How We Build Our AI Rating
Every rating is company-specific and relies exclusively on the company's official SEC filings. For every stock, we show the full file: category, rationale, and verbatim quotes with a source link. Not a quality judgment, not a buy recommendation — just a documented topical classification. Source: SEC filings (10-K annual reports and 10-Q quarterly reports).
The Four Categories
AI products or services are a demonstrable revenue source — documented in the product or segment description of the annual report (Item 1 Business, MD&A). Example: a chipmaker whose accelerator division supplies AI data centers.
The risk factors (Item 1A) cite AI as a concrete risk to the company's own business model — and AI is not a revenue source. Generic boilerplate ("AI may pose risks …"), found in nearly every filing, doesn't count: the link to the company's own business has to be identifiable.
The filings document operational AI use — such as efficiency programs or internal tools — but AI is not a revenue source in its own right. Example: a retailer that runs demand forecasting and logistics with machine learning.
No material AI connection in the filings reviewed. That, too, is a finding — we document which filings we checked and that they contained no substantive hits.
The Priority Rule
A company can meet several criteria at once — Walmart, for instance, is a supermarket chain that also uses AI heavily in its supply chain. That's why a fixed order applies, and the first matching category wins:
Sells AI → Threatened → Uses AI → Neutral
Data Basis and Evidence Requirement
- For each company, we review the last four quarterly reports (10-Q) and two annual reports (10-K) — sourced directly from SEC EDGAR, the official database of the US securities regulator.
- Every rating needs at least 2 verbatim quotes from these filings, each in the original English and in German translation, with a link to the source passage.
- If the evidence isn't sufficient, there's no rating — the column then shows "–". We don't guess, as a matter of principle.
- For "Neutral", we document the negative finding: which filings were checked and that they contained no substantive AI passages.
Limits of the Method
- Self-reported: SEC filings are written by the company itself — both overstatement ("AI-washing") and understatement are possible. The quote requirement keeps the basis verifiable at any time, though.
- Time lag: Time passes between business developments, the filing, and our review. Every file therefore carries its rating date and is updated each quarterly season.
- German stocks: Tickers without SEC filings (Xetra-listed stocks, for instance) are, for now, treated as "not ratable".
- Not investment advice: The rating says nothing about a stock's quality or valuation — it only provides a topical classification.
Frequently Asked Questions
What does the AI rating mean?
A purely topical classification of how a company relates to the AI boom: Sells AI, Uses AI, Neutral, or Threatened. It is not a quality judgment and not a buy recommendation.
Where does the evidence come from?
Exclusively from the company's official SEC filings — the last four quarterly reports (10-Q) and two annual reports (10-K). Every rating needs at least two verbatim quotes; for "Neutral" we document the negative finding along with the filings reviewed.
Why does a stock show only a dash ("–")?
That means no file exists yet, or the evidence wasn't sufficient. We don't guess, as a matter of principle — no rating beats a poorly documented one.
Why don't German stocks have a rating?
German tickers don't file SEC reports. Until we've connected an equivalent source, they're treated as "not ratable".
How current is the rating?
Every file carries its rating date and is updated each quarterly season as new filings become available. Some time can pass between a filing and the rating.
Can a company fall into more than one category?
Yes — in that case, the fixed priority rule decides: Sells AI before Threatened before Uses AI before Neutral. The first matching category wins.