How to Use ChatGPT—Free—to Analyze Which Shares to Buy or Sell?

This guide shows practical, repeatable workflows investors can run with ChatGPT at zero cost or for free to decide which shares to buy or sell. 

It focuses on screening ideas, analyzing fundamentals and technicals, modeling scenarios, managing risk, and building an audit trail—while being clear about what ChatGPT can and cannot do. Nothing here is investment advice; use your judgment and verify with primary sources.

What ChatGPT Can Do for Investors (Free Tier)

• Search & verify information on the web: Ask for timely summaries with links to sources (e.g., latest news, management changes, product launches). ChatGPT’s built-in search pulls in citations so you can click through and verify. 
• Analyze files you upload: Drop in CSV/Excel/PDF financials, watch ChatGPT preview tables, and then ask questions (“find revenue CAGR,” “compare margins vs peers,” “chart working capital”). Supported formats include CSV/XLSX/PDF/JSON; Drive/OneDrive uploads may be supported in your app. 
• Build reusable “mini-analysts” (custom GPTs): Create a custom GPT that follows your house style (e.g., “Quality at Reasonable Price screener”) and grants it tools like web search and file analysis. 
• Improve your prompts and workflows: OpenAI’s prompt-engineering guides help you structure instructions so outputs are consistent, auditable, and useful. 

Tip: Keep a single running chat (or a custom GPT) per company so the model remembers your context: tickers, peer set, key drivers, and “what good looks like.”

An End-to-End Research Workflow

1) Generate ideas & build a watchlist

Goal: Surface candidates that fit your strategy (quality, growth, value, dividend, turnaround).

Prompt template:

“You are my equity screener. Strategy: [e.g., profitable compounders]. Markets: [e.g., India large/mid]. Hard filters: ROCE > 15%, sales CAGR 5y > 10%, net debt/EBITDA < 2. Soft filters: consistent promoter holdings, low pledge. Return a table of 15 names with rationale, data fields, and links to sources.”

Then ask:

“Group them into thematic buckets (e.g., domestic consumption, specialty chemicals). Flag red-flags (one-off gains, customer concentration).”

Use Search follow-ups for each candidate’s recent news with links so you can validate quickly.

2) Rapid fundamental check (10–15 minutes per name)

Upload the latest annual report, investor presentation, and results deck (PDF/XLSX). Ask:

• “Extract five-year revenue, EBITDA, EBIT, PAT; compute 3y/5y CAGRs; show margin trend and a quick waterfall chart.”
• “Summarize business model in 5 bullets, and list top 3 unit economics drivers.”
• “List non-recurring items in the last 8 quarters; separate operating vs non-operating cash flows.”

File upload and interactive table views help you drill into rows/columns and then query them in natural language. 

3) Competitive landscape & moat scan

Ask ChatGPT (with search enabled) to map peers, market share talk, capacity additions, pricing power, and customer/supplier concentration. Request inline citations so you can click through, and insist on separating facts (with links) from model’s inferences.

Prompt:

“Build a peer set for [Company]. Return a comparison table: product mix, geographies, price vs value positioning, gross/EBITDA margin ranges, and where each enjoys switching-cost or network effects. Cite every factual claim with a source link; label anything inferred as ‘inference’.”

4) Quality of earnings & accounting red-flags

Upload financial statements; ask for:

• Cash conversion: EBITDA → CFO conversion, working-capital intensity, receivables days vs peers.
• Capital allocation: reinvestment rate, buybacks/dividends vs free cash flow, ROCE trend.
• One-offs: exceptional items, capitalized expenses, related-party transactions.

Prompt:

“From these files, compute FCF for 5 years (CFO – maintenance capex). Contrast with growth capex. Identify any changes in revenue recognition or depreciation policy noted in notes to accounts.”

5) Valuation scenarios (sanity > precision)

Ask ChatGPT to create conservative/base/bull cases with explicit assumptions and to show the math:

• DCF with key drivers (volume/price mix, margin, tax, WC turns, capex).
• Simple multiples cross-check (EV/EBITDA, P/E, P/S) vs peers and vs own history.
• Sensitivity tables (e.g., margin ±200 bps, WACC ±100 bps).

Prompt:

“Build a 5-year three-scenario model for [Company]: drivers listed, assumptions tied to sources or marked ‘judgment’. Output: key line items, FCF, terminal assumptions, EV/EBITDA triangulation, and a 2×2 sensitivity heatmap. Export to CSV.”

Upload/download support and table interactivity make iterating here quick. 

6) Technicals & timing (if you use charts)

Provide CSV of daily OHLCV data (from your broker/official exchange). Ask for:

• Trend regime (moving averages, higher highs/lows).
• Momentum/signal checks (RSI/MACD interpretation, only as a second opinion).
• Volume clusters, support/resistance levels, and risk points.

Prompt:

“From this OHLCV file, compute 20/50/200 DMA crossovers, RSI(14), ATR-based stop. Summarize in a risk plan: entry, invalidation, position size for 1% portfolio risk, and a trailing-stop rule. Include a chart.”

(If your app supports plots in file analysis, ChatGPT can render them after upload.) 

7) News, events, and thesis tracking

Use search to build a live brief:

• Regulatory actions, management changes, product launches, order wins/losses.
• Channel checks from reputable sources (avoid rumor mills).
• Earnings previews and post-results deltas to your model.

Prompt:

“Create a one-page ‘thesis monitor’ for [Company]: bullet current thesis, key risks, KPIs to watch. Then a dated log of news with links, color-coded as + / – / neutral, and a final ‘change vs thesis’ call.”

8) Build your own “Investment GPT” (optional but powerful)

You can package your instructions, preferred metrics, and templates into a custom GPT so every analysis is consistent. Give it access to Search and File Uploads, set prompt starters, and attach model spreadsheets as “Knowledge.” 

Governance tip: keep your custom GPT’s instructions explicit about disclosures, mandatory citations, and a checklist before issuing any “buy/sell/hold” phrasing. See OpenAI’s guidelines for writing instructions to make it robust. 

Concrete Prompts You Can Copy

1. Quick take (5 bullets + numbers)

“In 5 bullets, summarize [Ticker/Company]: business, 5y revenue/EBITDA CAGRs, margin trend, leverage, ROCE, promoter holding/pledge (if India), and 3 key risks. Add links for factual claims.”

2. Peer comp table

“Create a peer comp for [Industry] in [Region]: revenue, EBITDA margin, net debt/EBITDA, EV/EBITDA (ttm), P/E (ttm), 3-yr sales CAGR. Show quartiles and highlight outliers. Cite sources.”

3. Quality of earnings

“From these files, analyze CFO vs EBITDA, WC days, capitalized costs, and related-party transactions. Flag anything that inflates earnings quality. Show a short ‘smell test’ checklist.”

4. Valuation sanity check

“Triangulate value for [Company]: DCF (base case) + peer EV/EBITDA + P/E vs 5-yr median. Give a range and what must be true operationally to justify the top end.”

5. Risk plan

“Propose risk controls for initiating/adding to a position in [Company]: position sizing rule for 1% max portfolio risk, invalidation level, and a catalyst calendar.”

Building an Audit Trail (so you can defend your decisions)

• Keep sources & links: When you use ChatGPT’s search, request inline citations and paste the ones you rely on into your notes. 
• Save files and outputs: Upload your CSVs/PDFs and ask ChatGPT to export cleaned tables back to CSV for your records. 
• Standardize via a custom GPT: Store your checklist, ratio definitions, and thresholds so every company gets judged the same way. 

What ChatGPT Can’t Do (and How to Stay Safe)

• It can be wrong or overconfident. Always click through links, read the primary filings, and re-run the math. Use citations and label inferences vs facts. 
• Not a data vendor. Treat numbers as working notes until you’ve verified them with filings/exchange data.
• No personalized financial advice. It can structure analysis and generate scenarios, but you decide suitability and execution.
• Privacy & compliance. Don’t paste confidential, MNPI, or license-restricted data. If you build custom GPTs or connect actions/APIs, review the relevant Help Center articles and your firm’s policies. 

A Simple “Buy or Sell” Decision Framework You Can Automate

1. Business quality (0–5): moat, durability, industry structure.
2. Financial quality (0–5): ROCE, FCF conversion, leverage.
3. Execution (0–5): management, capital allocation, guidance hit-rate.
4. Valuation vs history/peers (0–5): cheap/fair/expensive.
5. Risk controls: position sizing, invalidation level, catalysts.

Ask ChatGPT to compute the sub-scores from your uploaded data + sourced facts and return a scorecard (with source links). Require a final section titled “What has to go right/wrong” before any Buy/Add/Hold/Reduce/Sell label appears.

Optional: Automate & Scale

Tasks & agents (where available): Set a daily/weekly brief on your watchlist or have an agent gather filings and update your model. Review capabilities and safeguards in the Help Center before using automation. 

Bottom Line

Used well, ChatGPT is a free research assistant: it searches and cites sources, ingests your spreadsheets and PDFs, builds quick comp tables and models, and standardizes your checklist—so your limited time goes to judgment, not data wrangling. Pair it with your broker/exchange data, insist on citations and transparency, and you’ll have a fast, auditable way to decide which shares to buy, hold, or sell.