AI Prompts for Stock Valuation (2026) | DCF, Multiples, Reverse DCF
Ten AI prompts and one mega prompt for DCF, WACC, reverse DCF and peer multiples, plus the context limits and stale-data traps to watch.
Why AI prompts for stock valuation DCF
A DCF needs about a dozen inputs: five years of revenue, EBIT, D&A, capex, working capital change, tax rate, diluted shares, net debt, and four market inputs. A chatbot can organize these faster than a blank spreadsheet, but it answers from training data or whatever you upload, not from a live feed.
OpenAI's published context windows as of September 2026 are 27K tokens for Instant models on Free, 54K on Go, Plus and Business, and 128K on Pro, with reasoning models at 256K on Plus and 400K on Pro. These totals include output, so a full 10-K plus a model rarely fits on a lower tier. Free accounts get 3 file uploads per day, while Plus allows up to 80 files every 3 hours.
- ChatGPT tiers in 2026: Free at $0, Go at about $8/month, Plus at $20/month, and Pro tiers above that.
- Pro was split into two tiers in April 2026, at around $100 and $200; some aggregators also list a $500 tier.
- Claude 2026 plans: Pro at $20/month ($17 billed annually), Max 5x at $100/month, Max 20x at $200/month.
- Claude Research mode is available on Pro and above, not on Free; Deep Research is included with ChatGPT Plus and above.
- Michael Fritzell (Asian Century Stocks) uses Claude Projects with saved prompts for comparables tables and bull/bear reviews.
- Citadel launched an internal AI Research Assistant to generate customized reports for its research desks.
The mega prompt
Paste your ticker, timestamped price, five-year 10-K table, share count, balance sheet items, WACC inputs and growth view into the prompt below. You should get labeled inputs, a WACC check, a five-year unlevered free cash flow table, terminal value with its share of enterprise value, a reverse DCF, a 5x5 sensitivity grid and a peer cross-check.
Treat the output as a scaffold and recompute every figure in a spreadsheet before relying on it.
ROLE You are a senior equity research analyst building a discounted cash flow (DCF) valuation. Show every assumption, label every input, and flag anything you cannot verify. INPUT Use only the data below. If a value is missing, write NOT PROVIDED and ask for it. Do not invent figures. - Company and ticker: [TICKER] - Valuation date and current share price, with source and timestamp: [DATE], [PRICE] - Last five fiscal years of revenue, EBIT, D&A, capex, change in working capital, and effective tax rate, copied from the 10-K: [PASTE TABLE] - Diluted shares outstanding: [NUMBER] - Cash, debt, lease liabilities, and minorities: [NUMBER EACH] - Risk-free rate, equity risk premium, beta, pre-tax cost of debt, and target capital structure, each with source and date: [VALUES] - My five-year revenue growth and margin view: [ASSUMPTIONS] - Peer tickers for cross-checking: [PEER LIST] INSTRUCTIONS 1. Restate every input and tag it as provided, derived, or assumed. 2. Compute WACC, then compare it with a reasonable range for this sector and explain any gap. 3. Project unlevered free cash flow for five years and show the formula for each line. 4. Calculate terminal value by perpetuity growth. Cap terminal growth at [MAX G, e.g. 2.5%]. Report the share of enterprise value that comes from terminal value, and the implied exit EV/EBITDA multiple. 5. Reverse DCF: holding WACC constant, solve for the revenue growth rate or steady-state margin that the current price implies. 6. Build a five-by-five sensitivity grid for WACC and terminal growth, showing implied value per share. 7. Cross-check against peer median EV/EBIT and P/E, and explain any divergence. 8. List the three inputs most likely to be wrong and how to verify each one. FORMAT - Use headings for steps 1 to 8. - Present projections and the sensitivity grid as tables. - End with a short verdict on what the current price implies, a confidence level, and a 'Verify before using' checklist.
10 prompts for AI prompts for stock valuation DCF
Each prompt below covers one step of the workflow, from input audit through red-team review. Replace every bracketed field with your own figures and their as-of dates before running it, and do not accept a number the model supplies from memory.
Input Audit Checklist
You are a careful financial data reviewer. Using only the figures I paste below for [TICKER] from the [FISCAL YEAR] 10-K, list every input needed for a DCF: revenue, EBIT, D&A, capex, working capital change, tax rate, shares, net debt, and market inputs. Mark each as present, missing, or inconsistent, and flag any figure that does not reconcile with the others. Do not fill gaps from memory; instead, give me the exact filing line item to pull for each missing value. [PASTE DATA]
WACC Sanity Check
Act as a valuation specialist. I will give you my WACC inputs for [TICKER]: risk-free rate [RF], equity risk premium [ERP], beta [BETA], pre-tax cost of debt [KD], tax rate [TAX], and debt weight [D/V]. Recalculate WACC step by step, then judge whether the result is plausible for a company in the [SECTOR] industry with [RISK PROFILE]. Name the single input that drives most of the result, and show how WACC changes if that input moves by one percentage point. Do not look up live market data; state the date of each input I supply.
Terminal Value Trap Test
You are a skeptical valuation reviewer. Given my forecast of free cash flow of [FCF SERIES] for [TICKER], a WACC of [WACC], and a terminal growth rate of [G], calculate the terminal value under the perpetuity growth method and under an exit multiple of [MULTIPLE]. Report what share of enterprise value comes from terminal value. Then flag three traps: growth above nominal GDP, terminal-year capex below depreciation, and a margin that is not sustainable. State whether the two methods agree and what that says about the assumptions.
Reverse DCF Solver
Act as an equity analyst running a reverse DCF. Holding WACC at [WACC] and terminal growth at [G], find the five-year revenue compound annual growth rate that justifies the current share price of [PRICE] for [TICKER], given a starting revenue of [REVENUE] and a steady-state free cash flow margin of [MARGIN]. Show the solution path, then express the result in plain language: what the market is pricing in for growth and margins. Compare the implied growth with the company's last three years of actual growth and with [PEER] consensus, and note any gap.
Peer Multiple Cross-Check
You are a comparables analyst. For [TICKER] and these peers: [PEER LIST], I will paste each company's price, shares, net debt, EBIT, net income, and a forward estimate with its date. Calculate EV/EBIT, P/E, and EV/sales for each, and the peer median and interquartile range. Then state where [TICKER] trades relative to the median and whether its growth, margin, and return on invested capital justify the difference. Flag any peer whose fiscal year-end or accounting basis differs enough to distort the comparison.
Sensitivity Table Builder
Act as a modeling assistant. Using my base-case free cash flow series [FCF SERIES], net debt [NET DEBT], and diluted shares [SHARES] for [TICKER], build a five-by-five table of implied value per share. Rows should step WACC from [LOW] to [HIGH] in equal increments; columns should step terminal growth from [LOW G] to [HIGH G]. Highlight the cell closest to the current price of [PRICE]. Show the formula used so I can reproduce it in a spreadsheet, and state which cells are unreliable because terminal growth is at or above WACC minus two points.
Stale Data Audit
You are a data-quality auditor. I will paste the market and financial figures I plan to use for [TICKER]: [LIST FIGURES WITH DATES]. For each one, state its as-of date, its likely source type (filing, exchange feed, consensus, or model estimate), and whether it is stale relative to today, [DATE]. Flag any number that mixes fiscal periods, currencies, or share classes. Do not supply replacement values from memory. Instead, list the exact source and field I should check to confirm each figure.
Revenue Scenario Builder
Act as a senior analyst building scenarios for [TICKER]. Starting from trailing revenue of [REVENUE] and the segment mix [SEGMENTS], write three five-year revenue paths: bear, base, and bull. For each, state the specific drivers (volume, price, new products, market share), the assumed margin path, and the evidence I would need to see to believe it. Assign probabilities that sum to 100%, and compute probability-weighted value per share using the outputs of my DCF of [VALUE BULL/BASE/BEAR].
Capex and FCF Normalizer
You are a cash-flow analyst. Using the five-year history of operating cash flow [OCF], capex [CAPEX], D&A [D&A], and working capital changes [WC] for [TICKER], normalize free cash flow for one-off items and cyclical swings. Show reported free cash flow and normalized free cash flow side by side, and explain each adjustment. Then state the maintenance capex estimate and whether the terminal-year reinvestment rate is consistent with my assumed growth of [G] and return on new capital of [ROIC].
Valuation Thesis Red Team
Act as a hostile reviewer of my valuation. My thesis for [TICKER] is [THESIS], and my DCF gives a value of [FAIR VALUE] against a price of [PRICE]. Argue the strongest case against my result in five numbered points, each tied to a specific input from my model. For each point, state which number would have to change, by roughly how much, for the conclusion to flip. Finish by listing the one piece of evidence that would most quickly confirm or refute my thesis.
ChatGPT or Claude vs Fintwit
The chatbot workflow handles structure and narrative; a dedicated platform typically handles the data layer and repeatable screens. The contrast breaks down along four lines.
- Live prices: a chatbot answers from training data or uploads, while a platform with a connected market-data feed can show a timestamped quote.
- Screening: a chatbot evaluates only the tickers you paste into the session, while a platform can run one screen across its full coverage universe.
- Persistence: a chatbot session resets, while a platform can keep saved models with version history and locked assumptions.
- Arithmetic: a chatbot generates valuation math as text, while a platform computes it in code and exposes the formulas.
- Lineage: chatbot output carries no source line by default, while a platform can attach a source and date to each input.
| Feature | ChatGPT Plus ($20/mo) | Fintwit ($9/mo) |
|---|---|---|
| Real-time market data | No | Yes — live prices & signals |
| AI stock analysis | Manual prompts required | Automatic for any stock |
| AI portfolio health check | Not available | Built-in with recommendations |
| Curated AI stock picks | Not available | Daily AI top picks & ETFs |
| Stock screener | Not available | Full screener with filters |
| Trending stocks from X | Not available | Real-time signal extraction |
| Earnings & dividend calendar | Not available | Built-in calendars |
| Brokerage integration | Not available | Connect Robinhood, Schwab, etc. |
| Price | $20/mo | $9/mo |
Where ChatGPT or Claude falls short
The limits below are documented in research on LLM stock analysis and in vendor guidance on financial data. Each one can change a valuation without any visible error in the output.
- Stale data: an LLM's knowledge has a shelf life of months or years, and telling the model today's date does not supply live prices (Valyu, DEV Community).
- Fabrication: an arXiv study (2603.19944) found that time-sensitive prompts often return outdated, inconsistent or fabricated figures, sometimes mixing fiscal periods or firms.
- Ratio misreading: the same study documented models treating lower P/E or P/B values as negative, which distorts peer comparisons.
- Arithmetic: Wisesheets warns that math performed during text generation can introduce errors and recommends computing valuation outside the model.
- Quota and plan changes: Free uploads cap at 3 per day, and Research or Deep Research access can change with plan pricing.
Pro tips
Use the steps below to keep the model in the interface role and the spreadsheet in the calculation role.
- Write an as-of date beside every price, risk-free rate and share count before pasting, and ask the model to repeat each date in its output.
- Trim the 10-K to the five-year rows the prompt needs so the input stays inside the 54K Plus window.
- Compute WACC, the DCF and the sensitivity grid in a spreadsheet, then paste the results back for narrative and consistency checks.
- Keep terminal growth at least two points below WACC; the sensitivity prompt flags any cell above that threshold as unreliable.
- Save the input audit and red-team prompts in a Claude Project with a fixed persona and a brevity instruction, as Fritzell does for recurring reviews.