You can copy the prompts below right now and adapt them with a five-part structure: Role, Personality, Context, Success criteria, and Stop rules. We built this framework on OpenAI’s own prompt guidance and years of coaching real estate investors. Below, you’ll find ready-to-use examples, a modular template, safety guardrails, and tips for applying all of it to cold calling practice.

TL;DR:

  • Use constraints rather than vague traits: a seller who needs a sibling’s approval or will not commit before speaking with her son creates actionable resistance.
  • Set stop rules for distress, self harm, harassment, or exiting character; debrief afterward with a human instead of relying on automated scores alone.
  • For foreclosure, add a real deadline; for probate, make another decision maker’s approval necessary; for divorce, teach callers to redirect sensitive questions without losing rapport.
  • Use AI for repeated drills and immediate scorecards, but rely on a human partner to catch hesitation and emotional subtext; rotate scenarios across practice days.
ClosersLeague
Practice More Confident Seller Calls
ClosersLeague coaches real estate investors and wholesalers to make better calls with homeowners facing foreclosure, probate, divorce, or other distress.

Explore ClosersLeague

Table of Contents

1. Copy-ready prompts for every roleplay scenario

Pick the category that matches what you’re building, then swap the names, stakes, or constraints to fit your situation. These prompts work in any chatbot that accepts a system or instructions message.

Character scenes

  1. “You are Mara, a skeptical landlord who inherited a rental property she doesn’t want. Stay in character, answer only as Mara, and reveal her real motivation only if the user asks about her late father.” When to use: practicing empathy-driven discovery questions. Success looks like: the AI reveals backstory only when earned, not volunteered upfront.
  2. “You are a retired teacher facing foreclosure who feels embarrassed about the situation. Respond with short, guarded answers until trust is built.” When to use: training tone and pacing for sensitive openers. Success looks like: guarded replies that loosen only after reassurance.

NPC and game dialogue
3. “You are a tavern keeper NPC named Brogan who trades rumors for coin. Never break character and keep responses under three sentences.” When to use: fast, replayable game dialogue. Success looks like: consistent voice across multiple exchanges.
4. “You are a quest-giver guarding a locked gate. Refuse entry unless the player answers a riddle correctly.” When to use: puzzle-gated game scenes. Success looks like: the AI enforces the rule without exceptions.

Collaborative storytelling
5. “Co-write a mystery with me. You control the detective; I control the witness. Introduce one new clue per turn.” When to use: building interactive fiction. Success looks like: balanced pacing with neither party dominating the plot.
6. “You are a narrator for a fantasy adventure. Describe settings vividly but never make decisions for my character.” When to use: solo interactive fiction. Success looks like: narration stays descriptive, not directive.

Structured training drills
7. “You are a homeowner in pre-foreclosure, anxious and defensive. Grade my call afterward on rapport, objection handling, and clarity.” When to use: cold calling skill drills. Success looks like: a clear, honest scorecard at the end.
8. “You are a probate heir unsure whether to sell. Push back twice before agreeing to a follow-up call.” When to use: practicing persistence without pressure. Success looks like: realistic resistance followed by a believable close.

Light conversational characters
9. “You are a friendly barista who remembers regular customers. Keep replies brief and warm.” When to use: casual conversation practice. Success looks like: natural, low-stakes small talk.

Minimal edits like swapping a name, raising the financial stakes, or adding a time constraint turn any of these into a fresh scenario in seconds.

2. A five-part framework that keeps any prompt on track

Most roleplay prompts fall apart because they skip structure. OpenAI’s prompt guidance recommends defining the outcome, constraints, and stopping point up front rather than hoping the model figures it out mid-conversation. We use five modules for every prompt we build:

  • Role: Who the AI plays. Fill-in: “You are [name], a [role] who [defining trait].”
  • Personality: How they sound and react. Fill-in: “Speak in a [tone] way and react with [emotional pattern].”
  • Context: The scenario’s facts and rules. Fill-in: “The situation is [setup]. You know [fact]. You don’t know [unknown].”
  • Success criteria: What a good exchange looks like. Fill-in: “A strong response from you does [behavior]; a weak one does [behavior].”
  • Stop rules: When to end or escalate. Fill-in: “If [condition], stop the roleplay and [action].”

Assembled, a complete prompt reads: “You are Mara, a landlord who inherited a rental property. Speak cautiously and warm up only after trust is built. You’re unsure whether to sell and haven’t told your siblings yet. A strong response from the user builds rapport before asking about the property. If the user becomes pushy or rude, end the roleplay and state that this behavior wouldn’t work on a real call.”

Pro Tip: Keep personality instructions separate from task logic. OpenAI’s cookbook on prompt personalities treats personality as an operational lever that controls consistency, not just flavor text, so blending it into your rules invites drift over a long conversation.

3. Reusable templates for games, stories, and training sims

These four templates have variable slots you fill in once and reuse indefinitely.

  1. GM/NPC encounter template: “You are [NPC role], motivated by [motivation]. You will not [constraint]. React realistically to player choices and never narrate for the player.” Swap the role, motivation, and constraint for new encounters in seconds.
  2. Collaborative storytelling template: “Co-write a story where I play [protagonist]. Introduce [twist] by the third exchange. Pace the story with [cues, e.g., short chapters or cliffhangers].” This keeps pacing consistent across sessions.
  3. Scored training simulation template: “You are [role] with the objective of [objective]. After the exchange, score my performance on [rubric categories] and ask [debrief questions] to help me reflect.” This format works well for cold calling practice, interview prep, or negotiation drills.
  4. Safe character template: “You are [character]. Keep all content appropriate for a general audience. If the conversation moves toward [restricted topic], redirect politely and stay in character.” This template keeps roleplay respectful without feeling robotic.

Each template takes under a minute to customize, and the slot-based structure means you’re never starting from a blank page.

4. Why stop rules matter and how to write them

Stop rules keep a roleplay from drifting into unproductive or risky territory. OpenAI’s updates on sensitive conversations call for explicit constraints around high-risk content, and the same logic applies to training simulations that touch on financial stress or grief.

  • “If the user expresses genuine distress, stop the roleplay and suggest a short break.”
  • “If the conversation moves toward self-harm or harassment, end the scenario and provide a supportive resource.”
  • “If the user asks to exit character, immediately drop the roleplay and respond normally.”

After any serious training run, a short human-led debrief adds value an automated scorecard can’t replicate. Debriefing standards from INACSL stress that reflection, not just grading, is what makes a simulation stick.

Reporting on AI use suggests a small but real risk of emotional over-reliance in extended chatbot interactions, which is one more reason to pair automated practice with real feedback from a coach or peer (Digital Trends).

5. What practitioners get right when building roleplay prompts

Two habits separate prompts that work from ones that fall flat. First, calibrate realism with concrete constraints: a homeowner who “sometimes hesitates” is vague, but one who “won’t commit until she talks to her son” gives the AI something to act on. Second, pair short success criteria with immediate feedback so you know what to fix before your next attempt. We cover more of these patterns in our guide to real estate roleplay training for distressed calls.

6. Customizing prompts for foreclosure, probate, and divorce scenarios

Distressed seller situations each carry their own emotional undercurrent, and your prompts should reflect that instead of using one generic “upset homeowner” template.

For foreclosure, build urgency and shame into the Personality module: “Speak tensely and avoid admitting the full timeline unless pressed gently.” The Context module should include a real deadline, since pressure changes how people negotiate.

For probate, lean into uncertainty rather than urgency. The seller often isn’t the sole decision-maker, so your Context module might read: “You can’t sell without your sibling’s agreement, and you haven’t discussed it with them yet.” This forces practice around multi-stakeholder objections.

For divorce, personality matters most. Add a line like “Respond curtly if asked about your ex-spouse” to train callers on redirecting touchy subjects without losing rapport.

In every case, keep Success criteria specific to the scenario’s emotional terrain. A good response to a foreclosure lead looks different from a good response to a probate heir who’s still grieving. Rewriting just the Personality and Context modules, while keeping Role and Stop rules steady, lets you build a full library of distressed-seller scenarios without starting from scratch each time. Our pre-foreclosure cold calling practice page shows what a fully built-out version of this looks like in practice.

6. Customizing prompts for foreclosure, probate, and divorce scenarios — overview diagram

7. Fitting roleplay prompts into a cold calling training routine

Prompts work best as part of a repeatable routine, not a one-off exercise. Start each practice block with a clear objective, like handling a specific objection type, then pick or build a prompt that targets it directly.

Run the call, then immediately review the AI’s scorecard if your prompt includes a scoring rubric. Follow that with a short human debrief, even if it’s just five minutes of self-reflection on what felt natural versus forced. This two-step loop, automated feedback followed by reflection, mirrors what simulation researchers recommend for skill retention.

Rotate through scenario types across a week: foreclosure one day, probate the next, divorce after that. This keeps practice from becoming repetitive and builds comfort across the full range of distressed-seller conversations you’ll encounter on real calls. Our AI cold call simulator guide walks through how to structure a full weekly practice schedule around this kind of rotation.

8. Matching prompt difficulty to your skill level

Not every caller needs the same level of resistance baked into a prompt. Adjusting difficulty is mostly a matter of tweaking the Personality and Stop rules modules.

Starter-level prompts should feature cooperative sellers with mild hesitation: “You’re mildly unsure but open to hearing an offer.” This builds basic rapport and script familiarity without overwhelming a new caller.

Growth-level prompts introduce real objections and emotional complexity: “You’re defensive about your financial situation and will test the caller’s patience twice before softening.” This stage is where most reps build the instincts that carry into real calls.

Pro-level prompts should combine multiple pressure points: a skeptical, time-pressed seller with a hidden constraint the caller has to uncover, like an upcoming auction date they haven’t mentioned. Success criteria at this level should reward speed and precision, not just eventual rapport.

Moving between levels just means adjusting how much resistance and ambiguity you write into the Context and Personality modules, while keeping the same underlying framework intact.

9. When AI roleplay beats a human practice partner, and when it doesn’t

AI roleplay wins on scale: you can run ten scenarios before lunch, each with instant feedback. Human partners win on nuance, picking up on hesitation or emotional subtext an AI might miss. The strongest training programs use both: AI for repetition and skill drilling, a human for deeper reflection on the calls that matter most.

— Dave

10. ClosersLeague: scenario-based roleplay practice built for cold callers

We built ClosersLeague around exactly this kind of scenario-based AI roleplay practice, with instant scorecards, skill drills, and leaderboards covering nine distressed seller types. Plans run Starter, Growth, and Pro to accommodate different training needs. Check current pricing to find the plan that fits your call volume.

ClosersLeague

FAQ

How do I make my own roleplay AI?

Start with the five-part framework: define a Role, a short Personality, the Context the character knows, Success criteria for a good exchange, and Stop rules for ending the scene. Paste this into any chatbot’s system or instructions field and test it with a few sample turns before relying on it.

What are the best roleplay ideas to try first?

Character-driven scenes with a clear emotional stake, like a skeptical landlord or an anxious seller, tend to produce the most useful practice. Game-style NPC dialogue and collaborative storytelling prompts are also strong starting points because they’re easy to replay with small tweaks.

What are the most effective AI prompts for roleplay?

The most effective prompts define an outcome and a stopping point rather than leaving the AI to guess, a structure OpenAI’s prompt guidance recommends for reducing drift in longer exchanges. Pairing a short personality instruction with explicit success criteria consistently produces more usable practice sessions.

What makes a good personality prompt for a chatbot?

A good personality prompt is short and functional, describing tone and reaction patterns rather than a full backstory. OpenAI’s cookbook on prompt personalities treats personality as a control for consistency, so keeping it brief helps the model stay in character over a longer conversation.

Sources