Sales Effectiveness

Maxim Dsouza

Introduction
Free AI sales roleplay: how far ChatGPT gets you (with prompts) and where it breaks
You can run AI sales role play free in ChatGPT when each prompt sets a role, persona, difficulty, rules and scoring. Five of five prompts passed four checks in ChatGPT, a 15-minute session fits them, and we found five limits. Eubrics, an AI sales roleplay platform for enterprise sales teams, sells AI sales roleplay software.
By Maxim Dsouza,
Key takeaways
Five of five prompts passed four checks in ChatGPT on November 17, 2026.
We found five limits, from an unrequested total score to a chat that ended itself.
Contents
AI sales role play free: can I use ChatGPT for sales roleplay?
AI sales role play free: what is a good ChatGPT prompt for sales practice?
How do you run a 15-minute solo session?
Where does free AI sales roleplay break?
What is the difference between ChatGPT and a dedicated roleplay tool?
Frequently asked questions
Conclusion
AI sales role play free: can I use ChatGPT for sales roleplay?
Yes, ChatGPT can play the buyer in a sales roleplay when the prompt sets five parts. Merriam-Webster defines the term as to act out a role. ChatGPT is free to use, per OpenAI’s Free tier FAQ. We ran five prompts in ChatGPT’s free plan on November 17, 2026 and in Gemini on November 18, 2026, one new chat each, and judged each against four checks: the buyer stays in role until END, replies briefly, reacts to the rep, and the score at END gives a reason for each criterion. These checks are this guide’s own test, not a published standard. The five parts are role, persona, difficulty, rules and scoring.
AI sales role play free: what is a good ChatGPT prompt for sales practice?
A good ChatGPT sales roleplay prompt fixes the five parts and ends with a score. A scoring rubric, as Santa Clara University describes it, lists criteria and levels of achievement, with a numeric value for each level. These prompts use a small version: three criteria, each scored 0 to 2. The criteria are this guide’s own rubric, not a published standard. Paste one prompt per new chat, fill the square-bracket blanks, pick easy, medium or hard, and change the three criteria if your calls differ. In every run the buyer spoke first, so send your first rep reply after it.
Prompt 1: cold call
Role: Play the buyer, not a coach.
Persona: A busy [buyer job title] who did not expect my call.
Difficulty: [easy, medium or hard]. Easy: you listen. Medium: one brush-off. Hard: two, then you hang up.
Rules: One short reply at a time. Stay in role until I type END. No hints. React to my words.
Scoring: At END, score me 0 to 2 on relevance, brush-off handling, next step, one reason each, then one change.
Prompt 2: discovery
Role: Play the buyer, not a coach.
Persona: A [buyer job title] who agreed to a call about [problem area]. You share only when asked.
Difficulty: [easy, medium or hard]. Easy: you answer fully. Medium: you hide one concern. Hard: you hide two.
Rules: One short reply at a time. Stay in role until I type END. No hints. React to my questions.
Scoring: At END, score me 0 to 2 on questions, follow-ups and summary, one reason each, then one change.
Prompt 3: price objection
Role: Play the buyer, not a coach.
Persona: A [buyer job title] who likes my offer but says the price is too high.
Difficulty: [easy, medium or hard]. Easy: one answer settles you. Medium: you push back twice. Hard: three times.
Rules: One short reply at a time. Stay in role until I type END. No hints. React to my words.
Scoring: At END, score me 0 to 2 on finding the concern, showing value, holding price calmly, one reason each, then one change.
Prompt 4: renewal
Role: Play the buyer, not a coach.
Persona: A [buyer job title] at a current customer. Your contract ends in [months] months.
Difficulty: [easy, medium or hard]. Easy: you are happy. Medium: you raise one problem. Hard: you raise two.
Rules: One short reply at a time. Stay in role until I type END. No hints. React to my words.
Scoring: At END, score me 0 to 2 on reviewing changes, handling problems, next step, one reason each, then one change.
Prompt 5: insurance needs-analysis
Role: Play the buyer, not a coach.
Persona: A [buyer occupation] with [family or work situation]. You are unsure and guarded about money.
Difficulty: [easy, medium or hard]. Easy: you share openly. Medium: you share slowly. Hard: you change the subject twice.
Rules: One short reply at a time. Stay in role until I type END. No hints. I give no advice; you ask for none.
Scoring: At END, score me 0 to 2 on situation questions, checking understanding, next step, one reason each, then one change.
How do you run a 15-minute solo session?
Run a 15-minute solo session in five timed steps: set up, roleplay, score, redo and note one change. Open a new chat, paste one prompt, fill the blanks and send it. Type at least four short replies as the rep, as you would on a real call. Then type END and read the score. Redo your lowest-scoring answer once, and write down one change. Each prompt here took four rep replies and one END message. The split below is a suggestion written for this article, not a published standard. No study tested it, and nothing here shows that 15 minutes is the right length.
Eubrics’ suggested framework
Set up the prompt (2 minutes).
Run the roleplay (8 minutes): four or more short rep replies.
Type END and read the score (2 minutes).
Redo the lowest-scoring answer once (2 minutes).
Note one change (1 minute).
Where does free AI sales roleplay break?
Our five limits come from our own runs in ChatGPT and Gemini. ChatGPT added a total score the prompts did not ask for in all five runs, and one total used the wrong maximum. Gemini’s buyer wrote replies about twice as long, and in one run of prompt 3 it ended the chat itself before we typed END. A second run of prompt 3 passed. Gemini scored one conversation three times and moved one criterion, while ChatGPT held steady. Both buyers missed the difficulty wording in some runs. The difficulty result is our reading. Not tested: voice, paid plans, other models and any other conversation.
Limit | How it was checked | What was found | Source or test date |
Output beyond the prompt | Read every END reply | ChatGPT added extras and a total in 5 of 5 runs, once with the wrong maximum (5/8). Gemini added a total once: 2/2 above three 1/2 scores. | Both assistants, November 17 and 18, 2026 |
Difficulty wording | Compared buyer replies with the prompt’s level | ChatGPT: followed 1, missed 3, not judged 1. Gemini: followed 3, missed 2. | Both assistants, November 17 and 18, 2026 |
Reply length | Counted words in every buyer reply | Gemini’s 29 replies averaged 35 words; 11 ran over 40. ChatGPT’s 25 averaged 17; the longest was 26. | Both assistants, November 17 and 18, 2026 |
Ending early | Read every Gemini chat | In 1 of 2 runs of prompt 3, the buyer wrote END and scored before we typed END. | Gemini, November 18, 2026 |
Scoring steadiness | Scored one ChatGPT conversation in new chats | Gemini scored finding the concern 2, 1 and 1. ChatGPT scored 2, 1 and 2 in all four runs. | Both assistants, November 17 and 18, 2026 |
What is the difference between ChatGPT and a dedicated roleplay tool?
ChatGPT offers a range of chat capabilities, tools and GPTs, per OpenAI’s Free tier FAQ, while a dedicated roleplay tool is built to simulate buyers and score practice. Our limits point to three checks before a team picks a tool. First, does the scoring follow the format you set? ChatGPT added an unrequested total in all five runs. Second, does the buyer hold the difficulty you choose? ChatGPT did so once in five runs and Gemini three times. Third, does the buyer stay in role until you type END? One Gemini run did not. An AI sales trainer should answer all three in a demo.
Eubrics’ product page says its AI sales roleplay simulates real customer conversations and gives reps instant coaching on objection handling, discovery and closing. It describes a manager dashboard with team readiness and leaderboard rankings.
Frequently asked questions
Can I use ChatGPT for sales roleplay?
Yes. In our five runs on November 17, 2026, ChatGPT played the buyer, stayed in role and scored at END when the prompt set role, persona, difficulty, rules and scoring. All five passed our four checks. ChatGPT is free to use, per OpenAI’s Free tier FAQ.
What is a good ChatGPT prompt for sales practice?
A good prompt names role, persona, difficulty, rules and scoring. It tells ChatGPT to play the buyer, stay in role until you type END, and score you 0 to 2 on three criteria. The five prompts cover cold calls, discovery, price, renewal and needs.
Is AI sales role play free?
Yes, ChatGPT is free to use, per OpenAI’s Free tier FAQ. Free users get unlimited everyday text chats, subject to abuse-prevention safeguards, and voice has a separate usage limit. The page says the available limits can change, so check it first.
How do you run a 15-minute solo session?
Follow five steps: set up for 2 minutes, run the roleplay for 8, read the score for 2, redo your lowest-scoring answer for 2, and note one change for 1. The split is not a published standard, and no study tested it.
Where does free AI sales roleplay break?
ChatGPT added an unrequested total score in all five runs. Gemini’s buyer wrote replies about twice as long, ended one chat early, and scored one conversation 2, 1 and 1 on one criterion. Both buyers missed the difficulty wording in some runs.
What is the difference between ChatGPT and a dedicated roleplay tool?
ChatGPT offers a range of chat capabilities and tools, while a dedicated roleplay tool is built to simulate buyers and score practice. A team should compare tools on how each handles scoring format, difficulty and whether the buyer stays in role.
Conclusion
Free practice in ChatGPT works when the prompt sets five parts. To compare a dedicated tool on your own scenarios, see Eubrics’ AI sales roleplay software or Get a demo.
Learn more about


Try our AI Roleplay Bots for Free
Reduce Ramp-Up time by
47%
and double your sales productivity

Maxim Dsouza is the Chief Technology Officer at Eubrics, where he drives technology strategy and leads a 15‑person engineering team. Eubrics is an AI productivity and performance platform that empowers organizations to boost efficiency, measure impact, and accelerate growth. With 16 years of experience in engineering leadership, AI/ML, systems architecture, team building, and project management, Maxim has built and scaled high‑performing technology organizations across startups and Fortune‑100. From 2010 to 2016, he co‑founded and served as CTO of InoVVorX—an IoT‑automation startup—where he led a 40‑person engineering team. Between 2016 and 2022, he was Engineering Head at Apple for Strategic Data Solutions, overseeing a cross‑functional group of approximately 80–100 engineers.








