Short answer
Flirting is warm attention plus playfulness plus a clear, low-pressure signal of interest. Research on creepiness (McAndrew and Koehnke, 2016) points to ambiguity and unpredictability as the triggers, so be clear and make it easy for her to leave. Compliment choices, not her body, escalate only when she reciprocates and treat a soft no as a no.
Key takeaways
- People are unsettled by ambiguity and by not being able to leave, so plain words and an easy exit do most of the work.
- Flirting is warm attention plus light play plus a clear, low-pressure signal of interest.
- Compliment something she chose, say it once and leave her body out of it.
- Signs of interest are unreliable: in one lab study most real flirting went unnoticed, and men tend to read friendliness as interest.
- A soft no is a no, signs never equal consent, and at work, at the gym and with staff the bar is much higher.
Start the free meet-someone-new plan

What actually makes a man come across as creepy
Interest is not what reads as creepy. People feel creeped out when they cannot tell whether someone is a threat and cannot easily get away from him.
The first half of that is the conclusion of what its authors describe as the first empirical study of creepiness. The second half comes from the same survey, where making it hard to leave a conversation was one of the behaviors people flagged.
What the research says
McAndrew and Koehnke (2016) surveyed 1,341 people online (1,029 women and 312 men, aged 18 to 77). Participants imagined a trusted friend describing someone they had just met as creepy, then rated how likely that person was to show each of 44 behaviors and traits. Among the items rated most likely were watching someone before interacting with her, asking to take her picture, asking for details of her personal life and steering the conversation toward sex. Participants also strongly agreed that such a person makes them uncomfortable because they cannot predict how he or she will behave (4.33 on a 5-point scale), and 95.3 percent said a creepy person is more likely to be male. The authors conclude that feeling creeped out is a response to ambiguity about whether there is a threat [1]. Limit: an exploratory survey of a self-selected sample that was about three-quarters women, rating an imagined person, not observed behavior. Verdict: Supported, as a description of what people say unsettles them.
Almost everything on that list is either ambiguous or hard to escape. He watches but does not speak. He asks things he has no reason to know. The same survey also flagged standing too close, steering relentlessly toward one topic and making it nearly impossible to leave the conversation without seeming rude [1]. So do not hide your interest. Be plain about it and leave the door open.
| What unsettles people | What you do instead |
|---|---|
| Watching before approaching | Say hello early or not at all. Never hover |
| Standing too close | Keep a normal arm's length and let her set anything closer |
| Making it hard to leave | Keep it short, stand where she can walk away and end it yourself |
| Steering the talk toward sex | Keep sexual comments out unless she has clearly gone there first |
| Asking for personal details or a photo | Let her volunteer where she lives, where she works and who she is with |
| Behavior she cannot predict | Say what you mean in plain words and stay consistent |
Only 8.6 percent of participants thought creepy people know they are creepy, so your own sense of how you come across is not a reliable check [1]. Use rules you can verify, and watch her response.
Part of the list was grooming: unkempt hair, greasy hair and dirty clothes were all on it [1]. Some items were plainly unfair, like very pale skin or bags under the eyes. Clean clothes and a haircut are the part you control, and the guide on how to be more attractive covers them.
The flirting formula: warmth, play, a clear signal
Flirting is warm attention, plus a little playfulness, plus a clear and low-pressure signal that you are interested. Leave out the signal and you are just being friendly. Leave out the warmth and play and it is a proposition.
Warmth
Give her your full attention. Phone away, eye contact that breaks naturally, her name, and a follow-up on something she said ten minutes ago. If you want the mechanics, the loop in how to talk to women is the foundation.
Play
Play is light humor about the situation or about yourself, and the occasional tease about something she chose and clearly enjoys, like her loyalty to a terrible football team. It is never about her looks, her intelligence or anything she cannot change. Negging, the insult dressed up as a joke, has no place here. It is unkind, and it is exactly the unpredictable behavior that puts people on guard. If she does not play back, drop it.
How to show her you are interested
Say it in words. Hints fail because flirting is detected poorly even when it is real [2], and a vague signal leaves her guessing about what you want. A clear sentence with an easy exit is the opposite of ambiguity with no way out.
Example. "I've liked talking to you. I'd like to take you for a coffee this week, if you're up for it. No pressure either way."
The full method, including what to say if she declines, is in how to ask a girl out.
How to compliment without making it weird
Compliment something she chose, say it once and expect nothing back. Leave her body out of it.
| Works | Skip |
|---|---|
| "That jacket is great. Where did you find it?" | Anything about her body or how her clothes fit |
| "You tell a story really well." | "You're prettier than your photos." |
| "I like how seriously you take your coffee." | The same compliment repeated until she responds |
| "Good taste," after she names a band | A compliment followed straight away by an ask |
People overestimate how awkward a simple compliment will be. In the first of five studies by Boothby and Bohns (2021), 190 people took part on a university campus: givers were sent to tell a stranger of the same gender that they liked their shirt. Givers underestimated how good the compliment would make the other person feel, and in later studies they overestimated how bothered and uncomfortable the person would be [3].
Read the limit carefully. Those were same-gender compliments about clothing with nothing romantic at stake. They support the kind of compliment in the left-hand column: brief, about a choice, no agenda. They tell you nothing about a man commenting on a stranger's body, and the creepiness survey gives you a fair idea of how that lands.
On delivery: say it, smile and move on to something else. A compliment that waits to be paid back is pressure.
Signs she likes you: how unreliable they are
Signs of interest exist, but they are weak evidence, and men in particular tend to read too much into friendliness. No list of signs can tell you what she wants.
What the research says
Hall, Xing and Brooks (2015) had 52 pairs of heterosexual strangers talk for 10 to 12 minutes, then asked each person whether they had flirted and whether they thought their partner had. Conversations with no flirting were judged more accurately than conversations where someone was flirting [2]. The University of Kansas release on the study gives the figures: more than 80 percent accuracy when there was no flirting, but only 36 percent of men and 18 percent of women recognized flirting that was really taking place [4]. Limit: one lab study of college students, with flirting self-reported, and the percentages come from the press release, not from the paper. Verdict: Supported, by one study. In it, most real flirting went unnoticed.
The opposite error is better documented. In Abbey's 1982 experiment, an unacquainted man and woman talked for five minutes while a hidden man and woman watched. The men, both those talking and those watching, rated the woman as more seductive and promiscuous than the women did [5]. A 2008 review concluded that this pattern, men reading behavior more sexually than women do, has been replicated in scores of studies, most of them with White, heterosexual US college students [6]. In a speed-meeting study of 199 people, men who rated themselves as more attractive, or who were more oriented toward short-term relationships, were more likely to overestimate a woman's interest. How far the men overestimated was predicted by how physically attractive the woman was. Women tended to underestimate men's interest [7]. Trust your read least when you most want it to be right.
There is a reason the two get confused. A meta-analysis of 309 effect sizes from 5,422 participants found that attraction goes with eye contact, smiling, laughter, mimicry, standing closer and talking to the person. The overall link was modest, and the authors note that these are the same behaviors research has tied to building trust and rapport [8]. A woman who smiles and holds your eye may be interested. She may also be a warm person having a good day.
So you cannot detect your way to certainty. Stop collecting signs to justify a move. Make a clear, low-pressure invitation and let her answer in words.
The signals worth noticing, and the ones that mean stop
Treat positive signals as an invitation to keep talking, never as permission for anything more. Treat stop signals as final the first time.
| Worth noticing | Means stop |
|---|---|
| She asks you questions back | Short or one-word answers |
| She stays when she has an easy chance to leave | She turns away, steps back or looks around for her friends |
| She turns toward you or moves a little closer | Headphones go back in or the phone comes out |
| She brings up something you said earlier, or something you could do another time | "I should get back", "I have a boyfriend", "maybe" or a change of subject |
| She starts a touch, like a hand on your arm | She goes quiet or stiff, or laughs without joining in |
Move one small step at a time, and only after she has matched the last one. If she moved closer, you can stay close. If she touched your arm, a brief touch on her arm in return is reasonable. Keep touch rare and short, and never be the one who starts it with a stranger: touching someone frequently was one of the highest-rated items in the creepiness survey [1]. Signs never equal consent. Her number, a date, a hand held, a kiss: each needs its own yes, and when in doubt you ask out loud.
A soft no is a no. People find it hard to refuse directly, so the refusal often comes as a hint. In a survey of 942 STEM graduate students, 277 described a specific unwanted or unreturned advance: 176 as the person pursued and 101 as the pursuer. Among those who answered the questions about what happened next, 52 percent of the pursued said they tried to avoid the person afterward, while only 7 percent of pursuers thought the other person had done so. About 23 percent of those pursued said they had felt obligated to agree, more than four times the share pursuers assumed [9]. The data are recollections from labs and departments, so the exact numbers will not carry over to a bar. The gap is what matters: the person asking saw far less discomfort than there was.
Her discomfort is never something to push through. If you sense it, give her space and end on a friendly line. If you misread her and she was interested, she can tell you. That mistake is cheap. The other one is not.
How to flirt over text
Flirt over text the same way at lower volume: warm, lightly playful and clear about wanting to see her. Text strips out tone and facial expression, so the ambiguity that puts people on edge goes up. No study we could confirm tests flirting by text, so treat this section as practical advice.
- Refer to something real. A detail from your conversation beats any line.
- Tease lightly and rarely. Without your face, a joke at her expense reads as an insult.
- Keep it clean. Nothing sexual unless she has clearly gone there first, and never a photo she did not ask for.
- Match her length and pace. Three paragraphs to her one line is pressure.
- Move toward a plan. Within a few messages, suggest a day and a place.
Example. "Still thinking about your claim that pineapple belongs on pizza. I'm willing to be proven wrong. Thursday?"
Send when: you have talked in person, she gave you her number herself and her replies have had some energy.
Do not send when: she has not answered your last message, her replies are one word, or she has said no or asked for space.
If she stops replying, one light follow-up after a few days is the limit. The guide on being left on read explains what to send and when to stop.
Work, gym and service settings: the rules change
Where she cannot easily leave, or where you hold any power over her, the bar for flirting rises sharply. Sometimes the answer is simply don't.
At work
Never flirt with someone you manage, hire, grade or evaluate. A no costs her more than it costs you. In the Bohns and DeVincent survey, people on the receiving end worried more about the professional fallout of refusing than their pursuers imagined [9]. With a peer, read your employer's policy first, make one clear invitation outside work, and treat anything short of a clear yes as a permanent no. The US Equal Employment Opportunity Commission lists unwelcome sexual advances as a form of harassment and notes that the harasser can be a supervisor, a coworker or a customer [10]. If you are unsure where the line is, ask HR. This guide is not legal advice.
At the gym
She is there to train. Headphones mean no. Never interrupt a set, never comment on her body or her form, and never stand watching. Watching someone before interacting was the highest-rated behavior in the creepiness survey [1]. If you see the same person for weeks, a nod can grow into a hello and later into a sentence at the water fountain. Let months of normal behavior do the work. If she keeps it at a nod, so do you.
With staff
A barista, server or bartender is paid to be warm and cannot walk away from you. Her friendliness is the job, not a signal. Do not ask her out while she is working. Be a good regular and leave it there.
Confidence is evidence: the four-rung practice ladder
Confidence around women comes from a record of evidence that conversations usually go fine. You build that record one rung at a time.
The closest test of this we found is a week-long experiment. Sandstrom, Boothby and Cooney (2022) had 198 university students in the US and the UK talk to strangers every day for five days, while 88 others only observed strangers. On day one, only 40 percent of the talkers expected the first person they approached to talk to them, yet 92 percent of first approaches that day became a conversation. By the end of the week the talkers expected less rejection and felt more able to hold a conversation, and the change held a week later [11]. These were everyday chats by students, not flirting, so the study cannot tell you how a romantic approach will go. What should carry over is the mechanism: repetition replaces a guess with evidence.
- Rung one: ten seconds of warmth. Eye contact, a smile and a real hello or thank-you to people you pass in a normal day. Five a day for a week. Men, women, any age.
- Rung two: two-minute conversations. With anyone, anywhere talking is normal, using notice, ask, follow up and share.
- Rung three: warmth and play. In social settings, with women you would like to know: one compliment about a choice, one light joke, then you end the conversation first.
- Rung four: the clear signal. Tell one woman you have enjoyed talking to that you would like to see her again, with an easy out. If she says no: "No problem. Good to meet you."
Move up when the current rung has become boring. If you are shy, stay on the first two as long as you need. Our guide on where to meet women lists the settings where rungs two to four are welcome.
Do this today
Do rung one on your next trip out of the house. Give five people eye contact, a smile and a real hello, and expect nothing back. Thank the cashier and mean it. It adds no time to your day, and it is the first entry in your record.
Then make it a habit: Start the free meet-someone-new plan. The dating path builds from Say hello first (hello and one easy question, wherever you are) to Ask her out clearly (one message with a day, a place, a time and an easy out), then to Run a great first date. One move at a time. More guides on attraction, conversation and dating are in Meet someone new.
Questions men ask
How do I know if I'm being creepy?
You often cannot feel it from the inside: in a survey of 1,341 people, only 8.6 percent thought creepy people know how they come across. So check your behavior, not your intentions. Are you watching without speaking, standing close, asking personal questions, steering toward sex or making it hard for her to leave? Is she giving short answers or turning away? If the answer to any of these is yes, step back and end it politely.
Is she flirting or just being friendly?
Usually you cannot tell, and neither can most people. In a University of Kansas study of 52 pairs of strangers, only 36 percent of men correctly spotted flirting, according to the university's news release. Decades of research also suggest men tend to read friendliness as interest. Smiling and eye contact go with ordinary warmth too. Stop decoding. If you have had a good conversation, make one clear, low-pressure invitation and take her answer at face value.
How do I flirt if I'm shy?
Start smaller than flirting. Spend a week on eye contact, a smile and a hello to people you pass, then move to two-minute conversations with anyone. In a week-long study of 286 students, daily practice talking to strangers lowered their fear of rejection, and the change held a week later. When you do flirt, you need only three things: full attention, one light joke and a plain sentence saying you enjoyed talking to her.
Is it creepy to talk to a woman at the gym?
It can be, because she is there to train and cannot leave without cutting her workout short. Never interrupt a set, never comment on her body and never stand watching. Headphones mean no. If you see the same person for weeks, a nod can become a hello, and a hello can become a sentence at the water fountain. If she keeps it at a nod, so do you.
How do I show interest without being creepy?
Say it plainly and make it easy to decline: 'I've liked talking to you. Want to get a coffee this week? No pressure either way.' Clear words remove the ambiguity that unsettles people, and the easy out removes the trapped feeling. Say it once, at a normal distance, somewhere she can walk away. If the answer is anything short of yes, smile, say 'no problem' and do not ask again.
Sources
Every source below was opened during drafting and review. The note under each one says who was studied and what it cannot show.
McAndrew, Koehnke (2016). On the nature of creepiness. New Ideas in Psychology, 43, 10-15.
An online survey of 1,341 people (1,029 women, 312 men) recruited by snowball sampling; exploratory, self-selected and about an imagined creepy person, not observed behavior.
Hall, Xing, Brooks (2015). Accurately Detecting Flirting: Error Management Theory, the Traditional Sexual Script, and Flirting Base Rate. Communication Research, 42(7), 939-958.
52 pairs of heterosexual strangers (104 people) who talked for 10 to 12 minutes, plus 261 observers of video clips; a lab setting with self-reported flirting, first published online in 2014.
Boothby, Bohns (2021). Why a Simple Act of Kindness Is Not as Simple as It Seems: Underestimating the Positive Impact of Our Compliments on Others. Personality and Social Psychology Bulletin, 47(5), 826-840.
Five studies; in the first, 190 people on a university campus took part in compliments about clothing between strangers of the same gender, so nothing romantic was tested.
University of Kansas (2014). Flirting hard to detect, researcher says. KU News, University of Kansas.
The university's news release on the Hall study of 52 pairs of college students, used here for the accuracy percentages; a press release, not the paper itself.
Abbey (1982). Sex differences in attributions for friendly behavior: Do males misperceive females' friendliness? Journal of Personality and Social Psychology, 42(5), 830-838.
A lab experiment in which an unacquainted man and woman, both college students, talked for five minutes while a hidden man and woman observed; one campus in 1982, and the abstract does not report the sample size.
Lindgren, Parkhill, George, Hendershot (2008). Gender Differences in Perceptions of Sexual Intent: A Qualitative Review and Integration. Psychology of Women Quarterly, 32(4), 423-439.
A narrative review of published studies on how men and women perceive sexual intent; most participants were White, heterosexual US college students giving questionnaire ratings.
Perilloux, Easton, Buss (2012). The misperception of sexual interest. Psychological Science, 23(2), 146-151.
199 participants in a speed-meeting study who rated their own interest and their partners' interest; brief staged meetings measured by self-report.
Montoya, Kershaw, Prosser (2018). A meta-analytic investigation of the relation between interpersonal attraction and enacted behavior. Psychological Bulletin, 144(7), 673-709.
A meta-analysis of 309 effect sizes from 5,422 participants; the link between attraction and behavior was modest and the behaviors overlap with ordinary rapport, so no single behavior identifies interest.
Bohns, DeVincent (2019). Rejecting Unwanted Romantic Advances Is More Difficult Than Suitors Realize. Social Psychological and Personality Science, 10(8), 1102-1110.
A survey of 942 STEM graduate students, 277 of whom described a specific advance (176 as the person pursued, 101 as the pursuer), plus a vignette experiment with 385 people; recalled and imagined advances in academic and work settings, not observed behavior.
U.S. Equal Employment Opportunity Commission (2026). Sexual Harassment. U.S. Equal Employment Opportunity Commission.
An official US agency guidance page with no publication date, read in November 2026; general information about US workplaces, not a study and not legal advice.
Sandstrom, Boothby, Cooney (2022). Talking to strangers: A week-long intervention reduces psychological barriers to social connection. Journal of Experimental Social Psychology, 102, 104356.
286 university students in the US and UK (198 talked to strangers daily for five days, 88 only observed them) and 1,336 conversations; students only, self-reported, and the conversations were everyday chats, not flirting.