How to read an AI stock signal, and what it can’t tell you

An AI signal is a statistical model’s best guess about where a price might go over a fixed period, based on patterns in past data. It is not a recommendation, and it is often wrong. Used well, it is a prompt to ask better questions about a stock.

The parts of a signal

• Direction: BUY, SELL or HOLD. HOLD usually means the model sees no clear edge, which is the most common and most honest answer.
• Horizon: the period the call covers, such as one day, one week or one month. A one-day signal says nothing about next year.
• Confidence: how strongly the model prefers its answer over the alternatives. High confidence is not the same as being right.
• Target and stop-loss: price levels scaled to the stock’s recent volatility, useful for thinking about risk before you act.
• Top factors: the inputs that pushed the model most, such as momentum, distance from the 52-week high or strength against the NIFTY 50.

Why most signals should be ignored

Markets are close to efficient over short periods. Models built on price history alone often perform only slightly better than a coin flip, and sometimes worse once trading costs are counted. Any service that promises consistent profits from signals is either mistaken or selling something.

Judge the model, not the call

A single correct call proves nothing. What matters is the full record: every signal published, scored after its horizon ends, including the misses. Look for hit rate, average return after costs, and whether results hold up across many stocks and months.