Four Things to Check Before Trusting AI Stock Analysis
Verify the sources, dates, numbers, and reasoning in AI stock analysis. Match links to cited claims and recalculate figures to catch errors in units, quarters, and diluted share counts.
Check separately whether an AI analysis's sources exist, whether they actually support its claims, and whether its calculations are correct. Fluent writing or numerous links do not guarantee an accurate distinction between facts and estimates.
A working link is not the same as supporting evidence
Open the filing links provided by the AI yourself and check the company name, reporting period, and table location. A real page can still be irrelevant or contain figures from another year. If you cannot find a quoted sentence in the original, do not end verification by merely asking the model to explain again; inspect the relevant table in the original document. Even an accurately named official document can be connected to a claim incorrectly. It is better to leave claims without disclosed sources unverified and avoid using them as inputs to your next judgment.
Recalculate the figures in a small table
First copy the original currency and units, quarterly versus cumulative basis, consolidated versus separate basis, and actual versus forecast labels. Then recalculate growth rates and margins with a calculator. For earnings per share, also check that the scope of net income and the share-count basis match. Even if AI displays its calculation steps, it may repeat the same incorrect inputs, so separate input verification from arithmetic checks. If a result is wrong, identifying which claims depend on it helps establish what needs revision rather than immediately discarding the entire article.
Check the reasoning and what is missing
If AI jumps directly from revenue growth to improved cash flow, it may have omitted receivables and inventory. Summarizing multiple sources can also mix preliminary results with final results. U.S. investor guidance explains that AI output may be inaccurate, rely on outdated information, or invent information. Do not use a confident tone as a measure of reliability. Ask it to identify opposing scenarios and unknowns, then check those answers against public information too.
Worked example
Hypothetical example: Suppose a filing lists revenue as 12,000 in units of KRW million, but AI summarizes it as KRW 1.2 trillion. The correct conversion is KRW 12,000 million = KRW 12 billion. The incorrect figure is 100 times the actual amount. With operating profit of KRW 1.2 billion, the actual operating margin is 10%, but using the incorrect revenue gives 0.1%. One unit error changes the interpretation of both scale and profitability. After correcting the original figure, recheck the growth rates, multiples, and conclusions calculated from it.
What to check
- Have you located the relevant company, period, and table within each linked source yourself?
- Have you preserved the original units and cumulative basis?
- Have you checked inputs separately from arithmetic?
- Have uncertainty and opposing evidence remained in the summary?
Sources and primary materials
This is educational material about research methods. For an actual company decision, verify the filings and primary sources for the relevant period.