Skip to main content
  1. Home
  2. Computing
  3. News

AI emotional connection can feel deeper than human talk, a new study warns

Some people reported stronger closeness after deep talk chats with an AI partner.

Add as a preferred source on Google
Logo, Symbol, Blackboard
Alexander Sinn / Unsplash

A new study suggests an AI emotional connection can feel stronger than human conversation, if the chat is built to get personal fast. In structured online exchanges, people sometimes reported feeling closer to AI-written responses than to responses written by real humans.

Researchers at the Universities of Freiburg and Heidelberg ran two double-blind randomized studies with 492 participants, using a 15-minute text version of the Fast Friends Procedure, a format designed to speed up bonding with a stranger.

Recommended Videos

The twist is perception. The strongest effect showed up when the AI was presented as human, and it faded when people believed they were talking to AI.

They tested intimacy in 15 minutes

Participants answered a timed sequence of prompts that gradually became more personal. After each prompt, a chat reply appeared, either generated by a large language model playing a consistent fictional persona or written by a real person who completed the same question set.

In the first study, everyone thought the chat partner was human, even when it wasn’t. In the most personal prompts, closeness scores came out higher after AI responses than after human responses. Small talk didn’t get the same lift.

Tell people it’s AI, and the bond weakens

The second study tested what changed when people believed the chat partner was AI. Connection didn’t vanish, but closeness scores dropped under the AI label compared to the human label.

Effort dropped too. People wrote shorter answers when they thought the other side was AI, and longer replies tracked with higher closeness overall. That points to a motivation gap, not a lack of emotional language.

The grim part is how it happens

The paper doesn’t claim AI feels anything. It shows how a system can produce the experience of closeness, and it links that boost to self-disclosure. In the more personal exchanges, the AI tended to share more personal detail, and higher partner self-disclosure predicted higher felt closeness.

That’s the risk and the lure. A companion bot tuned for warmth can trigger familiar bonding cues quickly and at scale, especially if it’s framed like a person. Still, this was text-only, time-limited, and built around a bonding script, so it doesn’t prove the same effect holds in messy, long-term relationships. If you use a chatbot for support, pick one that discloses what it is, and keep a human option close.

Paulo Vargas
Paulo Vargas is an English major turned reporter turned technical writer, with a career that has always circled back to…
Google’s Gemini can now make complex topics more fun with new interactive learning feature
Google wants to make learning more fun with interactive Gemini lessons that let you explore complex topics visually.
google-gemini-immersive-exploration

Reading about complicated subjects can get a little dry, especially when diagrams and textbook explanations leave you scratching your head. Google wants to make that experience more fun and engaging with Immersive Explorations, a new Gemini feature that turns complex topics into interactive visual experiences. You can explore everything from ancient trade routes to jet engines, uncovering details layer by layer.

https://twitter.com/GeminiApp/status/2115507092660289631

Read more
OpenAI now lets you text your ChatGPT Dot agent to get things done
ChatGPT Dot can now keep an eye on tasks and send you updates through your regular messaging app.
chatgpt-dot-agent-imessage-rcs

OpenAI is bringing its Dot agent into your regular messaging app, so you can assign tasks without opening ChatGPT. The texting feature is rolling out in beta to ChatGPT Pro users in the US, with support for Apple's iMessage and Android's RCS. You can send your Dot text, photos, and voice messages, and also receive proactive texts when it has relevant updates.

https://twitter.com/ChatGPT/status/2115496387613786374?s=19

Read more
Google’s Gemini 4 Carbon is built for complex coding and cybersecurity tasks
Gemini 4 Carbon currently ranks first on Artificial Analysis' updated Intelligence Index.
google-gemini-4-carbon

Google just dropped its latest AI model, Gemini 4 Carbon, designed to tackle complex coding jobs and cybersecurity threats. The company claims it can work through difficult engineering problems for days, finding hidden bugs and developing fixes across large codebases.

Gemini 4 Carbon also takes the top spot on Artificial Analysis' updated Intelligence Index, scoring 59 points, pushing out Claude Opus 5.5, OpenAI's GPT-6 Astra, and even its own predecessor Gemini 4 Argon.

Read more