How to Screen Stocks in Plain English With AI
One copy-paste prompt turns a plain-English thesis into a rigorous, numeric, checkable screen you can run in any screener — plus the one guardrail that keeps AI from inventing the answer. Model-agnostic (Claude, ChatGPT, Copilot, Gemini, Perplexity).
For: Retail investors · analysts · idea generationTo screen stocks with AI, split the job in two. First, use AI to translate a plain-English thesis — say “profitable mid-cap software growing free cash flow but not overpriced” — into explicit, numeric criteria with defensible thresholds (market cap, revenue growth, FCF margin, valuation, leverage). Second, execute those criteria on real data: paste them into an actual screener, or a model with live data access. The mistake almost everyone makes is asking a plain chat model to name the matching tickers directly — it will confidently invent companies, prices, and ratios that don't exist. AI is excellent at defining a screen and terrible at being the database. Define with AI, run on real data, and verify every figure against a primary source before the list becomes a watchlist. The same approach works in ChatGPT, Claude, Copilot, Gemini, or Perplexity.
Natural-language screening is the feature every platform is racing to ship — type “cheap tech stocks with growing dividends” and get a list. It's genuinely useful, and it's also the single easiest way to get burned by AI in finance, because a model with no live data will happily hand you a beautiful screen of stocks that don't match, or don't exist. The fix isn't a better tool — it's a better division of labor. Let the model do what it's great at (turning fuzzy intent into precise, defensible criteria) and put the data step somewhere the numbers are real. Here's the prompt that does the first half correctly.
The workflow, step by step
- Write your thesis in one sentencePlain English is fine — “profitable mid-cap software growing free cash flow but not overvalued.” The prompt does the rigor.
- Run the translation promptIt converts your sentence into explicit, numeric criteria with thresholds you can actually defend.
- Run the criteria on real dataPaste them into a real screener or a model with live data access. Never let a plain chat model invent the matches.
- Get a one-line why + one risk per nameFor each surviving name, ask for a one-sentence reason it matches and the single biggest risk.
- Verify every numberCross-check each metric against a primary source before the list becomes a watchlist. Models fabricate financial figures.
The exact prompt (copy-paste)
Model-agnostic. Replace the [THESIS] with your one sentence.
You are an equity analyst turning an investing thesis into a defensible stock screen. My thesis: [THESIS IN PLAIN ENGLISH]. Step 1 — Translate it into explicit screen criteria: - For each criterion give the metric, the operator, and a numeric threshold (e.g. "FCF margin >= 10%"). - Cover, where relevant: market cap, revenue/earnings growth, margins, valuation, leverage, and liquidity. - State the assumptions behind each threshold in one line. - Flag any part of my thesis that is subjective or hard to measure, and suggest a measurable proxy. Step 2 — Tell me exactly how to run it: which fields to enter in a screener, or what live data you would need. Do NOT list matching tickers unless you have live, verifiable data — if you do not, say so and stop at the criteria. Never invent companies or numbers.
Get all 7 workflows in one pack
This is one of seven paste-ready workflows in The AI Finance Workflows Pack — the exact prompts, pro moves, and honest guardrails finance pros run at their desk. Free.
Instant access on the next page + the free weekly brief. Unsubscribe anytime.
Prefer to browse first? Open the Workflows Pack → · Finance Prompt Library →
Frequently asked
Can I use AI to screen stocks?
Yes, in two ways. First, use AI to translate a plain-English thesis into explicit, numeric screen criteria you can defend — market cap, growth, margin, valuation, and balance-sheet thresholds. Second, run those criteria on real data using a screener or a model with live data access. Do not ask a plain chat model to produce matching tickers from memory — it will invent them. Verify every figure against a primary source. Educational only, not financial advice.
Will ChatGPT give accurate stock screen results?
Only if it is pulling from live data at the moment you ask. A standard chat model answers from training data with a cutoff and can fabricate tickers, prices, and ratios. Use AI to define the criteria, then execute the screen where the data is real, and cross-check the output. Do your own research (DYOR).
What makes a good AI stock screen prompt?
A good prompt forces the model to convert vague language into specific, numeric, defensible thresholds; to state its assumptions; to flag any criterion that is subjective or hard to measure; and to add a one-line rationale and one risk per name. It should also instruct the model to use only data it actually has and to say so when it is guessing. This is not a recommendation to buy or sell any security.