encounter.” Save reports that open with “in the ever-evolving world of Football Manager.” Transfer roundups where every paragraph has the same rhythm and nothing has an opinion. The information might be accurate, but it reads like it was written by something that has never watched a game.
There’s no point being precious about this. Creators use these tools because the alternative is publishing three times a week instead of every day, and the audience has been trained to expect every day. The problem isn’t the assistance, it’s shipping the first draft.
If you write this way, the fix is a workflow, not abstinence. Use the model to get the structure and the facts down, then do the part it can’t do: put your actual take in it, cut the hedging, and let the sentence lengths vary the way human writing does. For the cleanup pass on the prose itself, a humanizer strips out the tells — the stock transitions, the flat cadence, the compulsive triple adjectives — and leaves you something worth editing rather than something worth deleting. Then read it out loud. If you wouldn’t say it to someone in a pub, it doesn’t go up.
The reason this matters isn’t just style. Readers can feel generic writing even when they can’t name what’s wrong with it, and once a site feels automated, they stop trusting what it says about transfers too. In a subject where credibility is the entire product, that’s an expensive thing to lose.
The short version
Assume video is unverified until you’ve checked where it came from. Assume that anything that perfectly confirms what you wanted to be true was built to do exactly that. And if you publish football content yourself, remember that the tools that generate it and the tools that expose it are improving at the same rate — the only durable advantage is being the person who actually checked, and who actually had something to say.