The worst AI investing prompt is also the most common one: "What stock should I buy?"
Ask that, and the model does exactly what it is built to do. It produces a fluent, confident answer. No 10-K behind it. No earnings call. No comps. Just a guess dressed as a verdict, and you have no way to check it.
The fix is not a better prompt for the same wish. It is a different job.
AI should structure the research process. It should not pretend to pick the stock.
Here is the swap, AI builds and you judge:
-> Stop asking for the answer. Ask for the framework: what would I need to verify before this is even a candidate? -> Have it pull the inputs, not the conclusion. Revenue drivers, margin trend, debt maturities, segment detail. Each with a source and a date. -> Have it surface the disconfirming case. What in the filings argues against the idea, in management's own words. -> Have it flag every gap. "Unclear" is a real output and goes on your follow-up list.
Notice the failure mode you just avoided. A model that answers "buy this" cannot tell you why, cannot cite a page, and cannot be wrong out loud. It just sounds right.
The inputs are AI's job. The assumptions, the valuation, and the decision are yours.
Human judges. AI builds.
Save this and run it on the next name you look at.
Educational content only. Not investment advice, and not a recommendation to buy, sell, or hold any security. Wall Street Prompt. Always verify against the primary source filing.