Enterprise AI Analysis: Deconstructing Persuasion in Political vs. AI-Generated Communication

Executive Summary: The AI Persuasion Blueprint

A groundbreaking study by Garassino et al. provides critical insights into how Large Language Models (LLMs) like ChatGPT 3.5 handle persuasive and potentially manipulative language compared to human politicians. By analyzing "presuppositions"implicit assumptions embedded in languagethe research reveals that AI doesn't just mimic human communication; it develops its own distinct, and often less nuanced, persuasive style.

For enterprises, this is a pivotal discovery. The research shows that out-of-the-box LLMs tend to be more repetitive, rely on generic slogans (e.g., "we must build a better future"), and use implicit persuasion more frequently but less diversely than their human counterparts. While AI can generate seemingly on-brand content, it may lack the strategic subtlety of human communication and overuse certain patterns, potentially diluting brand voice or even appearing inauthentic.

This analysis from OwnYourAI.com breaks down the paper's findings into actionable strategies for your business. We explore how to move beyond generic AI outputs and develop custom-tuned language models that align with your specific communication goalswhether in marketing, internal leadership, or customer engagement. Understanding these nuances is the key to unlocking true ROI from your AI investments. Ready to build an AI that communicates with purpose? Book a Consultation.

Deconstructing the Research: What Are Potentially Manipulative Presuppositions (PMPs)?

The core of the study by Garassino and his team revolves around a linguistic concept vital for any enterprise using AI for communication: presuppositions. Think of a presupposition as information presented as "already known" or "taken for granted," which allows a speaker to subtly guide an audience's assumptions.

For example, the phrase "We will continue our commitment to innovation" presupposes that a commitment to innovation already exists. This is a powerful persuasive tool. The study focuses on Potentially Manipulative Presuppositions (PMPs), which are subjective or unproven claims disguised as established facts.

Key Concepts for Enterprise AI Strategy:

• Linguistic Triggers: These are the specific words or phrases that create the presupposition. The study found two are dominant:
  • Definite Descriptions: Using "the" or other determiners to imply existence (e.g., "the solution to our industry's problem" implies a single, known solution exists).
  • Change-of-State Verbs: Verbs that imply a prior state (e.g., "continue," "stop," "regain," "build").
• Discourse Functions: This is the strategic goal behind using a PMP. The paper identifies four main functions in political speech:
  • Stance-taking: Expressing an opinion or position on an issue.
  • Criticism: Attacking a rival or their policies.
  • Self-praise: Highlighting one's own achievements.
  • Praise-of-others: Appreciating supporters or allies.

Understanding these building blocks is the first step toward creating an AI that not only generates text but communicates with strategic intent. A custom AI solution from OwnYourAI.com can be fine-tuned to use (or avoid) specific triggers and functions to match your brand's voice and ethical guidelines precisely.

Key Findings Re-imagined for Enterprise Strategy

The study's quantitative results reveal a clear "fingerprint" for both human and AI communication. For businesses, these patterns are not just academicthey are direct indicators of how out-of-the-box AI might represent your brand and where custom solutions are needed to refine its voice.

Finding 1: Frequency - AI Is More Repetitive with Implicit Persuasion

The research found that, on average, texts generated by ChatGPT contained a higher density of PMPs than the speeches from human politicians. This suggests that LLMs, in an attempt to be persuasive, may overuse these implicit techniques, leading to content that feels formulaic or less genuine.

Finding 2: Form - Humans Prefer Subtlety, AI Prefers Action-Oriented Slogans

The type of PMP trigger used also differed significantly. Politicians favored Definite Descriptions, a more subtle way to frame reality. In contrast, ChatGPT heavily relied on Change-of-State Verbs, often creating repetitive, action-oriented slogans like "build a better future."

Finding 3: Function - AI Avoids Conflict, Humans Embrace It

While both groups used PMPs for "Stance-taking," the distribution of other functions was telling. Politicians frequently used PMPs for Criticism of opponents. ChatGPT, likely due to its safety training, almost completely avoided this function, resulting in a much less confrontational and more uniform style.