Translation will always require a critical and human perspective

Diana Alonso

If it's a lawyer, doctor, or other professional, it's most likely a man, and if we're talking about nursing or a person dedicated to caregiving, we'll probably imagine a woman.

That is one of the logics with which some translation engines and artificial intelligence (AI) process texts, according to the experience of Alicia Cortés, translator, editor and teacher of the Department of Languages (DEL) of ITESO, who for several years worked correcting and translating metadata, synopses and reviews of audiovisual content for different entertainment platforms.

However, this does not mean that AI is inherently bad or sexist. During the conference “The AI Slop Virus in Translation,” held as part of the Second Translation and Interpreting Forum 2026, “Translation in the Face of AI: The Irreplaceable Value of Human Judgment,” Cortés explained that the responses generated by these tools depend on the data with which they were trained, the information available on the internet, and the feedback obtained from millions of human interactions. In other words, AI also reproduces the values, prejudices, errors, and biases present in the information it consumes.

Translation involves much more than finding equivalents between words in different languages. It requires understanding historical, social, and cultural contexts. Therefore, knowing the limitations and biases of these tools is fundamental for those who work with language and allows us to challenge the myth that AI will completely replace translators. This is a threat that is often directed at almost every profession.

“Understanding the intention behind words, recognizing cultural nuances, interpreting silences, weighing the ethical implications of a linguistic decision, and taking responsibility for that decision [...] remain profoundly human tasks,” said Elsy Villegas, coordinator of the Bachelor's Degree in Translation and Interpreting at ITESO.

According to Cortés, some of the most frequent biases in AI are algorithmic bias, when the system incorrectly interprets an instruction; cognitive bias, related to human errors present in both users and those who develop these technologies; confirmation bias, which leads to reproducing erroneous information without distinguishing whether it is true or false; sampling bias, which occurs when certain groups or perspectives are left out of the training data; and prejudice bias, which replicates existing stereotypes and social trends.

Cultivating your own style and maintaining critical thinking is fundamental in the face of phenomena such as the AI slop , a term that refers to the enormous amount of AI-generated content circulating on the internet, characterized by its low quality, lack of verification, and high viral potential.

The translator warned that constant consumption of this type of material is beginning to be linked to phenomena such as brain rot , a cognitive decline associated with overexposure to digital content designed to capture immediate attention. She explained that this affects the ability to concentrate, fosters dependence on quick stimuli, and impairs basic skills such as reasoning and simple calculations.

One of the effects that worries her most is the loss of the translator's "voice." For her, the AI's style can be identified relatively easily: short, impactful phrases; repetitive structures; examples organized in groups of three; contrasts like "it's not this, but that"; and vocabulary that aims to sound technical or sophisticated. The problem, she said, is that these formulas end up unconsciously creeping into our own way of speaking.
“It is true that as translators we have to be invisible, but we also have our own style and way of achieving that goal. That personality we have is ultimately what gives us value as professionals,” he stated.

Although Cortés acknowledges that, in many cases, it's faster to start a translation from scratch than to correct an AI-generated text, he believes the solution isn't to completely reject these technologies. The key, he argues, lies in learning to use them critically, understanding how they work, and knowing how to communicate with them using precise instructions, especially in an industry where their use is becoming a job requirement.

“The only way to fight AI slop is by creating original content,” he concluded.