Promptism – When Fluency Becomes Authority with Sune Selsbæk Reitz

Key Takeaways

• Promptism is a growing cognitive vulnerability: users often accept AI-generated responses as truth because of their fluency and confident delivery.
• Processing fluency is mistaken for factual accuracy: Humans prefer easy-to-process information, leading them to question credibility rather than truth.
• AI manipulates the human preference for narratives: AI generates clear text but oversimplifies complexity, eliminating uncertainty and space for doubt.
• The removal of cognitive friction leads to dependence: AI’s validation can become a cognitive crutch, hindering true intellectual development and critical thinking.
• Professional competencies must pivot from answering to judging: As AI generates responses quickly, human value must focus on critical evaluation and logic challenges.
• The “responsibility gap” inevitably falls on the end user: AI’s complexity complicates accountability; users must verify information fully before trusting and sharing.
• Foundational domain knowledge is more critical than ever: Without expertise, it’s hard to judge an AI’s output on unfamiliar subjects; risky to rely on.
• We must intentionally reintroduce intellectual friction: To ensure critical thinking, employ deontological design and encourage adversarial questioning from AI.

Webinar Details

Title: Promptism – When Fluency Becomes Authority with Sune Selsbæk Reitz
Date: 2026-05-06
Presenter: Sune Selsbæk Reitz
Meetup Group: Book Launch with Technics Pub x MWS
Write-up: Author Howard Diesel

Is “Promptism” Influencing Our Acceptance of AI Responses?

Sune Selsbæk Reitz, a data and AI strategist with a background in history and philosophy, introduces the concept of “promptism”. He defines this phenomenon as the pervasive habit of accepting machine-generated responses as factual without engaging in critical examination. This observation stems from a personal experience in which Sune accepted an AI chatbot’s incorrect output simply because it was presented in a clear, structured, and convincing manner. He cautions that this behaviour represents a broader systemic pattern in which users substitute rigorous verification for reliance on the superficial familiarity and fluency of AI-generated text.

Figure 1 Promptism: When Fluency Becomes Authority

Figure 2 Introduction Slide

Figure 3 Reasoning for Writing the Book

Figure 4 Definition of “Promptism”

Is Promptism Undermining Critical Inquiry into Truth?

The cognitive vulnerability to promptism is rooted in “processing fluency,” a psychological concept established by Daniel Kahneman and Amos Tversky in the 1970s. This principle posits that humans are predisposed to believe information that is cognitively easy to process, regardless of its factual accuracy. Sune notes that artificial intelligence systems are engineered to optimise this fluency by eliminating linguistic friction, delivering prose that sounds confident and complete. Consequently, the traditional scholarly methodology—which relies on tracing historical clues and examining structural evidence—is undermined. Users are increasingly discarding the critical inquiry of “Is this true?” in favour of the much less rigorous standard of “Does this sound right?”

Figure 5 How Humans Respond to Fluency

Figure 6 What Sounds Correct is not always Correct.

Figure 7 The Acceptance of Explanations that Seem Clear and Concise

Figure 8 Being Misguided by What Sounds Truthful

Do AI Systems Undermine Critical Thinking and Doubt?

Artificial intelligence systems capitalise on the human cognitive predisposition for structured narratives. Rather than providing fragmented or nuanced data, these models consistently generate resolved text featuring a definitive beginning, middle, and end. Invoking Nassim Taleb’s concept of the “narrative fallacy,” Sune explains that individuals often prefer simplistic, cohesive explanations to the inherently complex, messy nature of reality. Language models excel at producing these clear explanations, systematically stripping away uncertainty and loose ends. Drawing a parallel to totalitarian ideologies that offer all-encompassing explanations, Sune warns that when systems offer complete, instantaneous resolution, the cognitive space for doubt and critical thinking is effectively eliminated.

Figure 9 Large Language Models Respond with Answers that Appear Complete and Concise

Figure 10 Humans are Prone to Accept Simple Truths over Complicated Ones

Figure 11 Reference to Hannah Arendt

Is AI Creating Cognitive Dependence in Users?

A more insidious consequence of AI integration is the technology’s tendency to continuously align with and validate user assumptions. While this artificial agreement promotes operational speed, it fundamentally impedes cognitive development, which relies on intellectual resistance and contradiction to refine ideas. By prioritising speed and ease, artificial intelligence systematically removes cognitive “friction”. Sune draws an analogy to utilising GPS navigation for a familiar route; while the technology reduces immediate mental effort, it simultaneously usurps the individual’s decision-making faculties. When users repeatedly rely on language models to initiate their thought processes, the relationship shifts from mere technological assistance to cognitive dependence.

Figure 12 Falling Victim to the Usefulness and Awe of Generative AI

Figure 13 We Need to Be More Critical

Figure 14 How do You Engage with Generative AI?

How Can We Balance Efficiency and Critical Thinking?

During the interactive segment, attendees highlighted the tension between organisational efficiency mandates and critical thinking. Instead of leveraging AI to reduce working hours, corporations frequently utilise these systems to extract greater output, treating them merely as efficiency engines rather than analytical partners. To preserve critical faculties, one participant shared a methodological approach: triangulating outputs by deploying the same queries across multiple Large Language Models (such as Copilot, Claude, and Gemini). This comparative analysis, coupled with robust prompting frameworks, is essential for maintaining user awareness and mitigating the risks associated with AI hallucinations, particularly when navigating unfamiliar subject matter.

How Does AI Change Responsibility in Decision-Making?

The instantaneous delivery of AI-generated responses fundamentally alters cognitive dynamics by framing problems and directing inquiries before human thoughts are fully conceptualised. In response, Sune argues that professional competencies must pivot from producing answers to critically evaluating them. This paradigm shift exposes a critical organisational vulnerability termed the “responsibility gap”. Because AI outputs are the culmination of training data, developmental parameters, organisational deployment, and user prompts, establishing clear accountability for errors becomes exceedingly difficult. Ultimately, Sune asserts that because the technology lacks a definitive author, the onus of verification and accountability must revert entirely to the end user.

Figure 15 Take Time to Validate the Answer You Have Recieved

Figure 16 It is Easy to Accept the Answer at Face Value

Figure 17 Be More Critical When Using Generative AI

Figure 18 The Social Problem of Promptism

Figure 19 Generative AI Outputs can not be Traced Back to a Single Answer to Validate Accuracy

Figure 20 Take Control of Your Engagement with Generative AI

Figure 21 Slow Down and Take Time to Evaluate the Answer

Figure 22 “An Answer is not an Insight”

Figure 23 QR Code to the Book ‘Promptism’

Is Human Intuition at Risk with AI Reliance?

The discourse subsequently addressed the risk that human intuition and domain expertise would atrophy due to overreliance on artificial intelligence. Sune advocated for a “deontological design” framework, emphasising the need to engineer trustworthy, explainable systems that incorporate mandatory human-in-the-loop interventions. Attendees noted the difficulty of challenging AI outputs without foundational domain knowledge. To counteract the inherent politeness and subservience of language models, one effective strategy discussed was to explicitly prompt the AI to adopt the persona of a critical academic professor. This directive forces the system to interrogate the user’s understanding, thereby reintroducing essential cognitive friction.

How Should We Verify Information in an AI World?

The concluding dialogue scrutinised the provenance of digital information and the subjective nature of truth in an AI-saturated environment. Attendees expressed concern over the social normalisation of prioritising rapid, unverified AI assertions over meticulously curated expertise. Sune explicitly advised against utilising Large Language Models as primary educational tools for entirely unfamiliar subjects, suggesting a continued reliance on traditional literature or constrained systems like NotebookLM that exclusively query designated references. Ultimately, the consensus was that, regardless of the technological tools used for drafting or ideation, professionals must maintain absolute, uncompromising accountability for any information they choose to disseminate.