How to Actually Get Better at Talking to People (Practice, Not Scripts)

You’ve read the threads. You know you’re supposed to “ask open questions,” “be confident,” and “just be yourself, but better.” And then a real conversation starts, your heart rate spikes, and every one of those tips evaporates while you say “so… do you come here often?” to someone standing in their own kitchen.

Here’s the uncomfortable part: none of that advice was wrong. It’s just useless in the moment, for the same reason reading about deadlifts doesn’t move the bar.

Conversation is a motor skill

Talking to people under mild pressure (a date, an interview, a stranger) is not a knowledge problem. You already know what a good conversation looks like. It’s a performance problem: producing words, in real time, while your nervous system is busy treating a first date like a fire alarm.

Performance problems have exactly one known fix, and it isn’t more reading. It’s reps under conditions that resemble the real thing. Musicians don’t get stage-ready by studying stage fright; they get there by playing in front of people until the fear runs out of material.

The catch with conversation is that reps are expensive. Every practice date is a real person’s evening. Every practice interview is a company you actually wanted. So most people do the rational thing: they don’t practice. They rehearse in their head, where they are, notably, extremely charming, and then wonder why it doesn’t transfer.

Silent rehearsal skips the exact part that fails: thinking while talking. Your inner monologue never stumbles, never gets interrupted, and never has to recover from a joke that died. Out loud is a different sport.

Why scripts make it worse

Memorized lines fail for a boring mechanical reason: conversations branch. A script survives until the other person says something you didn’t plan for, which is immediately, because that’s what people do: they interrupt, they give one-word answers, they ask “why?” at the worst time.

And when the script breaks, you’re worse off than if you’d had nothing, because now you’re doing two hard things at once: having the conversation and mourning the script.

What actually transfers between conversations isn’t lines. It’s a small set of trainable reflexes:

Recovering from a dead silence without panicking into a monologue.
Listening while planning nothing. Most people just wait for their turn to talk, and everyone can tell.
Being interrupted without losing the thread.
Letting a failed joke die with dignity instead of explaining it.

Notice that every one of those is a reaction. You cannot practice reactions alone in your head. You need something on the other end that pushes back.

Pressure is a feature, not a bug

There’s a well-worn finding in skill research: practice transfers best when it matches the conditions of performance. Comfortable practice makes you good at comfortable conversations, which are the ones you were already fine at.

The practice that moves the needle is the kind with something at stake, even something silly. A timer. A score. Someone who can visibly lose interest. Mild stakes teach your body that the spike of adrenaline is survivable, and survivable turns into ignorable, and ignorable is what confidence actually is. Confidence isn’t a personality trait; it’s an absence of novelty.

This is why “just talk to more people” is technically correct and practically terrible advice. The stakes are real people’s feelings and your own evenings. The feedback is nonexistent: nobody tells you why the conversation died; they just get quieter and leave. You bomb, you learn nothing, you’re now slightly more scared than before. Reps without feedback aren’t practice. They’re just exposure with extra steps.

What good practice looks like

Whatever you use (a friend who role-plays, a voice AI that pushes back, a mirror if you must), the checklist is the same:

Out loud. Non-negotiable. The skill is producing speech, so practice producing speech.
Unpredictable. If it can’t surprise you, it can’t train you. The other side has to interrupt, go cold, and change the subject.
Failable. There has to be a way to lose: a hang-up, a timer, a visible loss of interest. If nothing can go wrong, your nervous system knows it’s fake and learns nothing.
Scored. You need to know which sentence hurt you, not a vague sense that it “went okay.” Specific feedback is the difference between 50 reps and one rep repeated 50 times.
Repeatable within minutes. The magic of cheap reps is running it back immediately while the mistake is still warm. That’s the loop real conversations can never give you.

Do that a few minutes a day for a month and something undramatic happens: the fire-alarm feeling shrinks. You hear a dry answer and instead of oh no it’s dying, you get curious. You stop performing conversation and start having it. People will say you “came out of your shell,” which will be funny, because the shell was just a lack of reps.

(Full disclosure of our bias: we built RizzCall: 3-minute phone calls with AI characters who get bored, interrupt, and hang up on you, then score the call and quote your worst line back to you. It is reps, unpredictability, failure, and feedback in one loop, including a job-interview mode. It is also, we’re told, mildly traumatic in a useful way.)

The part nobody wants to hear

There is no version of this where you skip the bombing. Everyone who’s good at talking to people has a mass grave of dead conversations behind them; they just buried theirs early, cheaply, and mostly in private.

Your only real choice is where to bomb: in front of people whose opinions you care about, or somewhere with a retry button. Pick the retry button, run the reps, and let the real conversations catch you unusually warmed up.

She might still hang up on you. But it’ll take her a lot longer than it used to.