---
title: How AI explanations on X can shift political positions
author: Martin Ebner
date: 2026-11-03
url: https://postgenerative.at/indications/2026-11-03/how-ai-explanations-on-x-can-shift-political-positions
lang: en
type: Indication
summary: An Oxford study shows how AI can shift positions when editing and explaining posts. Its investigation of “Explain this post” on X highlights how platform instructions influence Grok’s responses.
tags:
  - Media & Public Sphere
  - Communication
  - Politics
  - Research
sources:
  - title: "University of Oxford: AI tools that can steer public opinion fall outside current regulation, Oxford study finds"
    url: https://www.ox.ac.uk/news/2026-07-21-ai-tools-that-can-steer-public-opinion-fall-outside-current-regulation-oxford-study
  - title: "Tsirtsis et al.: AI-Mediated Communication Can Steer Collective Opinion"
    url: https://arxiv.org/abs/2605.16245
license: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
alternate:
  de: https://postgenerativ.at/indizien/2026-11-03/wie-ki-erklaerungen-auf-x-politische-positionen-verschieben
---

# How AI explanations on X can shift political positions

On July 21, 2026, the University of Oxford reported a study of AI mediating public communication. Writing and explanation tools can subtly alter positions; the researchers identify gaps in existing regulation. [1]

X is a central example: the team replicated “Explain this post” using its published prompt template and Grok’s API. Across 78 abortion-related posts, the generated explanations aligned more often with the original position of anti-abortion posts than with posts supporting abortion rights. [2]

The analysis traces this asymmetry to a platform instruction to challenge mainstream narratives when necessary. The operator’s configuration thus helps shape supposedly explanatory AI. [1]

Network simulations demonstrate potential amplification of such biases. They do not establish that real X users have already changed their opinions. [2]

Read as an indication, this suggests that AI increasingly mediates how people understand each other. Even seemingly incidental explanations become a matter of public responsibility.