Analyst recommendations have become so tainted that buy side professionals have long treated a ‘Hold’ recommendation by a sell side analyst as a ‘Sell’ recommendation in disguise. Yet, in academic studies about the influence of analysts on share prices, we still tend to use analyst recommendations and earnings forecasts as the key input. This is mostly because up until recently, we couldn’t systematically ‘read’ and ‘quantify’ analyst opinions as expressed in the words and phrases they use in their research reports. Thanks to AI that has changed, with interesting results.

Shikun Ke from Yale has unleashed AI to ‘read’ more than 1.1 million sell side research reports from January 1998 to the end of September 2023. The charts below show what subject matter analysts pay most attention to.

Over the whole sample, more than 40% of the discussions in analyst reports focuses on profitability, which is what they should do, if you ask me. But note in the chart below that the average time spent discussing profitability in any given report is declining steadily. Meanwhile, a discussion of the financial conditions of a company (things like debt ratios) is increasingly prominent. Note also that during recessions, the amount of time spent on discussing profitability declines while analysts focus more on the financial stability of a company. Similarly, discussions of growth opportunities or the quality of management decline significantly in a recession. These are discussions to be had when times are good.

Average attention to different topics

Source: Ke (2025)

All of that isn’t much of a surprise. Analysts are paid to focus on what matters and what matters for the share price changes over time.

But what I really find interesting is that the words and phrases analysts use predict future earnings revisions and forecast changes. Analysts tend to foreshadow a change of mind and assessment in the text and that can be used to make money.

Starting in 2009, Ke used the previous five years of reports to train the AI on the relationship between analyst words and phrases and future share price returns. Then he used these results to read the analyst reports of the current quarter and predicted share price returns for the next twelve months. The chart below shows the performance of a long/short strategy that invests in the 10% stocks with the highest predicted share price return and shorts the 10% stocks with the lowest forecast return. This long/short strategy is compared to a similar long/short strategy based on analyst consensus price targets in IBES and the market overall. Clearly, there is significantly more forecasting power in examining what analysts write than just looking at their price targets or earnings forecasts.

Returns of trading strategies based on analyst textual analysis and price targets

Source: Ke (2025)