Can a chatbot's 'I'm just an AI' disclaimer really work when its chat-bubble, typing-delay design keeps making it feel human?
Can disclaimers alone reduce anthropomorphism if the interface structure reinforces it?
This explores whether adding warnings like "I'm just an AI" can stop people from treating chatbots as human-like when the chat interface itself, with its back-and-forth turns, typing delays, and messaging-app look, keeps nudging them to do exactly that.
This explores whether a disclaimer such as "I'm an AI, not a person" can work against anthropomorphism when the interface around it keeps signaling "you're talking to someone." The corpus suggests a disclaimer is up against more than one force. Research on chat interfaces argues that turn-taking, artificial "thinking" delays and social-media-style message bubbles make an LLM seem human before it has written a word How much does chat interface design contribute to anthropomorphism?. A disclaimer is itself language, so it works on the layer that comes after the interface has already set expectations.
Another finding explains why one fix is unlikely to be enough. A review of twenty-one anthropomorphism risks found that they come from several separate triggers: surface resemblance, task performance, context, and language Do anthropomorphism risks share one root cause or many?. A disclaimer deals with the explicit belief that "this is a person." It does nothing about the pull of a fluent, competent partner who answers you instantly. Nielsen Norman Group's four escalating levels (courtesy, reinforcement, roleplay, companionship) show that users can be drawn in by their own wish to feel understood, even when they know intellectually that it's a machine How far can personality push users toward misplaced AI trust?.
A useful parallel comes from bias research, which the anthropomorphism question doesn't obviously connect to. Persona prompts change how a model's output reads, but the underlying gaps in sentiment between groups stay the same. The prompt shifts where the bias shows up without removing it Can persona prompts actually reduce bias in language models?. Disclaimers may behave the same way: a surface-level message added on top of a system whose deeper design keeps producing the effect. Samsonov makes a related point about trust. Confident tone and polished formatting create a halo effect, and the chat interface adds to it by offering no accuracy warnings while pushing users toward their next question Does confident AI tone make users skip fact-checking?.
The more promising alternative in the corpus is to treat anthropomorphism as a dial designers can turn, not a fixed default. One paper proposes a spectrum that runs from anti-anthropomorphic to human-like designs. At the anti end, metaphors based on what the system is physically made of reveal the human labor, data and infrastructure behind the model Can interface design reduce harmful anthropomorphism of language models?. Shanahan offers a mental model that could replace the disclaimer's flat denial. A dialogue agent is best understood as a character being role-played, so human-style explanations apply to the persona and not to the system underneath Should we treat dialogue agents as role-playing characters?. That framing doesn't ask users to stop seeing a character. It tells them what the character is.
The corpus has no direct experiment that tests disclaimers alone against interface redesign, so the answer is an inference, but a consistent one. Disclaimers probably help at the margins and are unlikely to win against a structure that reinforces anthropomorphism at every turn. The less obvious takeaway is that the main lever is the shape of the conversation itself: its rhythm, its framing, and whether it ever shows you the machinery behind it.
Sources 7 notes
Research argues that turn-taking, artificial thinking delays, and social-media-style messaging conventions reinforce anthropomorphism structurally, operating before any text is generated. This means interface-level changes may be needed alongside language-focused interventions.
A review of twenty-one anthropomorphism risks identifies five analytic categories (epistemic, affective, human agency, normative, societal-institutional) driven by distinct triggers: superficial resemblance, task performance, context, and language. Risks are not reducible to consciousness attribution alone.
Nielsen Norman Group research identifies four escalating levels of anthropomorphism (courtesy, reinforcement, roleplay, companionship) driven by users' innate preference for being understood. Personality and emotional design are named as specific levers that amplify trust risk, especially in task-oriented work where perceived competence should matter more than perceived sentience.
Across three models, persona conditioning makes models follow trait instructions but fails to eliminate underlying bias. Between-group sentiment gaps persist unchanged, showing prompts operate only at the output level.
Samsonov argues that AI's confident tone, correct grammar, and polished formatting create a halo effect that makes users trust accuracy without verification. The chat interface compounds this by offering no accuracy warnings and actively prompting users toward new questions instead of source-checking.
Show all 7 sources
The paper argues anthropomorphism should shift from a fixed interface default to an adjustable design variable. Using materiality-based metaphors across a spectrum can counteract uncritical user trust and expose LLMs' dependence on human labor, infrastructure, and data.
Shanahan's framework treats LLM outputs as character-consistent text production rather than authentic mental states. The dialogue prompt establishes a character; the model generates continuations matching that character, making folk-psychology applicable to the simulated persona, not the underlying system.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- A Taxonomy of Linguistic Expressions That Contribute To Anthropomorphism of Language Technologies
- Anthropomorphism in the age of Large Language Models: An overview of potential risks and mitigations
- Beyond Anthropomorphism: a Spectrum of Interface Metaphors for LLMs
- When "A Helpful Assistant" Is Not Really Helpful: Personas in System Prompts Do Not Improve Performances of Large Language Models
- LLMorphism: When humans come to see themselves as language models
- CompanionSim: Synthetic Data for Evaluating Anthropomorphism in Human-AI Relationships
- The Tool-to-Entity Threshold: Parasocial Dynamics of Personalised AI Agents in Shared Social Spaces
- Does My Chatbot Have an Agenda? Understanding Human and AI Agency in Human-Human-like Chatbot Interaction