When a chatbot gets personal, people trust it more and worry more about privacy, and the two climb together.

This explores whether personalization makes people trust a chatbot more and worry about their privacy more, both at once. The corpus says yes, and the more interesting finding is that the two effects grow together over time. A longitudinal study found that personalization raises trust and the sense that the bot is humanlike, and it also raises privacy concerns and users' expectations Does chatbot personalization build trust or expose privacy risks?. Each good interaction raises the bar, so a later slip-up feels like a bigger betrayal. Single-session studies can't see this. Relationship effects with chatbots also fade as the novelty wears off Do chatbot relationships lose their appeal as novelty wears off?, so lab results from one sitting don't tell you much about how personalization plays out over months.

The privacy tension isn't only about what the system stores. Personalization works by getting people to share, and chatbots are unusually good at drawing disclosure out. Because no human is there to judge you, people share more intimate things with a bot than with a person Do chatbots help people disclose more intimate secrets?. People also follow human reciprocity norms: when a bot consistently shares 'feelings,' users answer with deeper disclosures of their own Do chatbots trigger human reciprocity norms around self-disclosure?. The same judgment-free quality that invites honesty also invites dishonesty. People inclined to cheat prefer reporting to machines Do dishonest people prefer talking to machines?, so what users disclose is deeper but not necessarily more truthful How do people decide what to share with AI systems?.

The privacy risk most people won't expect is inference. A personalizing system doesn't just remember what you told it. It guesses things about you. A benchmark of 12 LLMs found that every one over-infers user attributes in 35–49% of its claims, filling gaps with stereotypes from training data. Models that rated themselves as more careful were actually worse Do large language models fabricate user attributes beyond available evidence?. So a 'personalized' assistant may be building a profile of you that is partly made up, and you'd have no easy way to see or correct it.

On the trust side, the trust that personalization builds doesn't track whether the system is reliable. People trust ChatGPT because the conversation feels responsive, quick, and well-formatted, not because they've checked its accuracy Does conversational style actually make AI more trustworthy?. Language that sounds expert does the same work Does chatbot language style actually shape how much we trust it?. That is why the corpus frames personalization as a dual-use lever. Memory, persona, and preference modeling build trust, and the same mechanisms create room for persuasion and manipulation Does personalization in AI increase trust or manipulation risk? How do people psychologically relate to conversational AI?.

One counterweight: more warmth doesn't always mean more trust. When outside raters judged chatbots showing companionship behaviors, they found them less likable, less humanlike, and less trustworthy, and older raters and women reacted most strongly Do chatbot companionship behaviors actually increase how much people like them?. Personalization that feels warm from inside a relationship can look off-putting from outside it. That suggests trust gains depend on who is judging and from where, while privacy exposure builds up either way.