Stock screening and stock analysis are two completely different jobs. Most traders either confuse them or try to do both with the same tool – which is how you end up doing both poorly.
Screening is a filter. It narrows thousands of stocks down to a manageable shortlist based on quantitative criteria you define. Analysis is what happens after that – the deeper work of understanding whether the shortlisted companies are actually worth owning, at what price, and why.
AI is excellent at analysis. It is not a screener. It cannot scan thousands of live stocks in real time, apply your filter criteria, and hand you a results list. That’s what dedicated screening platforms are built for. The power of AI in fundamental research comes after the screener has done its job.
The right workflow: screener first, AI second.
Why the Distinction Matters
A screener like Finviz or Koyfin pulls from a live database of thousands of US-listed stocks and applies your filters instantly. P/E below 20, revenue growth above 20%, positive free cash flow – it runs that across the entire universe and returns a list of matches in seconds. No AI tool can replicate that function because no AI chat tool has access to a live, constantly updating financial database.
What the screener cannot do is tell you which names on that list are actually high quality. A stock can pass every quantitative filter and still have a deteriorating margin profile, a debt structure that’s about to become an issue, or a business model that’s losing competitive ground. The screener sees the numbers. It doesn’t interpret what those numbers mean together.
That’s where AI comes in. Once you have the screener output – 15 stocks, 20 stocks, whatever the list is – you bring it to your AI and ask it to do the analytical work that the screener can’t: rank by combined quality, identify red flags, flag the names that deserve a deeper look versus the ones that passed the filter by coincidence.
What Fundamental Data to Export and Paste
When you export from Finviz or Koyfin, you want to pull the columns that actually speak to business quality – not just the ones the screener used as filters. A well-constructed export gives the AI enough dimensionality to separate genuine quality from screening coincidences.
The columns worth including:
Ticker and company name – obvious, but necessary so the output maps back to real stocks.
Sector – critical for context. A P/E that looks expensive in one sector may be a bargain in another. Including this column lets the AI consider sector-relative norms without introducing external data.
Market cap – context for size and liquidity. A $400 million company and a $40 billion company with the same P/E are in very different risk categories.
P/E ratio – the starting valuation reference. Meaningful only in sector context, which is why the AI analysis step matters.
Forward P/E – the market’s expectation of next year’s earnings relative to the current price. More relevant than trailing P/E for growth-oriented analysis.
Revenue growth (year-over-year) – how fast the top line is expanding. For growth screening, revenue growth above 15% year-over-year is a typical minimum threshold to separate genuine growers from businesses expanding in line with inflation.
Gross margin – the percentage of revenue left after direct costs. This tells you how efficiently the business converts sales into profit at the most fundamental level. High and stable gross margins suggest pricing power and a defensible cost structure.
Operating margin – the percentage of revenue left after both direct costs and operating expenses. It reveals how much of the gross profit survives overhead, R&D, and selling costs. A wide gap between gross and operating margin signals high expense intensity that warrants scrutiny.
Free cash flow margin (or free cash flow per share if margin unavailable) – how much cash the business generates as a percentage of revenue after capital expenditures. A positive margin means the company is genuinely self-funding. For technology stocks specifically, a free cash flow margin above 15% is generally considered strong; below that starts raising questions about how durable the profitability is.
Debt-to-equity – total debt relative to shareholder equity. A common conservative starting filter is debt-to-equity below 1.0, but this varies enormously by sector. Capital-intensive industries like utilities or telecoms routinely carry ratios of 2.0 or higher, while software companies often carry almost no debt. Let the AI flag out-of-norm leverage rather than applying a one-size-fits-all filter at the screening stage.
For momentum-sensitive ranking: If you want the AI to identify a fundamental momentum leader (accelerating growth, estimate revisions), add one or both of these columns:
- Quarter-over-quarter revenue growth
- 3-month earnings estimate revision (%)
This set of columns gives AI enough dimensionality for a meaningful quality ranking. More columns add noise. Fewer columns leave out context that changes the interpretation of individual metrics.
How to Prompt AI to Rank Screener Results
The prompt structure for screener analysis is straightforward once the data is formatted as a clean table.
“Act as a fundamental analyst reviewing a stock screener output. I’ve run a screen for [describe your criteria] and the results table is pasted below.
[Paste the table]
Based only on the data in the table I’ve provided:
(1) Rank the stocks from strongest to weakest on overall fundamental quality – considering revenue growth, free cash flow strength, margins, and valuation relative to the growth rate together, not as isolated metrics. If a sector column is present, weigh each metric relative to what is typical for that sector, but do not introduce any external data.
(2) Identify the top three names that represent the strongest combination of growth quality and valuation – and state specifically what in the data supports each inclusion.
(3) Flag any names where a strong headline metric is offset by a significant weakness in another dimension – name the specific offset.
(4) Identify any names I should remove from further consideration based solely on the data provided – and state the disqualifying condition.
Do not introduce external information about these companies. Base the entire ranking on the table I’ve pasted.”
The constraint at the end – base the analysis only on what’s in the table – is not optional. Without it, the model may reach into its training data and pull in company information that’s months or years out of date. You’re asking for a ranking of the data you’ve verified, not a recall of what the model learned about these companies during training.
Pro-Tip – Level Up with a Weighted Composite Score: While a qualitative ranking is a great first pass for triage, you can upgrade this step by forcing the AI to calculate an exact mathematical score relative to the sector median. This produces a fully auditable, reproducible ranking that also feeds directly into cross-sector comparisons. For the complete scoring framework and prompt, see our guide on How to Compare Companies Across Sectors Using AI.
A Finviz Screen Walkthrough – From Filter to AI Analysis
Here’s what this looks like in practice.
A trader runs a Finviz screen on the S&P 500 with three criteria: P/E under 20, year-over-year revenue growth above 20%, and positive free cash flow. The idea is to find companies growing fast but not priced at the extreme multiples that growth stocks often carry – a value-within-growth approach.
The screen returns 14 stocks. The trader copies the results table – ticker, sector, market cap, P/E, forward P/E, revenue growth, gross margin, operating margin, debt-to-equity, and free cash flow margin – and pastes it into the AI with the prompt above.
The AI output identifies three key findings:
The Alpha Leader: A mid-cap healthcare technology company with 28% revenue growth, a 71% gross margin, a forward P/E below the S&P 500 average, and debt-to-equity of 0.3. The combination of high growth, high margin, reasonable valuation, and a clean balance sheet puts it ahead of every other name in the screen across multiple dimensions simultaneously.
The Margin Trap: One company shows strong revenue growth but an operating margin significantly below its gross margin – a sign that operating expenses are consuming most of what the top-line growth is generating. The growth looks impressive in isolation but the profit that actually reaches the bottom line is thin.
The Mismatched Style: Another name shows a low P/E alongside weak revenue growth – meaning the screener has likely captured a value stock rather than a growth stock. The trader needs to decide whether that fits the original intent of the screen.
The AI also recommends removing one name entirely: a company at the top end of the P/E range, with the weakest revenue growth in the screener and a gross margin at the lower boundary of the filter criteria – the weakest across every primary dimension with no compensating strength visible in the data.
The trader is left with three priority candidates worth deeper research. The other 11 are set aside without spending another hour on them. That triage – from 14 names to 3 priority candidates in 20 minutes – is the core value of the screener-to-AI workflow.
What AI Cannot Do Here
This is worth being direct about.
Ask AI to “find me all S&P 500 stocks with revenue growth above 20% and free cash flow positive” and you will not get a reliable answer. You may get a list of company names with plausible-sounding metrics. But those figures are drawn from training data that may be a year or more out of date, and the list is not a live screen – it’s an approximation. Do not act on it.
The correct process: run the screen yourself on Finviz, Koyfin, or TradingView using current live data, export the results, and bring that output to AI for analysis. The screener provides current data. AI provides the analytical layer on top of it.
Building Screening Into the Research Workflow
Fundamental screening isn’t a daily exercise. It’s a periodic process – run when you’re looking to initiate a new position, when a sector rotation analysis has identified a sector worth exploring more deeply, or when you want to systematically survey an area of the market rather than relying on names you already know.
The screening process feeds the deeper research queue. The screen defines the universe. AI ranks the universe by quality. The top names from that ranking go into the deeper analysis – transcript review, ratio contextualisation, peer comparison using the scoring framework from our cross-sector guide, filing detail – that produces the actual trade thesis.
For traders who combine technical setups with fundamental context, the overlap between the fundamental screen output and the technical watchlist is often where the strongest setups live: fundamentally sound businesses that are also setting up technically for an entry.