How to Do Stock Due Diligence With AI
Last updated June 2026
Short answer
Stock due diligence with AI is a checklist you run before you buy, with AI speeding up each step and you verifying its outputs. Work through six areas: the business model and moat, financial health (debt, cash flow, margins), valuation versus peers and history, management and insider activity, risks and red flags, and how much to own. AI tools like ChatGPT and Claude summarize filings, compute ratios, and argue the bear case, but they state wrong figures confidently, so check every number against the source. A connected assistant like Walnut can reason over holdings you already own, which helps most on the sizing question. AI tools and Walnut are not investment advisers.
“Do your due diligence” is easy advice and vague instruction. This guide turns it into a concrete checklist you can run on any stock, with AI doing the heavy lifting on each step. It is deliberately more rigorous than a quick look at a chart or a headline: the goal is to understand the business, confirm it is financially sound, judge whether the price is reasonable, size up the people running it, hunt hard for what could go wrong, and only then decide how much (if any) belongs in your portfolio. AI can compress hours of this into minutes, but every step ends the same way, you verify what the model told you, because a confident wrong number is the most common failure in the whole exercise.
What stock due diligence is (and how AI fits)
Due diligence is the structured homework you do before committing money to a stock. It is not a single lookup; it is a sequence of checks, each of which can talk you out of the idea. The point is to be systematic instead of falling for a good story, and to reach a position with a clear reason to own it and clear reasons you would change your mind.
AI changes the speed of this work, not the responsibility for it. Concretely, a good assistant can:
- Summarize and navigate. Condense a 10-K, an earnings call, or a shareholder letter and point you to the sections that matter, so you read the primary source with a map.
- Compute and compare. Pull key figures, calculate ratios, and build a peer comparison faster than you would by hand.
- Argue the other side. Lay out the bear case and the top risks explicitly, which is often the most valuable thing it does.
What it cannot do is be the final authority. Models misread figures, confuse fiscal periods, use stale prices, and occasionally invent numbers that sound right. So the rule running through every step below is the same: let AI accelerate the work, then verify its output against the actual filing or a live quote before you rely on it. If you want the lighter, learn-the-company version first, see how to research a stock with AI.
The due-diligence checklist, area by area
Here is the checklist itself. Each area lists what it is, what to actually check, how AI accelerates it, and the honest catch that tells you what to verify. Run them in order: the later steps only matter once the earlier ones hold up.
Business model and moat
Before any number, understand how the company actually makes money and why that is durable. What does it sell, to whom, how does revenue recur, and what stops a competitor from taking it? A moat is the structural advantage (network effects, switching costs, scale, brand, or a real cost edge) that lets a business keep earning above-average returns.
- What to check: How revenue is generated and how concentrated it is by product, customer, or geography; who the real competitors are; where pricing power comes from; and whether the advantage is widening or eroding.
- How AI helps: Ask an AI assistant to summarize the business model in plain language, list the main competitors, and lay out the bull and bear case for the moat, then use that as a map for your own reading of the filings.
- The catch: AI is confident about narratives even when they are dated or wrong, so treat its moat story as a hypothesis to test against the 10-K and recent results, not a conclusion.
Financial health (debt, cash flow, margins)
The part that separates a good story from a solvent business. You are checking whether the company generates real cash, carries a debt load it can service, and keeps margins that hold up over time. Cash flow is harder to dress up than earnings, so it is where careful due diligence spends its time.
- What to check: Free cash flow trend and whether it funds the business; total debt versus cash and versus operating cash flow; interest coverage; gross and operating margin direction over several years; and whether share count is rising through dilution.
- How AI helps: AI can pull the key lines from recent filings, compute ratios, and flag a deteriorating trend faster than reading statements line by line, and it can explain what a given ratio means in context.
- The catch: Models routinely misread a figure, mix up fiscal periods, or invent a number that sounds right, so re-check every ratio against the actual statement before you trust it.
Valuation versus peers and history
A great business at the wrong price is still a bad investment. Valuation asks what you are paying for the earnings, cash flow, or growth you just verified, and whether that is cheap or expensive relative to close peers and to the company’s own history.
- What to check: Common multiples (price to earnings, price to sales, price to free cash flow, enterprise value to EBITDA) compared to a peer set and to the stock’s own multi-year range; and what growth or margin assumptions the current price implies.
- How AI helps: Ask AI to assemble a peer comparison, explain why two similar companies trade at different multiples, and stress-test the assumptions baked into the current price so you are not anchoring on one number.
- The catch: AI often uses stale or wrong current prices and multiples, and it can pick a flattering peer set, so confirm the live figures and sanity-check the comparables yourself.
Management and insider activity
You are buying a team and a capital-allocation record, not just a ticker. This area looks at whether management has delivered on past promises, how they spend cash, how they are paid, and whether insiders are buying or selling their own stock.
- What to check: The track record on prior guidance and targets; capital allocation choices (buybacks, dividends, acquisitions, reinvestment); how incentive pay is structured; and the pattern of insider buying versus selling in public filings.
- How AI helps: AI can summarize recent shareholder letters and earnings calls, surface how compensation is tied to performance, and pull the gist of insider-transaction filings so you know where to look closer.
- The catch: Insider selling has many innocent explanations and AI will over-read it, so use its summary as a pointer to the primary filings rather than a verdict on management’s intentions.
Risks and red flags
Good due diligence spends real time trying to talk itself out of the investment. This is the deliberate search for what could break the thesis: customer or supplier concentration, regulatory exposure, litigation, aggressive accounting, or a secular decline the story is glossing over.
- What to check: The risk factors section of the latest 10-K, pending litigation and regulatory matters, unusual accounting or restatements, dependence on a single customer, product, or supplier, and any gap between the narrative and the numbers.
- How AI helps: Ask AI to argue the bear case explicitly, list the top risks from the filings, and point out where the reported story and the financials disagree, which is a fast way to find the questions you should be asking.
- The catch: AI can both miss a real red flag and manufacture a scary one that does not exist, so verify any risk it raises against the source and do not treat a clean AI summary as an all-clear.
The position-sizing question
Due diligence is not finished when you decide a company is good. The last step is how much, if any, belongs in your portfolio given what you already own. A strong company can still be a poor addition if it doubles a bet you have already made.
- What to check: How the position would sit against your existing holdings and concentration; how correlated it is with what you own; how it fits your goals and risk tolerance; and what would make you trim or exit later.
- How AI helps: A general assistant can reason through sizing in the abstract if you paste in your holdings, and a connected assistant that already sees your portfolio can frame how a new name overlaps with what you own.
- The catch: Sizing depends on your actual portfolio and goals, which a general model cannot see, so its answer is generic unless it is grounded in your real positions, and it is still your decision, not advice.
At a glance
| Area | What to check |
|---|---|
| Business model and moat | How revenue is generated and how concentrated it is by product, customer, or geography; who the real competitors are; where pricing power comes from; and whether the advantage is widening or eroding |
| Financial health (debt, cash flow, margins) | Free cash flow trend and whether it funds the business; total debt versus cash and versus operating cash flow; interest coverage; gross and operating margin direction over several years; and whether share count is rising through dilution |
| Valuation versus peers and history | Common multiples (price to earnings, price to sales, price to free cash flow, enterprise value to EBITDA) compared to a peer set and to the stock’s own multi-year range; and what growth or margin assumptions the current price implies |
| Management and insider activity | The track record on prior guidance and targets; capital allocation choices (buybacks, dividends, acquisitions, reinvestment); how incentive pay is structured; and the pattern of insider buying versus selling in public filings |
| Risks and red flags | The risk factors section of the latest 10-K, pending litigation and regulatory matters, unusual accounting or restatements, dependence on a single customer, product, or supplier, and any gap between the narrative and the numbers |
| The position-sizing question | How the position would sit against your existing holdings and concentration; how correlated it is with what you own; how it fits your goals and risk tolerance; and what would make you trim or exit later |
Which AI tools to use, and how
No single tool does the whole checklist well, so match the tool to the step and always keep the verification habit:
- General assistants (ChatGPT, Claude) for reasoning and summaries. They are strong at explaining a business model, condensing filings, arguing the bear case, and walking through valuation logic. They cannot see your accounts or live prices on their own, so verify every figure and confirm current quotes separately.
- Cited answer engines for fast market context. A tool that links its sources helps you check an earnings result or recent news quickly, but it is shallow on deep fundamentals and does not know your portfolio.
- A connected assistant for the fit-and-sizing step. Most tools cannot see what you own, so the position-sizing question stays generic. Walnut is an AI investing assistant you chat with on the broker you already own: it connects your brokerage through SnapTrade (read-only by default) and lets you reason over your real holdings through Claude, ChatGPT, or a built-in assistant, each framed against the S&P 500, and you approve every trade.
For a wider view of what is out there, see the best AI investing tools and the best AI stock research tools roundups.
Where AI helps most, and where it fails
Being clear about both sides keeps you out of trouble. AI is genuinely strong at the breadth-and-speed parts of due diligence and genuinely weak at the parts that require live accuracy and your specific context.
- Helps most: summarizing long filings, drafting a bear case, explaining an unfamiliar concept or ratio, and building a first-pass peer comparison you then check.
- Fails often: live prices and current multiples, exact figures pulled from statements, distinguishing real red flags from imagined ones, and anything that depends on knowing your actual portfolio.
- The discipline: treat every specific number as unverified until you have seen it in the filing or a live quote, and treat every risk it raises as a question to investigate, not a verdict.
The position-sizing step, grounded in what you own
The last step is the one most tools cannot help with, because it depends on your real portfolio. A company can clear every earlier check and still be a poor addition if it doubles a bet you have already made or adds to a sector you are heavy in. Sizing is where due diligence meets your actual holdings.
A general model can reason about this only if you paste your positions in, and even then it is working from a snapshot. A connected assistant sees the real thing. Walnut connects your existing brokerage through SnapTrade, reasons over what you actually own, and frames how a new name overlaps with your holdings, with each position shown against the S&P 500. It is read-only by default and you approve every trade, and because broker feeds rarely pass cost basis it frames returns as window returns rather than realized profit and loss, and says so. Walnut is not an investment adviser; the sizing call, like every other step, is yours.
The bottom line
Due diligence is not a vibe, it is a checklist: business and moat, financial health, valuation, management and insiders, risks and red flags, and position sizing. AI makes each step faster, summarizing filings, computing ratios, building comparisons, and arguing the bear case, but it does not make you right. The whole discipline is verification: check every figure against the source, confirm live prices independently, and treat AI’s risks as questions to answer. Use general assistants like ChatGPT and Claude for the reasoning, and a connected assistant like Walnut when the question is how a name fits the holdings you already own. AI tools and Walnut are not investment advisers.
Try Walnut on top of your broker
Walnut connects any major US broker in a few clicks, then lets you reason over what you actually own through Claude, ChatGPT, or its built-in AI, with each position framed against the S&P 500. Read-only by default; you approve every trade.
FAQ
How do you do stock due diligence with AI?
Run a checklist and use AI to speed up each step: understand the business model and moat, check financial health (debt, cash flow, margins), compare valuation to peers and history, review management and insider activity, hunt for risks and red flags, and decide position sizing. AI drafts summaries, pulls figures, and argues the bear case, but you verify every output against the primary filings. AI tools and Walnut are not investment advisers.
What is stock due diligence?
Due diligence is the disciplined homework you do before buying a stock: confirming how the business makes money, whether it is financially healthy, whether the price is reasonable, who is running it, what could go wrong, and how much of it you should own. It is different from a quick look because it is a rigorous, repeatable checklist meant to talk you out of a bad idea as much as into a good one.
Can AI do due diligence for me?
AI can accelerate almost every step, summarizing filings, computing ratios, building peer comparisons, and arguing the bear case, but it cannot do due diligence for you. Models state wrong figures confidently, misread fiscal periods, and use stale prices. Treat AI as a fast research assistant whose work you check against the source, not as a decision-maker. The judgment, and the accountability, stay with you.
What is the difference between due diligence and researching a stock?
Researching a stock is the broader activity of learning about a company. Due diligence is the rigorous, checklist-driven version you run before committing money: business and moat, financials, valuation, management, risks, and sizing, each verified. If you just want to get up to speed, see how to research a stock with AI. If you are about to buy, run the full checklist on this page.
Which financial metrics matter most in due diligence?
No single metric decides it, but a few carry weight: free cash flow and whether it is growing, debt relative to cash and to operating cash flow, interest coverage, gross and operating margin trends, and whether share count is rising from dilution. Valuation multiples (price to earnings, price to free cash flow, enterprise value to EBITDA) matter only next to peers and the company’s own history. Read them together, not in isolation.
How do I check a company’s moat with AI?
Ask an AI assistant to explain the business model in plain language, name the real competitors, and lay out the bull and bear case for the moat (network effects, switching costs, scale, brand, or cost advantage). Then test that story against the latest 10-K and recent results. The AI narrative is a hypothesis to verify, not a conclusion, because models are confident even when their moat story is dated or wrong.
Can AI read financial statements and filings?
Yes, AI can summarize 10-Ks, earnings calls, and shareholder letters and pull key lines quickly, which is genuinely useful for finding where to look. But it regularly misreads a figure, confuses fiscal periods, or invents a number that sounds plausible. Use it to navigate and summarize, then confirm every specific figure against the actual filing before you rely on it.
How does AI help with valuation?
AI can assemble a peer comparison, explain why two similar companies trade at different multiples, and stress-test the growth and margin assumptions implied by the current price, which keeps you from anchoring on one number. The catch is that models often use stale or wrong live prices and can pick a flattering peer set, so confirm current figures and sanity-check the comparables yourself before drawing a conclusion.
Can an AI tool see my portfolio for position sizing?
Most cannot. General assistants like ChatGPT and Claude have no view of your accounts and can only reason about sizing if you paste your holdings in. Walnut connects your existing brokerage through SnapTrade (read-only by default) so the chat can reason over what you already own and frame how a new name overlaps with your holdings, each framed against the S&P 500. It is still your decision, and Walnut is not an investment adviser.
Is it safe to rely on AI for due diligence?
It is safe to use AI as a research accelerator, but not as a source of truth. The failure mode is a confident, wrong figure or a fabricated risk, so the whole discipline is verification: check every number against the filing, confirm live prices independently, and treat AI’s bear case as a list of questions to investigate. Used that way it saves hours; used as an oracle it will eventually mislead you.
How does Walnut fit into stock due diligence?
Walnut is an AI investing assistant you chat with on the broker you already own. It connects your brokerage through SnapTrade (read-only by default) and lets you reason over your real holdings through Claude, ChatGPT, or a built-in assistant, with each position framed against the S&P 500. It helps most on the fit and sizing question, how a name sits against what you own, and you approve every trade. Walnut is not an investment adviser.
What should I never skip in due diligence?
The risks and red flags step and the position-sizing step. It is tempting to fall for a good story and stop once the business and numbers look attractive, but the whole point of due diligence is to argue the other side and to check that a new position does not overload a bet you already have. AI is good at drafting the bear case; verify what it raises and decide sizing against your real portfolio.
Walnut is informational and is not an investment adviser. App features, pricing, and availability change; verify current details on each provider's site before deciding. Nothing on this page is a recommendation to buy, sell, or hold any security or to use any particular product.