A one-sentence stock screen in Claude, and three ways it can quietly mislead you

Claude, a stock screening tool connected to the Equibles MCP server, can transform a one-sentence description of desired stocks into a set of filters and a table, revealing three ways such screens can silently mislead users. On November 3, 2026, the tool identified three potential issues with this approach.

The initial screen sought US stocks valued over $2 billion, profitable, with revenue growth exceeding 20%, gross margins above 50%, priced below 30 times earnings, and daily trading volume surpassing $10 million. This translated to six arguments: minimum market cap of $2 billion, zero minimum net income, 20% minimum revenue growth, minimum gross margin of 50%, maximum P/E ratio of 30, and minimum daily volume of $10 million.

The screen returned 32 stocks, with prices dated to the November 2 close, short interest from October 15 FINRA settlements, and 13F holder counts from the June 30 quarter.

First, running the same screen on a watchlist of Apple, Microsoft, Amazon, Reddit, Palantir, SoFi, Robinhood, and Uber revealed that only Reddit met all criteria. However, Robinhood and Uber failed to meet any filter despite not showing any red flags. This occurred because the screen excludes stocks lacking bounded metrics, and neither has a reported gross margin.

When building or using a screener, it's crucial to distinguish between "no data" and "fails the test" and request the assistant to specify which filter removed each stock.

Second, the same growth percentage can represent different types of growth depending on the industry. Sorting the top ten stocks by revenue growth showed Micron at 379%, Sandisk at 378%, Royal Gold at 115%, Nvidia at 106%, Chord Energy at 84%, TG Therapeutics at 70%, Franco-Nevada at 57%, Western Digital at 56%, Reddit at 53%, and Hinge Health at 53%. While these companies represent diverse industries, the filter cannot differentiate between commodity-focused growth and other types.

Third, the concept of "cheap" is not explicitly defined in the screen. Reddit passed the screen with a trailing earnings multiple of 18.7 times, while early February data showed a multiple of 53.4 times. Despite revenue growth and a lower P/E multiple, the price fell due to search referrals dropping by over a million per week in October.

The CEO explained that daily active users could fall by a low double-digit percentage if this trend continued. Screens can identify cheap stocks, but further analysis is needed to understand why a particular stock is cheap, including comparing valuation to peers, recent historical data, insider sales, and short interest.

The full breakdown of the screening process, including six prompts, answers, and verification methods for each stock, is available at https://equibles.com/blog/how-to-screen-for-stocks-with-chatgpt-or-claude. When handling missing data in personal screens, one can choose to exclude, flag, or attempt to fill in the gaps using additional sources like SEC filings, FINRA short interest, 13F filings, and market prices, all processed through tools like Equibles.