Rehearsing a hard conversation with an AI before you have it for real — asking for a raise, ending something, giving hard feedback — has a name now: "dry-chatting." And it's not a fringe habit. A Forbes report published in March 2026 found that 28% of people surveyed admit to practicing a difficult conversation with an AI companion before facing the real one, and one in four say they do it specifically to settle their nerves beforehand.
What “Dry-Chatting” Is
The term describes exactly what it sounds like: running the conversation you're dreading — a confrontation, a difficult ask, a breakup — with an AI playing the other person first, purely to get the words out of your mouth once before it counts. It's not about getting a script to memorize. It's about lowering the activation energy of starting.
Why Rehearsal Works
This isn't a new idea — actors, lawyers, and negotiators have rehearsed high-stakes conversations for as long as those professions have existed. What's new is having a rehearsal partner available at 11pm the night before, for free, with no one to report back to. The mechanism reporting on "dry-chatting" points to a few specific effects: it lowers emotional intensity by giving you something concrete to focus on, it forces you to put vague dread into actual words, and it delays your first real exposure to the stressful moment just long enough to walk in with some composure already spent.
The same underlying idea shows up in AI-roleplay training used for workplace negotiation and feedback conversations — the premise being that most people only get a handful of real chances a year to practice something like giving hard feedback, which isn't enough repetition to get good at it the normal way.
How to Do It Well
- Set the real stakes up front. Tell the AI who the person is, what's actually at risk, and how that person tends to react — vague prompts get vague rehearsal.
- Say the actual opening line out loud (or type it), not a summary. The value is in producing the real words under mild pressure, not planning bullet points.
- Let it push back. A rehearsal partner that just agrees with everything you say isn't rehearsal — ask it to react the way the real person plausibly would, including the annoying parts.
- Do it more than once. The value compounds — the first pass gets the dread out, later passes actually sharpen what you say.
Build a Practice Partner
You can dry-chat with any character, but a persona built specifically for this works better than a generic assistant. Building one at /create/character takes a few minutes:
- Basic Info — name it after the role, not a person (e.g. "Practice Partner," not your actual boss's name).
- Bio & Personality — describe how the real counterpart tends to react: defensive, calm, dismissive, whatever's true. Ask explicitly for realistic pushback, not agreement.
- First Message — set the scene the way the real conversation will actually start, so you're rehearsing the real opening, not a hypothetical one.
- Example Dialogue (optional) — if you know how this person talks, a couple of example lines locks in their voice more accurately.
- Visibility — Private, unless you want to reuse the same persona for other rehearsals later.
Rehearse it before you say it for real
Build a practice-partner character in a few minutes and get the hard part out before it counts.
Build a practice partner →Frequently Asked Questions
What is dry-chatting?
Does practicing with AI actually help with real conversations?
What kinds of conversations do people practice this way?
Sources
- Bryan Robinson, “‘Dry-Chatting’ With ChatGPT Helps You Face Difficult Work Situations,” Forbes (2026-03-13)
Garret Williams is the founder and CEO of chatbrat.ai, building at the frontier of AI companions and roleplay chatbots. Before pivoting to tech, the Michigan native studied marketing at UCLA and directed acclaimed film projects—including his TV pilot Self-Care (nominated for Best TV Episodic at the 2023 Mammoth Film Festival) and Eco-Riot (featured on MasterClass). He writes The Bratlog to document the uncharted territory of AI relationships, sharing real-time lessons and tackling the open questions nobody has answered yet.