An AI stock screener should start from evidence, not price momentum. The most underused evidence is free: every public company's 10-K sits on SEC EDGAR, and its AI mentions can be counted by anyone. Anyone can do it. Almost nobody does. Price charts show what happened. Filings show what a company is actually doing.
The method is simple. Count AI mentions per filing. Compute the year-over-year change. Compare against R&D spend. Flag the ones where mentions spike while R&D stays flat. That is the whole trick, and it works. It takes an afternoon with a spreadsheet, or you can use a database that already did it.
GPTTicker ran this on 499 S&P 500 companies across 988 filings and added an AI-wash score separating real AI investment from hype. A screener whose inputs are filings, not opinions. The score is a heuristic, not a verdict, and every input is visible for you to challenge.
Sources: SEC EDGAR
Screen on filings and the hype filters itself out.