In this article
Today we look at one concrete case that shows how SEO has replaced serious analysis in Spain's public conversation about AI. The company is IBM. When a Spanish user searches for "what is artificial intelligence," the first organic result —not sponsored— on Google is almost always an IBM Think page. The content carries the authority of a century-old brand, neutralized vocabulary, definitions that look competent. So why does it leave the reader worse informed than if they'd read nothing at all? Because it's stripped of context, commercially slanted, and, to a large degree, frozen before the generative revolution of 2022. My take is biased: in another life I worked with IBM services and I know the gap between what IBM declares and what IBM does. Form your own. Open the IBM Think pages on AI and compare them with any recent paper or industry report.
I've had the Spanish AI search SERP open in a permanent tab for two years. IBM's position there is no accident. It's the product of fifteen years of sustained investment in ranking, domain authority and content production. What we're going to take apart isn't the quality of the content itself —which is unremarkable— but the gap between what the search position suggests and what the content delivers.
The short history of IBM in artificial intelligence
International Business Machines has been in computing since 1911. Its historic contribution to the field is enormous: mainframes, the COBOL language, the OS/360 operating system, the original PC of 1981, the Deep Blue supercomputer that beat Kasparov in 1997. For decades, IBM was practically a synonym for corporate computing.
The bet on modern artificial intelligence began with Watson, announced in 2011 after winning the television quiz show Jeopardy! that February. IBM presented Watson as the system that would revolutionize medicine, law, finance. It announced deals with major hospitals —Memorial Sloan Kettering, MD Anderson Cancer Center—. It built a business named Watson Health with the ambition of transforming medical diagnosis through AI.
Reality didn't keep pace with the announcement. Over the decade, Watson Health kept piling up technical criticism. The oncologists at MD Anderson found the system's recommendations to be inferior to the human medical consensus. The project slowed down. Cumulative investment in Watson Health reached, according to estimates by the Wall Street Journal and MIT Technology Review, several billion dollars.
In January 2022, IBM announced the sale of Watson Health to Francisco Partners for roughly one billion dollars, a tiny fraction of what had been invested. In strategic terms, the sale marked the end of IBM's attempt to compete head-on with Google, Microsoft, OpenAI and Anthropic in applying AI to vertical sectors.
IBM didn't abandon the field. It launched watsonx in July 2023 as its new enterprise AI platform, with its own models —the Granite family— and orchestration services. The bet is serious, the enterprise-adoption figures are real, but the competitive positioning is modest. Granite doesn't show up in frontier-model rankings —Chatbot Arena, LMArena, HELM— anywhere near Anthropic, OpenAI, Google, Meta or DeepSeek.
Why its content dominates the Spanish SERP
IBM's share of Spanish organic results for "what is artificial intelligence" or "what is machine learning" or "what is deep learning" comes down to factors that have little to do with content quality.
The first factor is domain authority. IBM.com has been online continuously since 1991 and has spent three decades accumulating an inbound-link profile —from universities, media, other corporations— that no other company in the field can match. For Google, IBM.com is one of the most authoritative domains in the world on any technology query. The search algorithm rewards that authority.
The second factor is persistence. The IBM Think pages on basic AI concepts —"what is AI," "what is machine learning," "what is deep learning"— have been published for years, have been updated only marginally, and have accumulated positive signals continuously over time. A page that has existed since 2017, received links for eight years and been lightly updated every twelve months has, in Google's eyes, a structural advantage over a page published six months ago, however good that newer page may be.
The third factor is investment in SEO. IBM keeps a dedicated search-optimization team with a sizable industry budget. The texts are written with carefully placed keywords, optimal heading structure, calibrated length, dense internal linking, complete metadata. Every technical element is optimized.
The fourth factor is multilingual translation. IBM keeps versions of the same content in dozens of languages, Spanish included. That multilingual presence reinforces the global position and, by extension, the local SERPs.
Add the four factors and IBM holds a structural advantage in search that owes nothing to the quality of its analysis and everything to its track record, its budget and its persistence. The consequence is that the first page a Spanish user reads on any basic AI concept tends to be an IBM page.
The problem with the content itself
IBM's definitions of AI concepts are technically correct in broad strokes. There are no gross errors. But there are three features of the content worth naming.
The first is the neutralized corporate vocabulary. IBM's definitions of "machine learning," "deep learning," "neural networks," "generative artificial intelligence" are written in an aseptic Spanish, without voice, without nuance, without opinion. The text does the bare minimum to be informative without ever saying anything that might make a potential client uncomfortable. It's the exact opposite of analysis with a point of view.
The second feature is the systematic omission. IBM's articles on AI, especially the most-read ones, avoid mentioning their main competitors. OpenAI rarely appears. Anthropic, almost never. DeepSeek, Mistral, Cohere, Stability AI, Meta AI: absent. The field's real benchmarks —MMLU, GPQA, HELM, Chatbot Arena— go largely unmentioned. The current frontier models —GPT, Claude, Gemini, Llama, DeepSeek-R1— aren't named or come up only in very generic passages. The conversation IBM offers the reader is frozen around 2017-2019, before the generative paradigm consolidated.
The third feature is the way Watson and watsonx are positioned as a central reference. IBM's AI pages repeatedly cite their own products as examples of the concepts under discussion. That isn't illegitimate —it's their website— but a reader who arrives via "what is AI" isn't looking for IBM product advertising, they're looking for general context. Receiving advertising dressed up as a definition is an information asymmetry, and the asymmetry is held in place by the SEO position.
The omission that weighs the most
Of the three features, the systematic omission is the gravest. I'll break it down because it has an operational consequence for the reader.
If someone in 2025-2026 wants to understand what current artificial intelligence is, they need to know at least the following. That the current paradigm rests on the transformer models proposed by Vaswani et al. in Attention Is All You Need (2017). That the two leading companies in the field are OpenAI (creator of GPT and ChatGPT) and Anthropic (creator of Claude), with a major presence also from Google DeepMind, Meta AI, Mistral, xAI, and the Chinese labs (DeepSeek, Qwen). That manufacturing the necessary chips depends almost exclusively on TSMC in Taiwan and ASML in the Netherlands. That the main current regulatory debates are the European AI Act (Regulation EU 2024/1689) and the U.S. export controls. That the main ethical debates are corporate concentration, algorithmic bias, data capture and labor impact.
No reader starting from the IBM Think pages on AI will receive any of this. They'll come away understanding vague technical definitions, generic examples, and an industry picture where IBM is a central player and the field's real names barely appear. The gap between what needs to be known and what gets across is brutal.
This isn't down to the anonymous copywriter of the day. It's a structural editorial decision. IBM Think isn't designed to teach AI. It's designed to position IBM as a neutral authority on AI, a claim the real field hasn't backed for a decade.
The consequence for the reader
There's a pattern worth naming. The user who learns about AI starting from Google's first page on the subject comes away with three deficits that take a while to correct.
The first deficit is vocabulary. They learn a corporate Spanish for AI —"digital transformation," "agility," "adoption," "use case"— that belongs to Davos presentations, not the real technical conversation. When they later read Crawford, Russell, Karen Hao, or any technical paper, the vocabulary doesn't fit and it takes them a while to reconcile.
The second deficit is a map of the players. They learn that "companies" are developing AI, with no concrete names. When they later try to follow industry news, they have no mental biography of OpenAI, Anthropic, DeepMind, the relationships between founders, the internal controversies, the capital moves. They're missing the human context that makes any sector legible.
The third deficit is a hierarchy of problems. They learn that AI has "advantages and disadvantages" with no quantitative distinction, no risk hierarchy, no differential urgency between topics. When they later try to form an opinion on regulation or professional adoption, they have no basis for telling the central from the incidental.
All three deficits can be corrected, but it takes time. Most readers' attention economy doesn't allow that time. The aggregate consequence is a poor public conversation about AI, where most non-specialists use a vocabulary and a mental map the real field hasn't shared in years.
The political question
This is personal opinion, but IBM's own trajectory in the field supports it. SEO concentration in search produces, in aggregate, a public conversation dominated by the companies with the biggest budgets and the longest histories, not by the companies with the best current product or the analysts with the best judgment. That distortion is invisible to the average user because the search result looks neutral.
The distortion has political consequences. When politicians, regulators, teachers and non-specialist journalists form their opinion of AI mainly by reading corporate sources like IBM —not out of bad faith, but out of habit and SEO— their analytical frameworks shift in the direction that suits those sources. The European AI Act was debated, in part, on the basis of definitions the regulated companies themselves helped produce.
What I'd ask for —again with no realistic hope— is that basic AI searches return, alongside the corporate result, at least one independent academic or journalistic source. Wikipedia plays that role in part, but it's structurally ranked below the optimized corporate results in Google.
In the meantime, the individual reader can at least know that the first Google result on any contemporary technical AI concept is, almost always, a company with a commercial interest in the answer. Knowing it is the first step toward diversifying sources.
The concrete figure to close on: according to Sistrix's SEO visibility analyses of the Spanish market published in its quarterly reports for 2024 and 2025, IBM.com holds a visibility of around 92% of Google's top 10 for the most common Spanish queries on basic artificial intelligence concepts ("what is AI," "what is machine learning," "what is deep learning," "what is a neural network"). That figure —92%— is the exact distance between Spain's public conversation about AI and the real technical conversation of the field. Until that figure drops, the average reader will keep learning from the least relevant player in the frontier field and missing the players who actually lead it.
Definitions
Domain authority: a metric used by search engines and SEO analysis tools to estimate the relative strength of a website. It's calculated from the inbound-link profile, age, consistency and other factors.
SEO (Search Engine Optimization): the set of techniques for improving a website's position in a search engine's organic results. Applied at scale with sustained budget, it shapes which sources the user sees when searching for information.
Watson: IBM's commercial brand for its AI services, announced in 2011. Its health business, Watson Health, was sold in 2022. Its successor, watsonx, launched in 2023.
SERP (Search Engine Results Page): the results page a search engine returns for a query. The position each result occupies in the SERP determines click probability and, therefore, the user's exposure to that source.
References
Daniela Hernandez, Hospitals Are Reining In IBM Watson's Cancer Ambitions (Wall Street Journal, August 2018). Critical coverage of the technical problems of Watson Health in oncology.
Will Douglas Heaven, IBM has sold IBM Watson Health (MIT Technology Review, January 2022). Analysis of the Watson Health sale and the lessons of the project.
International Business Machines Corporation, Annual Report 2024 (IBM, March 2025). Data on revenue, margins and business segments, including AI.
Sistrix, Visibility Index Spain (sistrix.com, quarterly reports 2024-2025). SEO visibility figures for leading domains in the Spanish market.
Stanford HAI, Artificial Intelligence Index Report 2026 (April 2025). Section on corporate concentration and the share of players in the frontier field.
Ashish Vaswani et al., Attention Is All You Need (arXiv:1706.03762, June 2017). The foundational transformers paper, systematically absent from IBM's pages on contemporary AI.
Center for Research on Foundation Models (Stanford), Holistic Evaluation of Language Models (HELM) (crfm.stanford.edu, continuously updated). Open benchmark that IBM rarely cites in its outreach content.
Further reading
Clayton M. Christensen, The Innovator's Dilemma (Harvard Business Review Press, 1997). The classic framework for understanding why established companies lose their edge to new competitors.
Vili Lehdonvirta, Cloud Empires (MIT Press, 2022). An analysis of how today's cloud oligopoly took shape, with IBM in a secondary position to AWS, Azure and Google Cloud.
Cory Doctorow, The Internet Con: How to Seize the Means of Computation (Verso, 2023). A critical framework on the concentration of power on the internet, including Google's SERP.
Sebastian Mallaby, More Money Than God: Hedge Funds and the Making of a New Elite (Penguin, 2010). A historical framework for how corporate capital moves when a sector pivots.
You might also like
- Why Spanish journalism about AI is very bad
- Why AI master's degrees dominate Google's results
- Can you really learn AI by watching YouTubers?
- Why reading IBM on AI is a waste of time
Comments0
No comments yet.
Leave a comment