Does confident AI tone make users skip fact-checking?

Pavel Samsonov, writing for Nielsen Norman Group, argues that generative-AI chat interfaces actively discourage users from checking whether a response is correct. The outputs "mimic certain attributes that we associate with authoritative sources": "Their tone is unerringly confident, regardless of the accuracy of the response. They are grammatically correct and meticulously formatted." Through the halo effect, "users' positive perception of one attribute of the LLM's responses causes them to have a positive predisposition about all its other attributes — including accuracy." The interface compounds this: "nothing about the chat interface (a plain text box) acknowledges that the LLM's outputs might not be entirely accurate outside of a small label," and "chatbots typically encourage users to move on from the response by prompting them with a new question" rather than toward verification.

Samsonov frames this as a design failure, not a user failure: "It is irresponsible to blame the users for misinformation generated by LLMs. People are efficient (not lazy)." Users adopt genAI tools for the promise of speed, and "will not go out of their way to check the work if the effort of doing so does not seem proportional to the AI's likelihood of being wrong" — a likelihood the interface gives them no signal to revise upward. The confidence extends to users skipping verification entirely: Samsonov cites intern cover letters that left in the LLM's own sign-off, "Let me know if you need anything else!", as evidence that users often don't read the output end to end before sending it.

This extends Does polished AI output trick audiences into trusting it? from multi-modal artifacts — graphs, diagrams, presentations — to the plain-text chat response, and names the specific bias (the halo effect) driving the substitution, plus the interface-design failure to counteract it. It also gives a design-practitioner's causal account for a pattern Do LLM explanations actually help people judge answer quality? measured directly: that controlled experiment found explanation presence alone raises reliance on wrong answers; Samsonov's essay locates the same dynamic in tone and formatting rather than explanations specifically, and adds that the chat UI's ask something else affordance actively steers users away from the source-checking behavior the other study found protective.

The essay is a design argument built on professional anecdotes — lawyers citing nonexistent cases, scientists citing hallucinated papers, doctors trusting erroneous diagnoses, interns leaving in sign-offs — not a controlled measurement of the halo effect's size or of how much it changes verification rates. It does not test any specific interface remedy beyond noting that the existing disclaimer label is not enough. The claim is diagnostic: it identifies a mechanism and assigns responsibility to interface design, but leaves open how much a redesigned chat interface would actually change checking behavior.