# Are generative AI text annotations systematically biased?

arXiv: https://arxiv.org/abs/2512.08404 · HTML: https://mlsanity.com/paper/2512.08404/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 14 references that are in the map.
- First posted on arXiv 9 Dec 2025.

## Builds on (most cited first, 7 of 7)

- [ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks](https://mlsanity.com/paper/2303.15056/) (2023)
- [Do Prompt-Based Models Really Understand the Meaning of their Prompts?](https://mlsanity.com/paper/2109.01247/) (2021)
- [Challenges and Applications of Large Language Models](https://mlsanity.com/paper/2307.10169/) (2023)
- [Prediction-Powered Inference](https://mlsanity.com/paper/2301.09633/) (2023)
- [Fairness And Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, And Mitigation Strategies](https://mlsanity.com/paper/2304.07683/) (2023)
- [Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation](https://mlsanity.com/paper/2509.08825/) (2025)
- [Surprising gender biases in GPT](https://mlsanity.com/paper/2407.06003/) (2024)

## Built on by (most cited first, 1 of 1)

- [GR-Ben: A General Reasoning Benchmark for Evaluating Process Reward Models](https://mlsanity.com/paper/2605.01203/) (2026)

## Abstract (arXiv)

This paper investigates bias in GLLM annotations by conceptually replicating manual annotations of Boukes (2024). Using various GLLMs (Llama3.1:8b, Llama3.3:70b, GPT4o, Qwen2.5:72b) in combination with five different prompts for five concepts (political content, interactivity, rationality, incivility, and ideology). We find GLLMs perform adequate in terms of F1 scores, but differ from manual annotations in terms of prevalence, yield substantively different downstream results, and display systematic bias in that they overlap more with each other than with manual annotations. Differences in F1 scores fail to account for the degree of bias.