Thoughtful, Confused, or Untrustworthy: How Text Presentation Influences Perceptions of AI Writing Tools

TL;DR

This paper investigates how AI text presentation speed influences user perceptions of AI writing tools across creative and professional tasks. Using a within-subject online experiment with five text-appearance styles and two genres, the authors measure reading comfort, perceived quality, anthropomorphism, and adoption attitudes in a sample of 297 participants. The key finding is that medium-speed text (about 600 wpm) optimizes comfort and perceived quality, while nonstandard speeds reduce trust and humanness, with limited interaction from genre. The study highlights that seemingly innocuous interface decisions can bias judgments and usage, underscoring the need for context-aware, configurable designs that mitigate unintended perceptual effects and preserve user agency in writing processes.

Abstract

AI writing tools have been shown to dramatically change the way people write, yet the effects of AI text presentation are not well understood nor always intentionally designed. Although text presentation in existing large language model interfaces is linked to the speed of the underlying model, text presentation speed can impact perceptions of AI systems, potentially influencing whether AI suggestions are accepted or rejected. In this paper, we analyze the effects of varying text generation speed in creative and professional writing scenarios on an online platform (n=297). We find that speed is correlated with perceived humanness and trustworthiness of the AI tool, as well as the perceived quality of the generated text. We discuss its implications on creative and writing processes, along with future steps in the intentional design of AI writing tool interfaces.

Figures (2)

• Figure 1: This figure represents one example study order. For each participant, all marked conditions were randomized. Participants viewed ten scenarios, five in a row for one genre (in this case, beginning with professional), and the remaining five in a row for the second genre (here, creative). The genre order alternated between participants, and within each genre, the sequence of each scenario was randomized for each participant. Each of the five speeds was shown twice (once for each genre), and the sequence of speeds was randomized as well. For each scenario, the questionnaire appears after the AI completion fully appears.
• Figure 2: Survey interface. The viewable text area dynamically resizes according to the width of the display. On mobile, users must scroll down to the end of the text in order to see the "Add AI Suggestion" button. AI text displays in red after the user presses "Add AI Suggestion." The survey appears after the AI text fully appears.