Can AI Predict Stock Prices? What the Research Says
Can AI or ChatGPT predict stocks or beat the market? Not as an autopilot. What CNMV Working Paper No. 96 concludes and how to use AI well.
No: AI cannot predict stock prices and it is not a reliable autopilot for investing. CNMV Working Paper No. 96 concludes that language models make reasoning errors, can use old or invented data and only improve with human supervision and official sources.
Short answer: no. No model knows what a stock will do tomorrow, and AI does not work as an investing autopilot. CNMV Working Paper No. 96 concludes that language models make reasoning errors, can use old or invented data, and only improve with human supervision and official sources. Treat AI as a reading assistant, not a crystal ball.
Key takeaways
- The short answer is no: AI is neither a crystal ball nor an autopilot.
- Primary source: CNMV Working Paper No. 96 (March 2026).
- What AI does well: reading, summarizing and organizing information that you verify.
- An impressive number from one experiment is not a promise of returns.
See the Nvidia analysis in 10 seconds
What you just read, applied to Nvidia. A preview with real data:
- Making money? Yes
- Pricey vs its sector? Cheaper
See the full Nvidia analysis free → Already done: nothing to type.
1. What the primary source says
The paper is titled Large Language Models and Stock Investing: Is the Human Factor Required?. It is Working Paper No. 96 of the CNMV, Spain's securities regulator, written by Ricardo Crisóstomo and Diana Mykhalyuk. It tests several language models with different styles of prompts. The practical messages for an investor are these:
- Models fail. They make reasoning mistakes and can use old or simply invented data.
- Official sources matter. When the model is made to rely on a company's regulatory filings, results improve compared with letting it answer from memory.
- The human factor is still needed. The models need safeguards and validation before being used in financial markets.
The views in the paper are its authors' and do not necessarily reflect the CNMV's. Some news outlets have repeated return figures from the experiment. Do not take them as a promise: a result in one experiment, with its own period and rules, says nothing about what will happen to your money.
2. Why predicting prices is so hard
A stock price already reflects what the market knows and expects. To be right, you would need to know something the market has not yet absorbed, and to do it repeatedly. A language model does not solve that. It generates plausible text from patterns:
- It has no information from the future.
- It can mix data from different dates.
- It answers in the same confident tone whether it is right or inventing.
3. Can AI beat the market?
You will see this question phrased as "does AI beat the market?" or "can AI beat the market?". There is no guarantee for any investor. Even where a study finds an edge under specific conditions, that result does not carry over automatically to your own account once costs, taxes and your own behavior are included. The sensible position is skeptical: use AI to understand and decide for yourself.
4. "Investing automatically with AI": what it means and the risk
Some people look for an AI that invests on its own. Fully delegating the trading adds risks: data errors that become orders, no control when the market changes and nobody accountable. We do not teach how to build a bot that trades alone. The reasonable approach is to use AI to read and organize, and to decide yourself.
Keep reading free.
Create your free account to continue reading “Can AI Predict Stock Prices? What the Research Says”.
No card required
5. What AI can do for you
- Summarize filings you provide and point out disclosed risks.
- Calculate ratios with a formula you can check.
- Explain concepts with examples.
- Raise questions to build your own thesis.
With a real page: the NVIDIA page, with data as of November 9, 2026, shows sales growth of 105.9% and a P/E of 22.6. AI can help you understand why growth is so high. It cannot tell you whether tomorrow's price will rise. And that growth figure is not a prediction: it is what happened in the latest published periods.
6. How to use it with judgment
- Ask about facts from a document you paste, not about the future.
- Ask for the source of each fact and check it (how to verify ChatGPT's financial numbers).
- Distrust any tool that promises to make money for sure.
- Define your own horizon and risk: AI does not know your situation.
- Review the risks of AI stock analysis.
7. How to read a headline that says "AI beat the market"
You will meet headlines claiming that some AI model outperformed the market. Before you let one change what you do, ask these questions:
- Which period? A few good months say little. A result that holds across different market conditions says more.
- Which companies? Were they chosen with hindsight, or defined before the test began?
- Was the data available then? A model can only be tested fairly on information that was public on the date of each decision.
- Were costs included? Fees, taxes and spreads shrink real-world results.
- Who checked it? A single run is an anecdote; independent replication is evidence.
- Does it hold for you? A result for a regulated researcher with a fixed protocol is not a result for your account, your attention span and your emotions.
None of this means studies like the CNMV's are worthless. It means they are inputs to your thinking, not instructions. The most robust finding in that paper for ordinary investors is the one about process: official documents in front of the model, and a person responsible at the end.
Frequently asked questions
Can AI predict stocks?
No. AI can help you read and organize information, but it has no knowledge of the future and cannot predict prices. Treat any forecast as a guess written with confidence.
Can ChatGPT predict stock prices?
No. ChatGPT has no information from the future and can give old or invented figures. Any prediction it offers is a guess phrased with confidence.
Does AI beat the market?
There is no guarantee. CNMV Working Paper No. 96 suggests that relying on official documents and human supervision improves results, but that is not a promise of returns for any investor.
Can I invest automatically with AI?
It is not advisable to leave trading to an AI without control. Use AI to understand and decide yourself.
Continue with the AI guide, AI vs human stock analysts and the AI hub. This article is educational and is not investment advice.
Learn to analyze companies, one a week
Each week we explain a real company in simple words. Create a free account and try the analysis on a company you choose.
See it with real companies
Each of these pages shows the company's dated figures and how it compares with its sector:
Turn what you read into your first analysis.
Create a free account and run your first full analysis in a minute. No card needed.
- How it makes money
- Valuation scenarios
- What could go wrong
First full analysis free · No card
Keep learning
- AI vs Human Stock Analysts: Who Does What
- Risks and Biases of AI Stock Analysis
- How to Verify ChatGPT's Financial Numbers
- Semiconductor stocks
- Software and cloud stocks
- Internet and media stocks
- All Fintols articles in English
An article from the Fintols blog, adapted for US investors. Educational content, not investment advice.