Gemini and cheap stocks: when the algorithm hallucinates

Why AI shouldn't manage your stock portfolio

Can you trust an artificial intelligence to pick stocks on the Spanish stock market? A user asked Gemini to research cheap stocks and the model produced a detailed analysis of Tubacex, with specific figures: a strategic contract with ADNOC to supply corrosion-resistant alloys, a 115 million euro investment in working capital in 2025, net debt/EBITDA soaring to 3.5x, and a record order book of 1,266 million euros. All very nice on paper.

Skepticism rules: AIs don't invest better than monkeys

The reaction was unanimous: distrust. A simulation showed that nearly all the AIs given $10,000 to manage ended up more than 50% in the red. And when you push back that their recommendations are garbage, they fold: "you're right, here are ten more." That malleability, far from being a virtue, betrays the absence of any judgment of its own. As one analyst recalled, there have been experiments with monkeys throwing darts that produced similar results for 14 years. Using a language model to invest is on par with financial astrology.

Country risk and the diversification fallacy

Another argument that emerges is the wisdom of avoiding the Spanish stock market altogether. Concentrating your job, taxes, savings and future pension in a single country means taking on a country risk that is hard to manage. Whoever asked Gemini probably didn't expect the biggest piece of advice to be diversifying abroad. But that lesson, harsh as it is, is worth more than any algorithmic recommendation.

Conclusion: AI is an excellent text generator but a terrible financial advisor. If you want to lose money, ask Gemini. If you want to keep it, do the opposite of what it tells you. Or better yet, consult a human and, while you're at it, say a little prayer.