If you want to know how to be persuasive in a conversation, you’ll find plenty of advice that sounds clever and works badly. Talk faster. Flatter them. Create fake urgency. Some of it might get you a grudging “fine” today. Very little of it gets you a person who still agrees with you next month.
So let’s start somewhere less glamorous. Persuasion in real life rarely looks like a courtroom closing speech. It looks like a coffee with a colleague who doesn’t want the new process, a parent who thinks your career move is a mistake, a friend who’s dug in on something you think is plainly wrong. Two people, one table, a lot of feelings.
The good news? Over the past two years, researchers have run some unusually large experiments on exactly this situation, including field studies with trained canvassers, tens of thousands of AI-driven dialogues and national surveys of how people actually talk about contested topics. The findings are a bit humbling. They also point to habits you can practice tomorrow.
Below are 22 of them. None require charisma you don’t have. If you want the foundations first, our guide on how to have good conversation with anyone is a good warm-up.
1. Decide what “winning” looks like before you start
Most of us walk into a disagreement picturing a full conversion. They hear our points, pause, and say “you know what, you’re right.” That almost never happens, and expecting it makes you push too hard.
The numbers back this up. In Pew Research Center’s survey of 3,560 U.S. adults, fielded in December 2025 and published in February 2026, 59% said they at least sometimes change their own opinion when they discuss the news with others. Only 4% said it happens all or most of the time. People do move. They move occasionally, and usually by a notch.
That reframes the job. A successful conversation might end with “I hadn’t thought about it that way,” or simply with a willingness to talk again. If you measure yourself against total surrender, every realistic result will feel like failure, and you’ll start overselling.
Start small: before your next tricky conversation, write down one modest outcome you’d count as progress, such as “she agrees to look at the numbers” or “he admits the current system has a cost.”
2. Get curious first and ask real follow-up questions
Questions feel like a detour when you’re itching to make your case. They’re actually the shortest route, because they tell you what the other person believes and why, and they make you more likeable while you’re at it.
Alison Wood Brooks, a professor at Harvard Business School, describes in her January 2025 book Talk how her team analyzed 15-minute get-to-know-you conversations among 398 strangers. They sorted every question into four types: introductory, mirror (bouncing the same question back), topic-switching and follow-up. When people were nudged to ask more questions, the likeability boost came almost entirely from follow-ups, the ones that pick up on something the other person just said. Mirror questions did nothing for how people were seen. These were friendly chats, not arguments, so treat it as evidence about rapport rather than about changing minds directly.
Still, rapport is the soil persuasion grows in. A follow-up such as “what made you start feeling that way?” also hands you their actual reasoning, which you’ll need later (see item 15). If you tend to fill silences with your own stories, our piece on how to stop talking about yourself in conversations pairs nicely with this one. In your next disagreement, ask at least two follow-ups before you state your own view at all.
3. Listen to lower their guard, and don’t expect listening to do the persuading
This next finding cuts against a lot of popular advice.
Roni Porat of the Hebrew University of Jerusalem, with Erik Santoro, David Broockman and Joshua Kalla, published a preregistered field experiment in PNAS in March 2025. Nearly 1,500 Americans had ten-minute video conversations with trained canvassers about in-state tuition for undocumented immigrants. The team varied two things independently: whether the canvasser shared a persuasive personal story, and whether they practiced high-quality, nonjudgmental listening. Attitudes were measured right away and again five weeks later.
The stories produced real, lasting shifts. Listening made participants like the canvasser more and feel less emotionally defensive, but it did not make the story more persuasive.
So listening isn’t a magic persuasion lever. It’s closer to clearing the road. A defensive person can’t really hear you, and good listening lowers that barrier, even if something else has to carry the actual argument. It’s one study on one issue, so don’t read it as “listening never helps.” Read it as “listening is necessary, not sufficient.” Our guide on how to stop interrupting people when they talk covers the mechanics.
Next time, do this: let the other person finish their full point, wait two seconds, and only then decide what to say.
4. Restate their view in your own words before you answer
“I hear you” is the conversational equivalent of a shrug. Everyone says it. Nobody believes it.
Julia Minson, a professor of public policy at Harvard who studies disagreement, makes this point in an excerpt from her book How to Disagree Better published by the Harvard Gazette in March 2026. Her research packages receptive language into a recipe she calls H.E.A.R.: hedge your claims, emphasize agreement, acknowledge the other perspective and reframe to the positive. On acknowledgment, she is specific. Saying you hear someone doesn’t count unless you show evidence that you listened, which usually means restating their point. In her randomized experiments, people trained in these habits were rated by their opponents as more trustworthy, more objective and better future collaborators, on contentious topics like COVID-19 vaccines and affirmative action.
Picture this, purely as an illustration: your partner says the holiday budget is too high. You reply, “So your worry is less about this trip and more that we keep going over budget every quarter. Is that fair?” Now you’re arguing about the same thing. For more on this skill, see how to be an understanding person.
Your move: in your next disagreement, open your reply with a one-sentence summary of their position and ask if you got it right.
5. Hedge a little, especially on what you’re unsure about
It feels backwards. Surely confidence persuades and qualifiers make you sound weak?
Not quite. Jonah Koetke and Karina Schumann of the University of Pittsburgh, with Shauna Bowes and Nina Vaupotič, published five preregistered studies with 2,034 participants in Nature Human Behaviour in February 2025. When a scientist was described as intellectually humble rather than not, people saw that scientist as more trustworthy and believed their research more. In one study, they were also more willing to follow the scientist’s recommendations. The effects didn’t depend on the scientist’s gender or ethnicity, and they held across medical, psychological and climate topics.
You’re probably not a scientist, and these studies measured reactions to descriptions of people, not live conversations. The underlying logic transfers well, though. When you say “most of the evidence points this way” instead of “everyone knows,” you sound like someone tracking reality rather than defending territory. Minson’s H.E.A.R. recipe makes hedging its first step for the same reason. Swap one absolute (“always,” “never,” “obviously”) for an honest qualifier the next time you make your case.
6. Bridge the identity gap and name what you share
People don’t just weigh arguments. They weigh who’s making them. If you feel like “the other side,” your points get filtered through that.
Martin Naunov of Northwestern University, with Timothy Ryan and Carlos Rueda-Cañón of UNC Chapel Hill, tested this in a study published in the Journal of Politics in 2025. More than 400 ordinary people wrote arguments aimed at someone who disagreed with them on immigration, climate or transgender rights. Over 3,000 readers then rated their own views before and after reading a randomly assigned argument, producing more than 9,000 evaluations.
About 30% of the persuasion attempts moved the reader toward the writer’s side. Backlash, where the reader moved the other way, happened in about 11% of cases, a level the researchers couldn’t distinguish from measurement error.
The arguments that worked best bridged identity divides, took the reader’s perspective, used personal stories and emphasized common ground. That 30% is a decent hit rate for a single written message, and the backlash fear many of us carry turned out to be mostly noise. Working together across differences is a skill in its own right, and our guide on how to collaborate more effectively digs into it.
A quick experiment: before you make your main point, name one value you genuinely share with the other person, such as fairness, safety or wanting the kids to do well.
7. Leave your résumé out of it
The same Naunov study had a second finding worth its own item. The least effective persuaders were the ones who emphasized their own political knowledge or engagement, alongside those who bombarded readers with facts.
Why would expertise backfire? Follow the chain. When you lead with “I’ve read everything on this” or “I’ve been following this for years,” the listener hears a status claim. A status claim invites a status response, so their attention shifts from your argument to defending their own competence. Now they’re protecting their ego instead of considering your point, and you’ve turned a shared problem into a ranking contest. This is a mechanism inferred from the results, not something the study measured directly, so hold it lightly.
There’s a balance here. Credentials matter when someone asks for them or when the topic is technical. They just shouldn’t be your opening move. Showing humility (item 5) tends to earn the authority that announcing it doesn’t. Our piece on how to impress your boss at work makes a similar case for workplace credibility.
Give it a go: cut “as someone who knows a lot about this” from your vocabulary and let the quality of your reasoning speak instead.
8. Swap stories, not just arguments
Of all the findings in this article, this one has the deepest roots.
David Broockman of UC Berkeley and Joshua Kalla of Yale ran three preregistered field experiments, published in the American Political Science Review in January 2020. Some 230 canvassers had face-to-face conversations with 6,869 voters across seven U.S. locations. In the first experiment, conversations using arguments alone had no measurable effect on exclusionary attitudes toward immigrants. Otherwise identical conversations that added a nonjudgmental exchange of personal narratives did reduce those attitudes, and the change lasted at least four months. Two further experiments on transphobic attitudes replicated the narrative effect.
The effect sizes were small (d = 0.08 and 0.04), which is normal for real attitudes measured in the field. What matters is direction and durability. And the 2025 Porat study from item 3 found the same pattern five years later: personal stories carried the persuasion. This source is older than our usual cutoff, but it’s the foundational field test, and the newer work confirms it.
Notice the word “exchange.” You share a story, and you also invite theirs. When you next argue for something, tell one short, true story about a specific person affected, then ask, “Has anything like that ever happened to you or someone you know?”
9. Have one deep conversation instead of a dozen quick pitches
That research didn’t stay in journals. It became a training method called deep canvassing, and organizations have used it at scale.
According to a Common Dreams report republished by El Semanario, the Deep Canvass Institute, launched in 2021 as a project of the People’s Action Institute and the New Conversation Institute, trained people who went on to hold more than 245,000 deep-canvass conversations on climate, immigration and trans rights. In a 2022 experiment with partner groups in Missouri, Idaho and Georgia, these conversations raised agreement that all Americans should be able to vote by mail by 4.2 points and lowered the belief that mail voting contributes to fraud by 1.9 points, with the biggest shifts among moderates and independents. Rep. Ayanna Pressley’s team used the method in late 2021 on student debt cancellation and reported moving supporters about two points on a 1 to 10 scale. In North Carolina, the group Down Home NC held thousands of conversations backing Medicaid expansion, which the legislature passed in March 2023.
A caveat. These outcome figures come from the organizations themselves, reported through advocacy media, not from peer-reviewed papers, and the Medicaid vote had many causes. Still, the pattern matches the controlled studies, and the method’s whole design is slow: questions, stories, space to think.
Test it this week: pick one person you’d like to persuade and book a proper sit-down instead of making your point in a hallway or a text thread.
10. Tailor the conversation to the person in front of you
A generic case is written for nobody. The more you adjust to the actual person, the better you do, as long as you genuinely use what you know.
Francesco Salvi and colleagues at EPFL and Fondazione Bruno Kessler published a striking experiment in Nature Human Behaviour in 2025. Some 900 participants debated either a human or GPT-4 on a policy question. When GPT-4 was given basic facts about its opponent (age, gender, education, job, politics), it was more persuasive than humans in 64.4% of the debates where one side came out ahead, and it raised the odds of shifting someone’s opinion by about 81% compared with human-versus-human debates. Without that personal information, GPT-4 did no better than people.
The detail I find most useful is the one about humans. People who got the same personal information about their opponent gained little from it. The information was there; they just didn’t use it. So the lesson isn’t “be a chatbot.” It’s that tailoring works only when you consciously adapt your examples, priorities and language to the person. This is a big part of how to be charming too.
Here’s a simple way in: before a big conversation, list three things you know about the person, such as what they worry about, what they value and how they like information delivered, and pick your examples to match.
11. Let them argue themselves into it
The most convincing voice in anyone’s head is their own. That’s why asking people to generate reasons often beats handing them yours.
Why does self-persuasion work? A person who produces an argument feels ownership of it, so it doesn’t trigger the defensive “someone is trying to change me” alarm. The reasons also arrive in their own vocabulary and fit their own priorities. But there’s a catch, demonstrated by Rhia Catapano, Zakary Tormala and Derek Rucker in Psychological Science in 2019. When people generated counterarguments to their own view while imagining someone with opposing values, they were less persuaded by those arguments. When they imagined someone with similar values, attitude change went up. The arguments that persuade us are the ones that fit our values. This is a single paper and older than our usual window, kept because it shows the mechanism so clearly.
In conversation, that means prompting rather than lecturing, and framing the prompt in their values. Instead of “here are five reasons to hire a second engineer,” try “if we did hire someone, what would you hope it freed you up to do?” Then let them talk.
12. Bring up the strongest objection yourself
If you pretend the other side has no good points, the listener assumes you either don’t know them or are hiding them. Neither helps.
Jeff Conlin and colleagues at Penn State tested this with a hard audience: U.S. adults still unvaccinated against COVID-19. Their study, published in Health Communication in 2024, compared two-sided refutational “conversion” messages, which raise the opposing view and then answer it, with one-sided advocacy. The conversion messages significantly raised pro-vaccination attitudes compared with the control. They didn’t directly reduce resistance, though resistance fell indirectly. The pathway that mattered was homophily, meaning how similar the audience felt to the message source, rather than raw argument strength.
That second result ties back to item 6. Acknowledging the other side does more than look fair. It signals “I’ve been where you are,” which makes you feel like one of them. It was one study with written messages, so the conversational effect may differ. Next time, do this: say the best version of their objection out loud (“the strongest point against this is the cost, and it’s a real one”), then give your answer.
13. Bring specific, checkable evidence and keep it accurate
Item 7 warned against bombarding people with facts. So are facts useless? No. The difference is between a pile of numbers and specific claims that answer the question on the table.
One of the largest studies on this came out in Science in December 2025. Kobi Hackenburg of the Oxford Internet Institute and the UK AI Security Institute, with colleagues including Helen Margetts and David Rand, ran three experiments with 76,977 UK participants across 707 political issues, using 19 large language models as persuaders. Tweaking how the models were trained boosted persuasiveness by up to 51%, and prompting strategies added up to 27%. According to Oxford’s December 2025 summary, the single biggest driver was information density, meaning responses packed with fact-checkable claims, which alone explained about half of the explainable variation in persuasion. Conversations also beat static, one-way messages.
Then the warning. The same methods that raised persuasion lowered factual accuracy, based on checks of 466,769 claims. That’s the trap for humans too. When you’re straining to win, you start reaching for the shakier stat.
Your move: before an important conversation, prepare three specific, verifiable facts that address their main concern, and drop anything you can’t source.
14. Cut your weakest arguments
You’d think five reasons beat one. Often they don’t, because listeners average your arguments rather than add them up.
Magdalena Obermaier of LMU Munich and Thomas Koch of Johannes Gutenberg University Mainz ran four online experiments, published in Scientific Reports in September 2024. People read material with one strong argument, several weak ones, or both. In three of the four studies, adding clearly weak arguments diluted the strong one. In a health-check study, for example, the strong argument alone scored an average of 3.56 on a five-point attitude scale, while the same argument plus three weak ones dropped to 3.11. Adding moderately strong arguments, by contrast, increased persuasion (4.07). The authors conclude that “more is better” holds only when the extras are at least moderately strong.
The topics were low-stakes and unfamiliar, so the effect might differ when people care a lot. Still, it’s a useful editing rule. Your filler reason (“plus it’s trendy”) hands the other person the easiest target and drags down the average. This study sits just over two years old and is kept because nothing newer tests the effect this directly. Before you speak, rank your reasons and drop any you’d be embarrassed to defend under questioning.
15. Answer their actual reasons, not the ones you assume
Generic rebuttals bounce off. What moves people is a response to the specific reason they hold the belief.
Thomas Costello of American University, Gordon Pennycook of Cornell and David Rand of MIT tested this with one of the most stubborn groups imaginable: more than 2,000 people who believed a conspiracy theory. In their Science paper from September 2024, each participant first explained in their own words what they believed and why. Then GPT-4 Turbo held a personalized dialogue rebutting the participant’s own evidence. Belief dropped by about 20% on average, roughly one in four participants disavowed the conspiracy, and the effect lasted at least two months.
The study is a little over two years old. It stays in because the 2025 Hackenburg work points the same way (information tailored to the conversation persuades), and because it challenges the popular idea that facts never work on committed believers. The key detail is personalization. The facts were aimed at each person’s stated reasons, not at the theory in general. This is where your follow-up questions from item 2 pay off.
Start small: ask “what’s the main thing that convinced you?” and build your answer around that one point, ignoring the ten others you had ready.
16. Drop the minimizing and condescending words
Some words quietly tell the other person their view is silly, even when your argument is good.
Julia Minson’s 2026 Harvard Gazette excerpt names two families of them. The first is minimizing words such as “just,” “simply” and “merely,” as in “you just need to look at the data.” They imply the other person’s position is barely worth considering. The second is heavy reasoning language like “because” or “let me explain,” which can come across as lecturing in a heated exchange. Minson also recommends reframing toward the positive: same message, fewer emotionally loaded words. Her advice comes from her lab’s text analyses and experiments on receptive language, and it applies most in conversations where tempers are already up.
You won’t scrub these words entirely, and you shouldn’t try to talk like a hostage negotiator. Just notice them. Our guide on how to be more careful about what you say has more on catching words before they land.
A quick experiment: reread your last heated email or message and delete every “just,” “simply” and “obviously.” Check whether it sounds more persuasive or less.
17. Let your voice land when you’re sure
How you say a sentence changes whether people take it in. One small cue matters more than you’d guess: whether your voice goes up or down at the end.
Thomas Vaughan-Johnston of Cardiff University, with Joshua Guyer, Leandre Fabrigar, Grigorios Lamprinakos and Pablo Briñol, published three experiments in Personality and Social Psychology Bulletin (2026). Falling intonation at the end of a sentence signaled that the speaker was confident. It didn’t automatically win people over, though. Under moderate attention, a falling voice made listeners process the message more carefully, so they liked strong arguments more and weak ones less. A confident delivery magnifies the quality of what you say. It doesn’t replace it.
A separate October 2025 study in the Journal of Speech, Language, and Hearing Research, by Zhikang Peng, Chaoyi Wang and Xiaoming Jiang, found that voices morphed to sound more confident were judged more credible, while the speaker’s gender made no difference. That one had only 52 participants, so treat it as supporting evidence. Pair this with item 5 and the rule becomes: hedge your uncertain claims in words, and say your sure ones with a voice that drops at the end. More on nonverbal cues in our guide on how to improve body language in communication. Record yourself making your main point and check whether it ends on a question-like rise.
18. Keep the temperature down
Anger and persuasion don’t mix. The mechanism is fairly direct.
Jieyu Ding Featherstone and Jingwen Zhang ran an experiment with 609 participants, with data collected in 2018 and published in the Journal of Health Communication (covered by PsyPost in April 2021). Vaccine misinformation made people angrier, and that anger was linked to worse attitudes toward vaccines. Refutational messages lowered anger, and lower anger was linked to better attitudes. The authors suggested that correcting misinformation might work partly by calming people down. These were mediation links in one study, so the causal path is likely rather than proven. The study is older than our usual range and kept because it lays out the anger pathway clearly.
The step-by-step logic goes like this. An angry listener reads your words as an attack. An attack calls for defense, not reflection. Defense means hunting for flaws in what you say rather than weighing it. The 2025 Porat study from item 3 adds a piece: reducing defensiveness is exactly what good listening did. Keeping your own tone calm, and lowering theirs, keeps the door to reflection open. When the other person is already furious, our guide on how to deal with angry people is the place to start.
Give it a go: if you feel your pulse rising, say “let me think about that for a second” and take one slow breath before you answer.
19. Remind them they’re free to say no
Ending a request with something like “but it’s completely your call” is one of the best-known compliance techniques. The honest picture of its evidence is messier than the pop-psychology version.
Adrien Fillon and colleagues published a meta-analysis in Meta-Psychology in July 2023 covering 52 experiments with 19,528 participants. Overall, the “but you are free” technique showed a medium effect (g = 0.44), stronger face to face. But when they looked only at the seven studies with a low risk of bias, the effect shrank to g = 0.11 and was no longer reliable. The authors flagged low reproducibility across the field.
So why keep it? Because the downside is zero and the logic is sound for a different reason. Saying “it’s your decision” lowers the sense of being pushed, which links back to the defensiveness problem in items 3 and 18. Use it out of respect, not as a trick, and don’t expect miracles. This meta-analysis is over three years old and kept because no newer synthesis exists. Test it this week: close your next request with a genuine “and if it’s a no, that’s fine.”
20. Don’t use “explain how it works” as a gotcha
You may have read that asking someone to explain, step by step, how a policy works will humble them out of extreme views. It’s a neat idea. It doesn’t hold up well.
The original 2013 finding by Philip Fernbach and colleagues suggested that exposing people’s shallow understanding reduced political extremism. Richard Walker, Jiasheng Zhang and Edmund Cheng tested it again with 5,139 citizens in Hong Kong, publishing in Political Psychology in June 2024. Across three replications and extensions on more controversial issues, they found the opposite: positional extremity rose when people articulated their understanding of policy. Hong Kong’s postconflict setting may matter, and other replications have had mixed results. But it’s a real warning.
Asking someone to explain a mechanism can feel like a test they’re meant to fail. People who feel tested tend to dig in. Curiosity questions (“what would you change about it?”) work differently from quiz questions (“so how exactly does that work, then?”). The Walker study is a little over two years old and is kept because it’s the largest direct test. If you catch yourself asking a question you already know they can’t answer, swap it for one you genuinely want answered.
21. Think of it as a dance, not a speech
Picture two people learning to dance together. If one drags the other across the floor, both look awkward and someone gets their toes stepped on. If one leads gently and keeps adjusting to how the partner moves, they get somewhere together.
Persuasion in conversation works more like the second scene. You make a point, watch how it lands, adjust, make the next one. The Hackenburg study from item 13 supports the core of this: across tens of thousands of participants, back-and-forth conversation was more persuasive than one-way static messages. The analogy breaks down in one important way. In a dance, the leader decides the destination, while in a good conversation you have to stay open to the possibility that your partner is right and you’re the one who should move. Treat that as part of the deal, not a weakness. Our guide to improving emotional intelligence helps with reading those cues.
Here’s a simple way in: after each main point you make, stop and ask a short check-in question, such as “does that match what you’ve seen?” before moving on.
22. Protect the relationship so you get a second conversation
Real persuasion is cumulative. One conversation plants a doubt, a later one gives it room, a third finds them saying your point back to you as if it were theirs. That only happens if there is a later conversation.
The chain works like this. Attitudes rarely flip within a single talk, as items 1 and 8 showed. Durable change, when it happens, often builds through repeated, low-pressure exposure to a view from someone you trust. Trust needs contact. So anything that ends contact ends persuasion, which makes the relationship itself the most valuable asset you have in a disagreement. That’s a reasoned mechanism, not a measured one, but the survey data shows the risk is very real.
In Pew’s December 2025 survey, 56% of U.S. adults said they had stopped talking to someone about political or election news because of something that person said, up from 45% in 2024. Among liberal Democrats, the figure was 77%. Whatever your politics, that’s a lot of conversations ending for good. When things get heated, end the conversation warmly and on purpose, with something like “I don’t think we’ll settle this today, but I’m glad we talked, and I’d like to pick it up again.”
What to do when you slip
You will slip. You’ll interrupt, sneer at a weak point, say “obviously” in exactly the wrong tone, or press so hard the other person goes quiet. Everyone does. What matters is what you do with the next sentence.
There’s encouraging evidence that one good response can change the direction of a conversation. Michael Yeomans, Julia Minson, Hanne Collins and Charles Dorison published five studies with a combined 17,184 participants in the Journal of Personality and Social Psychology in August 2024. Receptiveness turned out to be contagious. When one person was trained to be receptive, their political opponent became more receptive too, and even a single receptive response improved how people judged each other. The researchers found this wasn’t simple mimicry. It was a deeper shift in conversational style. The paper is a little over two years old and is kept because it’s the most direct test of recovery within a disagreement.
So when you notice you’ve gone off the rails:
- Name it briefly. “Sorry, that came out more dismissive than I meant.”
- Go back to item 4 and restate their point properly.
- Lower the stakes. Remind yourself (and them, if it helps) that nobody has to decide anything today.
- Don’t spiral afterwards. One bad exchange doesn’t make you a bad communicator, and replaying it all night helps nobody. If you tend to do that, our guide on how to stop catastrophizing is worth a read.
Then try again next time. Persuasion is a skill, and skills come with clumsy reps. Working on your social skills in low-stakes moments makes the high-stakes ones easier.
Conclusion
Look back over these 22 ideas and a pattern shows up. The research rewards behavior that most of us would call good manners plus homework: ask real questions, listen, restate fairly, admit what you don’t know, bring a true story and a few solid facts aimed at their actual concern, and keep the relationship alive for another round. The tricks people associate with persuasion, from fake urgency to fact-dumping to flexing credentials, either don’t work or work only until someone notices.
You don’t need to use all of it at once. Pick two habits, maybe follow-up questions and restating their view, and practice them in your next three disagreements. Then add another.
If you’d like more practical, evidence-based guides like this one, browse our self improvement section and pick your next skill to work on.
