Content Tagging Error: Why a Televisa Reality Show Was Classified as Football

Sunday, 20 September 2026. An automated content-classification system pushed out a record, and it carried a single tag: football.

The record contained 11.8 million viewers, sourced to ACAM. It contained an elimination round. It contained a salvation test. It contained an audience vote that decided one person's fate. And it contained two names, repeated over and over: Cynthia Klitbo and Ernesto Laguardia. Not one club. Not one player. Not one goal, not one decisive pass, not one expected-goals figure.

I read that record three times. The first time I laughed, the laugh of a 41-year-old man who has grown used to watching systems cut corners. The second time I reopened it to check whether I had misread a line. By the third time I stopped laughing, because the party that needs to be questioned here is us — the people who taught the machine what football looks like, and taught it wrong.

The show in that record is La Casa de los Famosos Mexico 2026, broadcast by Televisa. It is a reality-television format in which a group of celebrities live together in a house, build alliances, position themselves, and are eliminated one by one through gala nights decided by the audience's vote. The 2026 season is entering its final stretch.

Sunday 20 September was the night Cynthia Klitbo, a Mexican actress, was eliminated from the game. That result shattered the alliances inside the house and, by the show's own description, dealt a brutal blow to the group known as "el cuarto".

The figure of 11.8 million viewers comes from ACAM, Mexico's television-audience measurement body. With that number, the gala became the most-watched programme in the Sunday slot. The same record also carries the story of Ernesto Laguardia, a television presenter, whom Cynthia Klitbo called "two-faced" and described as having a selfish attitude. That is the entire raw material: one elimination, one vote, one clash between two presenters, and one audience figure.

So why football?

Because in many newsrooms, Vietnam included, football is no longer handled entirely by human eyes. It is handled by a pipeline. A system scans thousands of items a day in Spanish, English, Portuguese and Vietnamese; it tags the domain, scores relevance, pushes the article to the page. Working at that speed, the system does not read meaning. It reads word frequency.

And here I have to be blunt, because I have sat on the other side of that desk: dependence on speed is turning football into a column produced assembly-line style, where errors go undetected simply because nobody has time left to detect them. Google's 2026 algorithm moves in the opposite direction — it rewards information gain and punishes derivative content. But by the time any algorithm punishes anything, the reader has already been fooled.

To understand how a Mexican television show slipped into a football tag, look at the language.

Spanish has a word that every football classifier loves: partido. It means "match". It also means "political party", and in reality-television context it can appear as an organisational unit. A model trained on football vocabulary sees partido and can hardly avoid thinking of football.

Then estrategia — strategy. In the house, that is survival strategy: who to stand beside, who to betray, when to stay silent. On the pitch, that is pressing strategy, a defensive block, a transition plan. The same word, two entirely different worlds.

Then alianzas — alliances. In football, alliances exist between clubs, between owners, between leagues. On the show, an alliance is three people promising not to put each other's names on the nomination list.

Then competencias — competition. In Spanish sports usage, that is the word for a tournament. In television, it is the word for any contest programme.

Then eliminacion — elimination. Football has knockout rounds. Reality television eliminates people. The same verb, the same sense of tension, but a completely different nature.

Then alineacion — in football, the starting line-up; in other contexts, alignment. Then plantilla — in football, the squad; in administration, a template. Then salvacion — on the show, the salvation test; in religion, salvation. Then audiencia — in media, the audience; in law, a court hearing.

A classifier that reads words without reading meaning will always find football wherever there is competition, alliance, strategy and elimination. That is the classic false-positive failure of natural-language processing: sports keywords scattered across a text that contains no sport at all.

Now look at the number. 11.8 million is a television audience figure, and I have to state this clearly because I know how many articles will rush to turn it into a football metric. In a football dossier we have broadcast revenue, wage bills, contract amortisation, financial-sustainability obligations. None of that is here. No transfer, no renewal, no release clause, no agent. There is exactly one number, and it belongs to the Mexican television advertising market.

But there is one genuine football question sitting behind this record, and it is my favourite question in the whole piece: where was Liga MX in that Sunday slot?

This is what a local reporter might skip but a half-insider, half-outsider like me cannot. In France I write about Ligue 1, and I can see plainly that football does not compete with football. It competes with everything else on the screen. A football league is no longer the only product selling tension on a Sunday evening. A reality show also sells tension, and it sells it cheaper, more accessibly, and without depending on the result of any single match.

In Vietnam I see another version of the same story. A V-League match has to compete with a Premier League match, a Champions League match, and entertainment programming in the same slot. The power of Vietnamese football is measured not only by the number of spectators in the stands, but by whether viewers still switch the channel on at the right hour.

Back to the technical problem. The principle of null handling states that when a record holds insufficient information for analysis, the correct move is to declare "insufficient information" rather than to guess. In this case, all the usual football analysis dimensions — tactics, finance, results, league landscape, rules, dressing room, risk, media narrative, transmission chains — return empty. Not because the analyst is lazy. Because football is absent from the text.

And here is the genuinely frightening consequence. If a language model is forced to "analyse football" from this record, it will have to invent. It will build a team for Cynthia Klitbo. It will describe Ernesto Laguardia's tactical shape. It will talk about form, about runs of results, about managerial pressure. All of it fluent, all of it linguistically plausible, and all of it false. That is data hallucination, and in football, data hallucination is not merely embarrassing — it corrodes trust.

The right guardrail is not better prose but a domain-validation gate: before any football analysis is allowed to run, the record must contain at least one genuine football entity — a club, a league, a player tied to a competition. No entity, no analysis.

For the Vietnamese market, this trap is not remote. Vietnamese has its own words that belong both to football and to other fields: "doi hinh" (line-up) appears in music and esports; "chien thuat" (tactics) appears in games, in business, in every meeting; "loai" (eliminate) appears in every contest; "gala" appears at every awards night. A Vietnamese-language pipeline running an old football keyword set will tag an article about a music awards night as football without blinking.

And I know that because I once operated a manual version of that very pipeline. In 2026, aged 32 and a senior analyst at a French football magazine, I watched Monaco win Ligue 1 with 95 points. Kylian Mbappe scored 15 league goals and 6 in the Champions League before moving to Paris Saint-Germain. I published a piece titled "Mbappe is only a product of the system", showing that 78% of his expected goals came from passes by Bernardo Silva. The article drew 2.3 million reads and a debate that ran for three weeks.

I once owned a truth I had carved myself, until Mbappe smashed it to pieces.

This is the part where I have to contradict myself, because if I sit here scolding a machine for mislabelling while I once mislabelled a human being, then I am putting on a farce.

In the summer of 2026 the editorial board sent me to Russia as a live commentator, effectively as payment for the 2026 piece. I sat in the stands and felt what no spreadsheet can ever carry: the energy of a stand full of flags, the singing, the silence before a corner. In the Luzhniki final, France beat Croatia 4-2 and Mbappe scored in the 65th minute. Standing amid a forest of French flags, watching the boy I had called a fleeting product lift the trophy, I burst into tears and hugged a Croatian reporter I had never met.

That night I wrote a piece arguing against myself: "I was wrong, Mbappe is not a number". Numbers are never wrong; only the person reading them and assuming he is right.

So when I say this pipeline got it wrong, I must add: I am that pipeline, only slower, capable of tears, and paid to apologise in public.

But there is a philosophical temptation here that I want to refuse before it convinces me. People will tell me: football is just a reality show too. A World Cup final draws 1.5 billion viewers. It has knockout rounds. It has a public vote of sorts. It has people cast out. Watching a gala eliminate someone feels no different from watching a penalty shootout. So tagging a television show as football is no catastrophe — it is simply honesty about the nature of both.

I understand the appeal of that argument. The pandemic podcast taught me that silence is also a form of interviewing, and I learned that the questions you turn against yourself are usually the best ones. Emotion is not the enemy of reason; it is the silent analyst.

But there is one point that argument deliberately skips. When a reader opens a football analysis, they are not opening a poem. They are placing attention, trust, and sometimes money into it. In French they say ambiguity is a gift of literature; in sports journalism, ambiguity is a typo. I could write a fine piece arguing that football and reality television are twins. I will not publish that piece under the tag "football analysis", because that piece has no Spain 4, no Croatia 2, and no 65th minute at all.

My prediction, and I always want my predictions to be testable: within the next 12 months, at least one football analysis will be published in Vietnam whose source text contains no football entity whatsoever.

When you come across such a piece, you need know nothing about machine learning. You need ask one question only: which club, which league, which player? If the article cannot answer, you are reading a record that was tagged wrongly. Every time I say "surely", football finds a way to prove I am an idiot.