AI-Generated Images and Transfer Rumours: The Three-Step Chain That Blinds Scouts

At 2:47 on the morning of January 31, with the winter transfer window hours from closing, a photograph of a young striker standing in an airport with a green suitcase spread through the group chats of the scouting world. I opened my laptop, zoomed in, and did what I always do with a new image: I checked the shadows, the reflection of the ceiling lights on the floor, and the letters on the sign behind the player's shoulder.

The image was impossibly smooth. The hair on the player's shoulder merged into the wall instead of separating into strands. The logo on the suitcase was warped at the corner. The sign behind him misspelled a single letter in the name of the city.

No such flight existed. No such player stood there.

By six in the morning the photo had passed through forty posts. By nine, two sports outlets had carried it under the familiar phrase "an image said to show". By noon, the club denied it. The denial always runs slower than the photograph, and that delay is built into the structure of the market.

I have tracked hundreds of chains like this one. They are not random. They have a structure, and the structure repeats often enough to be described in three steps.

A MARKET THAT PAYS FOR CONFIRMATION

The transfer window is the only period in the year when an unverified claim can move a club's commercial value, its shirt sales, and even a fund's decision to commit capital. A big window generates thousands of claims: true, half-true, and entirely invented. Journalists have minutes to check. Fans have seconds to share.

Fabrication used to be expensive. To fake a photo of a player in a new shirt, someone needed editing skill, a high-quality source image, and a few hours. That barrier accidentally worked as a filter. Liars had to invest effort, so there were few of them, and the serious ones left traces.

The barrier is gone. A single sentence now produces an image in seconds. A short audio clip produces a fake phone call. A blank page produces a medical report that looks real. The cost of faking has fallen to near zero, while the cost of verifying has not fallen at all. It has risen, because the verifier must now check more layers than before.

I began writing in 2026, at twenty, at a local newsroom. The discipline then was simple: see it yourself, hear it yourself, write exactly what you saw. Years later, in Jakarta, I learned something else.

In 2026 I attended the AFF U19 Championship. In the match between Vietnam U19 and Malaysia U19, which finished 4-1, I fixed on a sixteen-year-old named Nguyen Minh Quan, a product of the PVF academy. He recorded three assists, an 89 per cent passing accuracy, and one dribble past five players before scoring.

I wrote a 2,500-word piece calling him "Vietnam's Messi". The newsroom objected for lack of verification. I still believed I was right. Three years later, Quan vanished from the map of elite football. No club mentioned him again.

That was the first time I understood that the person who manufactures a false confirmation is not necessarily a liar; often he is simply an observer who got excited. I had no AI. I had one match, one data sheet, and one belief.

THE THREE STEPS OF A CHAIN

Step one is a real but ambiguous detail. In football it can be a photo of a player having lunch with a sporting director, a video of a player at an airport during a layover, or a social-media follow that lands at exactly the right moment. These details are not false. They simply do not say anything specific.

Step two is a fabricated piece of evidence made to confirm. This is what is new about the current decade. If step one plants a question, step two answers it with an image that looks like the answer.

Step three is a framing account. A parody handle or a reposting account shares the fabricated evidence with a neutral caption, something like "according to sources close to the situation". The screenshot then loses the parody label. From that point the image stands alone, without a source, and becomes fact.

The three steps work like an assembly line. Remove step one and the story has no anchor in reality, so nobody believes it. Remove step two and it is an ordinary rumour, and ordinary rumours are infinite. Remove step three and the fabricated evidence stays in a small corner, out of reach of the crowd.

I saw the cleanest version of this chain in a case unrelated to football, reported by The Express Tribune in 2026. An AI-generated image of the singer RAYE and the actor Michael B. Jordan at the MTV Video Music Awards circulated as evidence of a romance. It was pushed out by a parody account, alongside a genuine photograph of the two at a theme park, where the camera angle and distance led many readers to see something else. RAYE later denied it, saying the two were just friends. The paper also stated clearly that the kiss image was AI-generated.

The structure matches what I see in football. A real detail. A fabricated confirmation. A framing account. And a newsroom forced to run behind and correct it.

THE ECONOMICS OF A RUMOUR

Nobody stops the chain because the chain is profitable. The outlet that publishes first receives the traffic. The outlet that publishes the correction receives a small fraction of that traffic. This is a structural asymmetry, and every editor knows it exists.

One example, in numbers. When Cristiano Ronaldo moved to Al-Nassr in late December 2026, international outlets reported a package that could reach around two hundred million euros a year. The figure was large enough to spawn weeks of secondary rumours, most of them baseless. When Mykhailo Mudryk moved from Shakhtar Donetsk to Chelsea in January 2026, the reported fee was around eighty-eight and a half million pounds, and in the two weeks before the deal closed, the number of false claims about which club he would join exceeded the number of true ones.

Real deals generate real money. Fake rumours do too, except the money flows elsewhere: into traffic, into advertising, into the short-term reputation of whoever posted first.

There is a layer of interest that is rarely discussed. Ambiguity has value for agents and for clubs. An unconfirmed rumour can raise an asking price. Silence is a negotiating tool, and in some cases refusing to kill a rumour is a deliberate decision.

WHEN SCOUTING REPORTS BECOME GARBAGE

Scouting reports are built on video. A scout watches the match, takes notes, cross-checks, and writes. If the video is fake, the report is fiction. If the data is fake, the report is fiction with tables.

In recent years academy databases have grown larger, and the larger they grow, the easier they are to poison. A false height, a blurred injury history, an inflated record of goals can all be entered into a system and survive there for years. Nobody re-checks a file that has sat in the database since three seasons ago.

Based on my experience watching matches, I built a source-ranking table with three columns: type of source, verifiability, and the price I pay if I am wrong.

Grade A is the club's official announcement and the league's registration document. Grade B is two independent journalists with a record of being right. Grade C is a single journalist. Grade D is social media with an image. Grade E is an anonymous account claiming a private source.

The rule I set for myself is this: the more emotionally satisfying a claim is, the higher the verification cost must be. Rumours do not spread because they are true. They spread because they satisfy an expectation that already existed. When a young player is linked to the club fans want him to join, the fake image does not need to be perfect. It only needs to arrive on time.

I also learned that rewatching the full ninety minutes matters more than watching highlights. A highlight reel is a form of editing, and every edit carries the editor's intention. A two-minute clip can turn an average player into a compelling one, and that was happening long before AI appeared.

WHO PAYS IN THE END

Nguyen Minh Quan was not an AI product. He was a real player, broken by a real story written by human beings. I was one of the people who wrote that story.

In 2026, in the middle of the European Championship, I flew to Tokyo to cover the Olympics. With no Vietnam team to follow, I ran into Quan by chance at a training ground in the JFL, the fourth tier of Japanese football. He wore number eight, and told me he had left Benfica B after two seasons and only four appearances.

I wrote a 3,000-word profile. This time I did not call him Messi. I wrote about the psychological journey. It drew more than two hundred thousand reads, the highest of the year for the newsroom, largely because readers were curious about a man who had disappeared and reappeared.

What I learned from Quan is what I still tell younger colleagues: data is the site, psychology is the artefact. A data sheet tells me how far a player ran and how often he shot. It does not tell me how many hours he slept after reading an article that called him the heir to a legend.

In the era of synthetic images the burden is heavier. A young player used to answer only for what people wrote about him. Now he must also answer for what people build about him: a face attached to a statement he never made, a video he never recorded, a contract he never signed.

THE OTHER SIDE

The common worry today is that AI is manufacturing fake news. I think that framing misses the axis.

What AI does is lower the cost of something the market was already willing to buy: confirmation. Demand came first; the tool came second. If every image model vanished tomorrow, the three-step chain would still run, only a little slower, using editing software, a fan with a phone, a bored teenager.

In 2026, I was that tool.

Another counter-intuitive point: corrections do not fix much. A correction has a half-life measured in hours. The original has a half-life measured in years. In the RAYE and Michael B. Jordan case, the Express Tribune correction reached a small fraction of the people who saw the fake image. People remember the photograph. They rarely remember the correction.

And the most uncomfortable part: the parties with power in football do not necessarily want the fog to clear. Fog helps negotiations. A club may stay quiet when a rumour lifts a player's valuation. An agent may stay quiet when silence raises the pressure. Fighting misinformation costs money and wins no points in the table.

WHAT REMAINS

My probability estimate for the next few windows: synthetic content will grow, because the tools keep getting cheaper and the skills keep spreading. At the same time the market is adapting. Some clubs now require original metadata for images and video. Some leagues are building centralised player records to resist tampering. These responses are slow, but they are real.

I do not think this fight ends in a victory. I think it ends in a habit. Like checking fuel before a long drive, or checking an expiry date before buying medicine.

Fifty-three years have taught me this: fast wins, but slow sees.

That night of January 31, the airport photograph dissolved after two days, leaving nothing behind except a deleted post and a young player who had to answer a few questions from his roommates. I still keep the image in a separate folder, named by its date. I do not keep it as evidence. I keep it to remind myself that every touch is a mark cut into the stratum of a match, and anyone reading those layers must learn to tell real soil from soil that was just piled on top.

Rough gems are not on the map. They lie in the dust of the running track. And in this era, that dust may be real, or it may have been created seconds ago.

I dig through data, but what I excavate is people.