Skip to main content

    How to use ChatGPT for stock analysis: what it does well, where it breaks and how to check it

    To use ChatGPT for stock analysis, let it explain, summarise and structure, but never treat its memory as a data source. Ask for sources and periods, check the key figure yourself and always request the counter-argument.

    Peakwise Research · · Updated · 4 min read

    Contents

    To use ChatGPT for stock analysis, give it the jobs language models are good at, such as explaining a business, summarising a report you paste in and structuring the arguments for and against a stock, and never treat its built-in knowledge as a source for current numbers. Then check its work: ask for sources and reporting periods, verify the one figure your view rests on and always ask for the counter-argument. Used that way, ChatGPT is a fast and patient research partner. Used as an oracle, it is a confident source of outdated figures.

    What ChatGPT is genuinely good at

    • Explaining a business. "How does Visa make money, and who carries the credit risk?" has a stable answer that a model explains well.
    • Explaining concepts. Free cash flow, operating leverage, why banks are valued on tangible book value, at your level.
    • Summarising documents you provide. Paste the outlook section of an earnings release and ask for the three changes since last quarter. The model works from your text, not its memory.
    • Structuring your thinking. Bull case, bear case and the three questions that decide between them.
    • Finding what you forgot. "What risks am I missing for a European chemicals company?"

    Where ChatGPT stock analysis breaks

    Current figures

    A model's built-in knowledge stops at its training cutoff. If an answer did not come from a web search in that conversation, prices, P/E ratios and revenue figures may be months or years old, or a blend of periods that never existed together.

    Precision that looks like research

    "Gross margin of 46.2 percent" reads like a fact. Sometimes it is, sometimes it is a plausible guess. The format tells you nothing.

    Consistency

    The same question with slightly different wording can shift the emphasis. That is how sampling from a language model works. If an answer would change your decision, rerun it and compare.

    Your situation

    The model does not know your holdings, time horizon or tolerance for losses. Any answer to "should I buy this?" ignores what matters most for you.

    A five-step check routine

    1. Ask for sources and dates. Add "name the source and reporting period for every figure" to your prompt.
    2. Open two sources. The investor relations page and the latest report settle most questions. How to read an earnings report shows where the numbers sit.
    3. Check one number by hand. Usually revenue growth or a valuation multiple. If it is wrong, distrust the rest.
    4. Ask for the counter-argument. "What is the strongest case against this?"
    5. Write your own conclusion. One paragraph in your words.

    Prompts that work better than "Should I buy X?"

    • "Explain how {company} makes money, split by segment, using the latest annual report. Cite the report."
    • "Which three figures would tell me whether {company}'s growth is slowing? Where do I find each?"
    • "Write the strongest bear case for {company} in five sentences, then say which points the bull case would dispute."
    • "Here is the outlook section of the latest release: [paste]. What changed versus the previous quarter?"

    Our list of 50 ChatGPT prompts for stock analysis covers every research step. German readers will find prompts in ChatGPT für Aktien.

    ChatGPT vs a tool built for investing

    TaskChatGPTSpecialised tool
    Explaining conceptsStrongStrong
    Current financialsOnly with search, check the periodBuilt on market data
    Source linksVariesUsually standard
    ValuationAd hocStructured with method
    Portfolio viewManualBuilt in

    The Peakwise investment chatbot runs a web search for every question and lists the sources it used, so step 1 of the routine is already done. It answers in the language you write in and lays out arguments and risks without telling you what to buy. For a full picture of one company, the AI stock analysis builds a report with financials, news, analyst consensus, a fair value range with its method and risks. Peakwise is independent and not affiliated with OpenAI, and it can make mistakes too, so the routine stays the same.

    This guide is for education and is not investment advice.

    Frequently asked questions

    It can explain business models, summarise documents you give it and structure a bull and bear case. Whether its numbers are current depends on whether it searched the web for that answer, so ask for sources and open them before you rely on any figure.

    No. A chatbot does not know your goals, holdings or risk tolerance, and it can be wrong about the facts. Use its answer as a starting point for your own research.

    Ask for one thing at a time, ask for the source and date of every figure, and ask for the strongest argument against the idea. Our list of 50 prompts follows that pattern.

    No. Peakwise is an independent product and is not affiliated with OpenAI. Its investment chatbot runs a web search for every question and lists the sources it used.

    Ask about any stockwith sources you can check

    The investment chatbot runs a web search for every question and links the sources it used.

    Free for 7 days. Not investment advice.