Paper: arXiv 2501.01763

Authors: Lennart Ante, Aman Saggu

Abstract

Following an analysis of existing AI-related exchange-traded funds (ETFs), we reveal the selection criteria for determining which stocks qualify as AI-related are often opaque and rely on vague phrases and subjective judgments. This paper proposes a new, objective, data-driven approach using natural language processing (NLP) techniques to classify AI stocks by analyzing annual 10-K filings from 3,395 NASDAQ-listed firms between 2011 and 2023. This analysis quantifies each company’s engagement with AI through binary indicators and weighted AI scores based on the frequency and context of AI-related terms. Using these metrics, we construct four AI stock indices-the Equally Weighted AI Index (AII), the Size-Weighted AI Index (SAII), and two Time-Discounted AI Indices (TAII05 and TAII5X)-offering different perspectives on AI investment. We validate our methodology through an event study on the launch of OpenAI’s ChatGPT, demonstrating that companies with higher AI engagement saw significantly greater positive abnormal returns, with analyses supporting the predictive power of our AI measures. Our indices perform on par with or surpass 14 existing AI-themed ETFs and the Nasdaq Composite Index in risk-return profiles, market responsiveness, and overall performance, achieving higher average daily returns and risk-adjusted metrics without increased volatility. These results suggest our NLP-based approach offers a reliable, market-responsive, and cost-effective alternative to existing AI-related ETF products. Our innovative methodology can also guide investors, asset managers, and policymakers in using corporate data to construct other thematic portfolios, contributing to a more transparent, data-driven, and competitive approach.

Complexity vs Empirical Score

  • Math Complexity: 4.0/10
  • Empirical Rigor: 6.0/10 (capped from 7.5: transaction costs not modelled; no out-of-sample test; see penalty flags)
  • Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 7.5)

Why this score: The paper relies on established NLP and statistical methods like TF-IDF without introducing novel mathematics, but it is highly data-driven, analyzing thousands of filings, constructing indices, and validating results through event studies and risk-adjusted performance metrics.

Research Flowchart

  flowchart TD
  A["<b>Research Goal</b><br>Quantify firm AI engagement using<br>objective, data-driven methods"] --> B["<b>Methodology & Data</b><br>NLP analysis of 10-K filings<br>2011-2023, 3,395 NASDAQ firms"]
  B --> C["<b>AI Classification</b><br>Binary indicators & weighted<br>AI scores based on term frequency/context"]
  C --> D["<b>Index Construction</b><br>AII, SAII, TAII05, TAII5X<br>4 thematic indices created"]
  D --> E["<b>Validation & Testing</b><br>Event study on ChatGPT launch<br>Comparison vs. 14 AI ETFs & Nasdaq"]
  E --> F["<b>Key Findings</b><br>Higher AI engagement = superior returns<br>NLP indices outperform existing ETFs<br>Cost-effective, transparent methodology"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a new, objective, data‑driven NLP approach to classify AI stocks using 10‑K filings and builds AI stock indices.

Source sentence
This paper proposes a new, objective, data-driven approach using natural language processing (NLP) techniques to classify AI stocks by analyzing annual 10-K filings from 3,395 NASDAQ-listed firms between 2011 and 2023.

Assumes 1

  • All AI mentions in the latest annual filing are assumed to have equal impact for the equally weighted AI index. model
    Source sentence
    The Equally Weighted Artificial Intelligence Stock Index (AII) serves as a baseline measure, incorporating all companies that mention AI in their latest (i.e., last year's) annual filings, regardless of the extent of their communication.

Data

Backtest sample: 2019-06 – 2023-09 · NASDAQ-listed stocks · equities · daily

  • Annual 10‑K filings · SEC EDGAR text-news public
    2010‑2022 · 3,395 NASDAQ‑listed firms
    Source sentence
    we leverage natural language processing (NLP) techniques to analyze annual 10-K filings from 3,395 NASDAQ-listed firms spanning 2010 to 2022
  • Daily closing prices for stocks, ETFs, IXIC, SPX · S&P Capital IQ market-prices commercial
    2010‑2022 (sample period) · 3,395 NASDAQ stocks, 14 AI‑related ETFs, Nasdaq Composite, S&P 500
    Source sentence
    we obtained the historical daily closing price and Central Index Key (CIK) from the S&P Capital IQ

Recognised sources: SEC EDGAR Bloomberg S&P Capital IQ Refinitiv / LSEG (Datastream, Eikon, Tick History)

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
Notably, the AI‑themed ETFs exhibit expense ratios significantly higher than 0.25%, positioning them at the higher end of the ETF pricing spectrum. Our analysis does not reveal a positive correlation between expense ratios and returns

Finds 3

Reported performance (AI stock indices (AII, SAII, TAII05, TAII5X)): Sharpe 0.04 · max drawdown 4.7%

  • AI indices built with the NLP‑based methodology deliver higher average daily returns and superior risk‑adjusted performance (higher Sharpe and Sortino ratios) than existing AI ETFs, with comparable volatility. performance
    Source sentence
    Overall, our AI indices constructed through an NLP-based methodology outperformed existing AI ETFs by delivering higher average daily returns and demonstrated superior risk‑adjusted performance, as evidenced by higher Sharpe and Sortino Ratios.
  • The AI indices perform on par with or surpass 14 AI‑themed ETFs and the Nasdaq Composite in risk‑return profiles, market responsiveness, and overall performance, achieving higher returns without increased volatility. performance
    Source sentence
    ETFs and the Nasdaq Composite Index in risk‑return profiles, market responsiveness, and overall performance, achieving higher average daily returns and risk‑adjusted metrics without increased volatility.
  • Event‑study on ChatGPT launch shows firms with higher AI engagement earn significantly larger positive abnormal returns. predictive-accuracy
    Source sentence
    we validate our methodology through an event study on the launch of OpenAI's ChatGPT, demonstrating that companies with higher AI engagement saw significantly greater positive abnormal returns

Checked by others

Supported by

  • The Cross-Section of Stock Returns and AI Exposure (2026)
    On the question: Do AI‑engagement signals predict abnormal stock returns around AI product launches?
    Compare the findings

    This paper: Event‑study on ChatGPT launch shows firms with higher AI engagement earn significantly larger positive abnormal returns.

    That paper: Around frontier‑provider model releases, high‑AI‑beta firms earn a 1.9% five‑day abnormal return, 1.1% after FF5 and momentum adjustment.

    Both studies find positive abnormal returns: A reports significantly larger returns for high AI‑engaged firms; B reports a 1.9% five‑day abnormal return (1.1% after FF5+momentum).

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.