Open a Spanish general newspaper, open one of the popular YouTube channels, open an X thread about OpenAI's latest, open a conversation with your brother-in-law over a family lunch. You'll get three opposing accounts told by the same people at different times of day. In the morning, existential risk. In the afternoon, infinite productivity. At night, the Skynet joke. The one warning you of the apocalypse is the same brand selling you salvation. And while the two accounts compete for your attention, there's a third corridor — the one running under your feet, the one of the quiet transformation already happening — that almost nobody names. It's worth asking why.
Today we're talking about how ignorance takes the chair. The media don't know what artificial intelligence is, but they talk about it and pontificate. It isn't reason speaking. It's superstition dressed as reason. Which is more dangerous, ignorance or superstition? AI has become one of the great references of contemporary magical thinking. And reason — understood in its classic sense, look at the data, tell what we know from what we imagine, accept uncertainty when it's due — has been left out of the conversation. Let's use it here to look at the problem.
First reason: the broadcasters are judge and party
Let's go over who's speaking when someone speaks in Spanish about AI. SaaS brands selling AI: IBM, Microsoft, Salesforce. Banks integrating AI into their services: BBVA, Santander. Consultancies advising on AI: Deloitte, KPMG, EY, PwC. Private universities recruiting enrolments with AI master's degrees: UNIR, VIU, IEBS, ISDI, ESADE. Cybersecurity companies selling protection against "AI threats": Avast, Malwarebytes, Telefónica Tech. General media covering technology from a small department with few specialists. And, occasionally, some academic voice arriving late and badly translated.
The notable thing isn't that these broadcasters exist. The notable thing is that they occupy both poles of the debate at once. The same companies publish one day an article titled "10 dangers of AI you need to know" and the next day another titled "5 advantages of AI for your business". It's not a contradiction: it's traffic capture from both sides. If the reader searches in fear, they find the same broadcaster offering the cure. If the reader searches in enthusiasm, they find the same broadcaster offering the product. The public conversation isn't happening between independent voices with diverse perspectives. It's happening inside the same provider, which has sat in both chairs of the debate and answers itself depending on what the visitor is looking for.
This isn't debate. It's market occupation dressed as debate. And the difference matters, because when a real debate happens, the average reader can triangulate between opposing sources and form judgement. When what happens is occupation, the average reader believes they're triangulating between opposing sources but is actually bouncing between two corners of the same room.
Second reason: the format rewards the cinematic
Let's run a thought experiment. Which headline do you think gets more clicks in a general newspaper? Option A: "MIT study suggests intensive use of AI assistants produces measurable detraining in specific cognitive skills during writing tasks". Option B: "Expert warns: ChatGPT could leave your children unable to think". Option C: "Sam Altman talks about Skynet again and the experts are split".
A is the real story. B is the translation into clickbait. C is the noise. The SERP, the newspaper's front page, the YouTube algorithm, the X feed: all of them reward B and C over A. The cinematic takes the winning position because it generates more clicks, more dwell time, more comment, more virality. The structural — AI Index figures, the evolution of technical benchmarks, the sector's geopolitical dependencies, concrete cases of problematic use — gets relegated to a secondary section if it gets published at all.
The operational consequence is that the pyramid of visibility is inverted. The things that matter, in order of real impact on the reader's life, are roughly: the structural concentration of the sector in four to six companies, the capture of cognitive input, cognitive atrophy from heavy use, the homogenisation of language, military and surveillance use, the labour problems in the annotation chain, the energy cost of data centres, the geopolitical dependence of hardware. That list exists and is documented. What occupies the front page, by contrast, is: machine revolt, killer robot, Skynet, conscious AI, the end of the world. That other list doesn't exist in the empirical sense, only in the narrative sense.
The outlet fulfils its economic function by prioritising the second. The reader fulfils their biological function by getting hooked on the second. The corporate broadcaster fulfils its marketing function by feeding the second. Nobody in that circuit has any incentive to name the first. And so the first isn't named. Not by censorship. By ecosystem.
Third reason: serious journalism is in English
Here we have to be very explicit because it's worth not sugar-coating. Deep journalism about AI, in 2026, exists. But it's almost all in English. Stratechery, by Ben Thompson, publishes weekly strategic analysis of the sector with a depth that has no equivalent in Spanish. Platformer, by Casey Newton, covers the big firms' internal decisions and their cultural consequences. 404 Media does investigative journalism on the dark side of the sector: precarious annotators, non-consensual deepfakes, the use of models for harassment. MIT Technology Review publishes long reports with rigorous verification. The Information offers detailed business coverage. The Atlantic publishes long essays with a voice of their own. The Verge covers the news without losing depth. Wired is still Wired.
In Spanish the list is much shorter and much more uneven. There are excellent individual journalists — Karma Peiró, Marta Peirano, some byline at El Confidencial, the occasional piece at El País — but there's no sustained ecosystem. The general media cover AI with small sections, quick translations of the corporate press release, interviews with Sam Altman treated as if Altman were a neutral expert instead of a CEO with interests. The specialised Spanish-language podcasts can be counted on one hand. Translations of the major books arrive late or don't arrive. The Reuters Institute Digital News Report 2025 confirms a trend that affects all markets but is sharper in Spanish: trust in the media stalls, audiences migrate to alternative sources, and specialised technology coverage concentrates in very few places.
The result is a Spanish-speaking user with a very limited menu. If they want serious journalism about AI, they have to read in English. If they don't read English, they have to settle for what's there. And what's there is, mostly, noise and marketing. Not out of bad faith by Spanish professionals — there are many good ones doing what they can — but out of the industry's structure: few specialists, short budgets, fast publication cycles, dependence on international agencies for almost everything.
Contemporary magical thinking
It's worth stopping for a moment and describing the phenomenon with words that come from anthropology, not marketing. AI has become, in the Spanish-speaking public imagination, a quasi-magical entity. It has vague but immense powers. It can save or it can destroy. It's hard to understand. Its mechanisms are opaque. It speaks. It answers. Sometimes it seems to understand you. The public conversation around that entity — given the mix described above of judge-and-party, cinematic format and the void of serious journalism — has adopted the classic forms of magical thinking: invocation, spell, exorcism, pilgrimage.
Invocation: the SaaS brands that promise to "integrate AI" without the client understanding what that means, like a priesthood mediating between deity and faithful. Spell: the listicles of ten dangers, offering protective rituals ("implement governance policies") without those policies doing anything operationally. Exorcism: the apocalyptic headlines that summon the shadow of Skynet to sell the cure right beneath. Pilgrimage: the AI master's degrees at private universities, where the faithful pay their tuition to have knowledge laid upon them through a certificate.
It's an uncomfortable and deliberately provocative description, but it has the advantage of making visible a pattern you can't see when you're inside. The public conversation about AI in Spanish works, in many of its corners, like a folk religion of badly understood technologies. And folk religion has its logic, its professionals, its temples and its rents. What it doesn't have is any relation to the user's initial question: "what is this and what actually happens when I use it?".
What would an honest debate look like?
Let's imagine for two minutes what an honest debate about AI in Spanish would look like. It would have a few identifiable features.
The broadcasters would declare their conflicts of interest at the start of the text, not buried on a transparency page. "This article is written by X, who holds Y investments in Z". That alone would change half the conversation.
The formats would change. Fewer listicles, more long articles. Fewer alarm headlines, more deep reporting. Fewer "5 advantages" and "10 dangers", more concrete cases told with verification and nuance. When an outlet covered a technical advance, it wouldn't just translate the press release: it would read the paper, talk to unaffiliated authors, contrast with the competition.
The voices brought in would change. Crawford, Zuboff, Gebru, Bender, Hinton, O'Neil translated and cited regularly. Reference Anglo journalism imported into Spanish with credit and context, not as anonymous translation. Specialist Spanish academics — who exist — writing in the general media with enough space. Working professionals — doctors, teachers, judges, journalists, technicians — describing what they see in the field, not just the CEOs of the big firms talking to Vanity Fair.
And the conversation itself would change. Instead of swinging between "AI will save the world" and "AI will destroy everything" — two forms of the same magical thinking — it would happen more slowly, more concretely. We'd discuss corporate concentration with AI Index figures. Cognitive debt with MIT data. Language homogenisation with the Science Advances study. Precarious annotators with the reports already published. The energy cost with real figures. And the disagreements would be about what to prioritise, how to regulate, how to intervene, on what timescales. Not about whether AI is good or bad, which is an empty question.
The way out isn't centrism
It's worth clarifying what the way out isn't. It isn't centrism. It isn't equidistance between the two empty accounts. It isn't saying "well, there are risks and there are advantages, in their right measure". That formula is exactly the operation the corporate ecosystem needs: that the reader be left content with the sensation of having taken a balanced position without having looked at anything. Centrism is the last listicle, the one that sits in the centre and enumerates three advantages and three dangers to end up saying nothing.
The way out is to step off the frame the two accounts provide. Get out of the question "is it good or bad?" — a question that can't be answered and serves only to inflate empty conversation — and into concrete questions you can look at with data. How much has the sector concentrated between 2022 and 2026? Figures: AI Index. What effect does intensive assistant use have on measurable cognitive skills? Figures: Kosmyna et al. Who trains the models at the last link, under what labour conditions? Reports: 404 Media, MIT Tech Review. What hardware does the frontier depend on and where is it made? Data: Epoch AI, Stanford HAI. What documented incidents are there with military or surveillance use? Registry: AI Incidents Database.
Those questions have answers. The answers aren't cinematic. They don't fit in a headline. They aren't solved in four minutes of reading. But they lead to something the current public conversation doesn't offer: understanding. And understanding isn't sold or bought. It's built, slowly, reading the right people in the language you can read them in.
The public conversation about AI is broken because the breakage benefits the whole ecosystem. Repairing it isn't the ecosystem's job — it gets paid for the breakage. It's the reader's job. And it starts, I suspect, by no longer clicking the first link.
Quick definitions
- Judge and party: a situation in which a single player produces both the content that warns of a risk and the product sold as the solution to that risk. It defines the current corporate editorial ecosystem on AI.
- Bipolar traffic capture: an SEO strategy consisting of simultaneously occupying the search positions related to fear ("dangers of AI") and to enthusiasm ("advantages of AI") from the same broadcaster, ensuring the user arrives whatever their initial mood.
- Magical thinking: a mental frame in which a complex phenomenon is treated as an almost-personal entity with vague but immense powers, before which rituals of invocation and protection are developed that have no operational relation to the real phenomenon. Applied here to the folk imagination about AI.
- Third corridor: a metaphor used in the article to name the space of quiet transformation already happening (concentration, cognitive delegation, homogenisation) and that the two dominant accounts (apocalypse vs salvation) don't name.
Referencias
- Stanford HAI (2026). AI Index Report 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report
- Reuters Institute (2025). Digital News Report 2025. https://reutersinstitute.politics.ox.ac.uk/digital-news-report/2025
- Kosmyna, N. et al. (2025). Your Brain on ChatGPT. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/
- Doshi, A. & Hauser, O. (2024). Generative AI enhances individual creativity but reduces the collective diversity of novel content. Science Advances.
- Postman, N. (1985). Amusing Ourselves to Death. Viking.
- Zuboff, S. (2019). The Age of Surveillance Capitalism. PublicAffairs.
- Crawford, K. (2021). Atlas of AI. Yale University Press.
- Thompson, B. Stratechery. https://stratechery.com
- Newton, C. Platformer. https://platformer.news
También te interesa
- Why the critiques of AI aren't made by real experts
- The real dangers of AI
- Why reading IBM on AI is a waste of time
No comments yet
No comments yet. Be the first.
Leave a comment