ShouldaSaid is an AI argument simulator that gives you a real opponent — one that pushes back, makes counterpoints, and doesn't let you off easy. Practice your logic, stress-test your reasoning, and finally say what you actually meant.
Pick any topic — social, political, everyday. The AI automatically takes the opposite side. Three rounds, scored on logic and vocabulary. Good for building argumentation skills from scratch.
Describe a real argument you had — at work, at home, anywhere. The AI plays the other person. You get to replay it and say what you actually meant this time.
No scoring, no judgment. Just say what you need to say. The AI responds in character and lets you get it all out. Sometimes that's enough.
Each round is scored on three factors:
Argument length — a detailed, developed point scores more than a one-liner. You're rewarded for explaining, not just asserting.
Vocabulary variety — using different words to make your points scores higher than repeating the same phrases. The scorer looks for range, not complexity.
Logical connectors — words like "because", "however", "therefore", and "as a result" earn bonus points. These signal structured thinking.
No insults — personal attacks deduct points. The simulator rewards arguments, not aggression.
Most people are bad at arguing under pressure — not because they lack intelligence, but because they haven't practiced. Real arguments come with adrenaline, emotion, and time pressure. Simulating them in a low-stakes environment helps you:
The "Personal Redo" mode is built around a well-documented psychological phenomenon — l'esprit de l'escalier, or staircase wit. The term, coined by French philosopher Denis Diderot, describes the experience of thinking of the perfect reply only after the conversation is over. You're walking away and the ideal comeback, the decisive counterpoint, the thing you actually wanted to say arrives — too late to matter.
When we replay arguments in our heads, we're doing genuine cognitive work: organizing our thoughts, processing the emotional residue of the exchange, building a more accurate model of the other person's position. The problem is that this replay happens after the fact and doesn't directly improve our real-time performance. We process the experience, but that processing doesn't get stored in a way that helps us in the next high-pressure moment.
The Personal Redo converts that natural replay instinct into active practice. By simulating the other person with an AI and forcing you to actually make the arguments — not just think them — it gives the brain a chance to encode the better response. Research on "mental contrasting" and structured rehearsal shows that simulating a performance transfers to real-world improvement better than passive replay does. The argument you practise making becomes easier to make when it matters.
There's also an emotional processing component. The frustration of an argument you felt you lost, or a conversation where you didn't say what you needed to say, carries cognitive load — it tends to resurface as rumination. Getting the argument out, even in a simulated form, gives that loop somewhere to go.
The AI in ShouldaSaid doesn't just agree with you or serve up canned counterarguments. It takes the opposite position and actively looks for weaknesses in your reasoning. If your first argument is vague or unsupported, it will press harder. If you make a strong point, it will find the edge case or the exception. If you steelman its position — acknowledge the strongest version of the opposing argument — it will raise the stakes rather than backing down.
This makes it a meaningfully better practice tool than asking a friend to "play devil's advocate." Friends know when to let you win. They pull punches to preserve the relationship. Even when they try to argue the other side, they're limited by their own knowledge and motivated by their investment in you. An AI opponent has no such constraints — it will keep pushing until you've made a genuinely complete argument.
It's also better for practice than conversational AI tools that are optimized to be agreeable. ShouldaSaid's AI is specifically tuned to maintain its position and respond to the argument you actually made, not the argument it might prefer you to make. This creates a more demanding — and more useful — practice environment.
Debate Practice — Use this when you want to build general argumentation skills, prepare for a professional situation (an interview, a negotiation, a presentation), or practise in a second language. Any topic works; the daily rotating challenge removes the decision fatigue of choosing.
Personal Redo — Use this when you left a specific real argument feeling like you had more to say; when you're preparing for a specific difficult conversation you know is coming; or when you want to stress-test your position before committing to it in front of someone who matters. Describe the situation in as much detail as you're comfortable with — the more context you give the AI, the more realistically it can simulate the other person.
Vent Mode — Use this when you don't want feedback or scoring — you just need to say it all. When you're processing something emotionally and evaluation would be counterproductive. When you want to rehearse without being assessed. Vent Mode doesn't cost Sparks and doesn't run on a timer — it's there whenever you need it.
The argument simulator is at its most powerful when used before a difficult conversation you know is coming. A salary negotiation. A confrontation with a coworker about something they've been doing wrong. A hard talk with a family member about a decision they've made. These conversations are difficult precisely because the stakes are high — and high stakes make most people perform worse, not better, unless they've prepared.
By simulating the conversation in advance — with the AI playing the other person using the context you provide — you're giving your brain the preparation it needs to stay calm and clear when the real version happens. Lawyers moot their arguments before trial. Athletes visualize their performance in advance. Actors rehearse until the lines are muscle memory. This is the same principle applied to the conversations that actually shape your daily life.
The goal isn't to script the conversation — real conversations don't follow scripts. It's to ensure you've thought through your main points, anticipated the responses you're most likely to hear, and made the core arguments at least once before the stakes are real. After one or two simulations, you'll find that the real conversation feels less like a test and more like something you've already done.
How is this different from ChatGPT?
ChatGPT is optimized to be helpful and agreeable. It will tend to validate your position, offer balanced perspectives, and avoid sustained confrontation. ShouldaSaid is specifically designed to argue back — to maintain the opposing position, find weaknesses in your reasoning, and keep pressing. It's a purpose-built training tool, not a conversational assistant.
Can I replay the same scenario multiple times?
Yes. Running the same scenario multiple times is one of the most effective ways to use the simulator — you'll find that your arguments get sharper with each iteration. Each session starts fresh, so the AI won't remember your previous runs, but you will.
Does it remember previous sessions?
No. Each session is independent. This is intentional — it means you can replay scenarios freely without the AI adapting to you, which keeps the practice honest. Your scores and history are stored in your profile if you create an account, but the AI itself starts fresh each time.
Is it appropriate for serious personal situations?
For preparing for difficult conversations, yes. For processing genuine emotional distress, Vent Mode may be more appropriate than Debate Practice or Personal Redo. ShouldaSaid is not a mental health tool — if you're dealing with something that requires professional support, please reach out to a therapist or counsellor.
Can I use it as a debate coach?
In a limited but real sense, yes. The scoring gives you consistent feedback on argument structure, vocabulary, and connector usage across sessions. You can track which areas you're improving in and which you aren't. It won't give you the nuanced, human feedback of a trained coach, but it provides structured, repeatable practice that a coach can't — unlimited availability at no marginal cost per session.
Related reading:
No signup. No download. Works on any device. Available in English, German, Arabic, and Spanish.
Try the Argument Simulator →More from ShouldaSaid