For the past few months, we’ve heard a million and one things that ChatGPT can do. From writing poems to fixing your code (and telling you what you did wrong), this generative AI truly promises to be a groundbreaking tool. But can it actually help people make better stock market decisions? In a recent study, a team of researchers argues the answer could be ‘yes’.

Before we go any further, let’s make one thing clear. No one model or algorithm can predict the stock market with accuracy. If such a tool were developed, let’s just say global markets would be in for a wild ride. Instead, AI tools (and especially language models) can only be assistants. They offer some clues, but not certainty.

So what can ChatGPT actually do for stock investors?

Investing in stocks is not only about cold hard facts. Sometimes, it’s about how a certain company is perceived. Savvy investors try to stay ahead of the curve and gain an edge in the markets by looking at news and market trends — this is called sentiment analysis. In sentiment analysis, investors try to gauge the collective mood or sentiment of investors towards a particular asset, market, or the financial market as a whole based on available textual data. This is useful in everything from CFD investing to bond investing.

This sentiment analysis is what Alejandro Lopez-Lira and Yuehua Tang, both from the University of Florida, looked at. There are already dedicated AI tools for this. For instance, Morgan Stanley’s AI models look at things ranging from financial statements to social media comments and then they use this to look at patterns that can be used to derive future stock prices. Except the researchers didn’t use a dedicated tool — they used ChatGPT.

ChatGPT can determine whether a headline is good, bad, or irrelevant to a firm’s stock price. The researchers used this insight and then computed a score for multiple sources. They then correlated this with stock market movements, finding a correlation between the score and the future stock price.

Notably, this didn’t work for earlier, less advanced AI language models.

“We use ChatGPT to assess whether each headline is good, bad, or neutral for firms’ stock prices,” the researchers write in the article. “We document a significantly positive correlation between ChatGPT scores and subsequent daily stock returns. We find that ChatGPT outperforms traditional sentiment analysis methods. More basic models such as GPT-1, GPT-2, and BERT cannot accurately forecast returns, indicating return predictability is an emerging capacity of complex language models.”

“Overall, our results suggest that incorporating advanced language models into the investment decision-making process can yield more accurate predictions and enhance the performance of quantitative trading strategies.”

It’s important to return to that mental note that this doesn’t mean predicting the market. This means analyzing information that can be relevant to a stock price. In particular, monetary policy decisions (like interest rates) or other major changes can have significant effects on financial markets. To a lesser extent, news and other bits of information can also provide valuable insights.

Researchers also suggest other ways in which ChatGPT could offer financial guidance. For example, it can identify patterns and stocks that appeal to someone’s values, preferences, or risk attitudes. Basically, you can tailor your portfolio to your specific profile — for free, no specialized broker required.

Of course, there are no guarantees. ChatGPT can be really good at sentiment analysis and highlighting stocks that fit with a particular strategy in principle, but it was not designed for this. Furthermore, there’s a lot of potential bias in the data, and this is insufficiently explored.

The integration of advanced language models like ChatGPT in the realm of stock market investment is a fascinating development, offering a fresh perspective on how we can analyze and interpret vast amounts of textual data. However, it’s crucial for investors to approach these findings with a dose of caution. While ChatGPT’s capability in sentiment analysis is noteworthy, it should be treated as an assistant rather than a definitive guide.

The stock market’s intricacies are influenced by a myriad of factors, many of which are unpredictable and extend beyond the scope of any AI models.

You can read the paper here, in its entirety.