Why Most Would-Be Creators Never Press Go Live, and What Rehearsal Can Fix

On Thursday, the Streamer Awards will hand out its trophies live on Twitch and YouTube. Kai Cenat, IShowSpeed and Jynxzi lead the nominations with four each, according to Dexerto, and the categories run from best marathon stream to a Rising Star award for newcomers. The next morning, TwitchCon opens at the San Diego Convention Center for three days of tournaments, meetups and stage shows. Twitch itself calls it the biggest weekend in livestreaming.

Tucked into Saturday's schedule, between the tournaments and the meet-and-greets, is a one-hour session called "IRL Streaming for Beginners: It's Time to Touch Grass." KPBS, the San Diego public broadcaster, describes it in its guide to the weekend as a session for "anyone who might be nervous about streaming IRL." Earlier that same day, on another stage, a panel called "Stop the 'Trolls'" promises real stories and resources for dealing with harassment on stream.

Those two rooms say something the trophies don't. Everyone who makes it to a convention about streaming already streams. Behind them stands a much larger group that never shows up in any nomination list or analytics dashboard: people who installed the app, set up a lamp, opened the live screen, looked at their own face in the preview, and closed it again. Nobody counts them, because there is nothing to count. A stream that never starts leaves no data.

The standard explanation for that group is a courage gap, and the standard advice follows from it: just press the button, nobody is watching anyway, you will get used to it. This piece argues that the explanation is wrong in a way that matters. The research on performing in front of others, which runs from racing cyclists in 1898 to cockroaches in 1969 to a room full of avatars in a London lab in 1998, points to a different diagnosis. Camera fright on live video is mostly a rehearsal problem. And since the summer of 2025 the big platforms have made rehearsal harder while making the first live more exposed than it has ever been. What follows is the mechanism, the evidence, what the platforms changed, one new kind of rehearsal room built from AI, and a practical ladder for anyone whose thumb has hovered over Go Live and pulled back.

Two fears share one button

The fear of going live is really two fears at once, and they pull in opposite directions.

The first is the empty room. You start talking, the viewer counter reads zero, and you are now a person speaking alone into a phone in your kitchen. The scale of live video makes this the likeliest outcome for a beginner. The analytics firm Streams Charts, in a 2026 analysis of Twitch, counted about 3.2 million unique creators streaming each month and, on average, more than 41,000 channels live at any one time. Third-party trackers have long put the large majority of those channels in single digits of average viewers, although the exact share depends on who is counting and how. A new stream on any platform starts at the bottom of a very deep list.

The second is the hostile room: the stranger who arrives only to mock, the comment about your face, the coordinated pile-on. This is not an imagined risk. Pew Research Center's survey of 10,093 US adults, published in January 2021, found that 41 percent had personally experienced online harassment and 25 percent had experienced its more severe forms. Streaming has its own version. In the late summer of 2021, Twitch creators staged a walkout called #ADayOffTwitch over "hate raids," in which bot accounts flooded marginalized streamers' chats with slurs. Twitch then sued two users it accused of running them, alleging, as TechCrunch and Tubefilter reported, that one of them was linked to some 3,000 bot accounts.

Underneath both sits an older, more ordinary fear. In the National Comorbidity Survey Replication, a nationally representative US survey of 9,282 adults carried out between 2001 and 2003, public speaking or performing was the most common lifetime social fear, reported by 21.2 percent of respondents. The analysis, by Anne Marie Ruscio, Ronald Kessler and colleagues in Psychological Medicine in 2008, also found that 11.8 percent feared simply working while being watched. A livestream is both of those at once: you perform, and you work a camera, a chat and your own nerves, all in public.

An empty room tells you nobody cares. A hostile room tells you somebody does, in the worst way. The button never says which one you are about to get.

What an audience actually does to a performer

What is often called social psychology's first experiment is about exactly this. In 1898 the psychologist Norman Triplett noticed in racing records that cyclists posted faster times when they rode against other cyclists than when they rode alone against the clock. Something about other people's presence changed the performance.

For the next six decades the findings kept contradicting each other. Sometimes an audience helped. Sometimes it made people clumsier and slower. In 1965 the psychologist Robert Zajonc proposed the explanation that still anchors the field, in a short paper in Science titled "Social Facilitation." The literal mechanism goes like this: the presence of others raises arousal, and arousal strengthens whatever response is already dominant, meaning the thing you would most likely do anyway. If you have practiced something until it is automatic, the dominant response is the right one, and an audience makes it crisper. If the skill is new, the dominant response is a mistake, and an audience makes you more likely to make it.

Zajonc then tested it on an animal that cannot possibly care what anyone thinks of it. In a 1969 study with Alexander Heingartner and Edward Herman, cockroaches fled a bright light either down a simple straight runway or through a maze with a turn in it. Some ran alone. Some ran past an "audience" of other cockroaches sitting in clear boxes beside the track. The paper reported that the audience sped the roaches up on the easy runway and slowed them down in the maze.

The broader record is more modest than that tidy story. Charles Bond and Linda Titus pulled together 241 studies in a 1983 review in Psychological Bulletin and concluded that the effects are real but small: others' presence slightly helps simple performance and slightly impairs complex performance. And in 2020 a team led by Emma Halfmann, with Janne Bredehöft and Jan Alexander Häusser, ran a preregistered replication of the cockroach study for the journal Psychological Science. Their roaches ran slower with an audience present on both tasks. The researchers found no sign that the audience helped on the easy runway.

The half of the theory that says an audience can hurt you is the half that held up.

Here is the image worth carrying through the rest of this piece. A theater company does not send a play from the script to opening night. There is a technical rehearsal, where the lights, cues and sound get tested with nobody watching. Then there is a dress rehearsal, a full run with everything exactly as it will be on the night, sometimes in front of a few friendly invited faces. By opening night, the actors' dominant response is the show itself. A beginner's first livestream, by contrast, is opening night with no dress rehearsal. The mistakes it brings out are not a measure of character. They are what an audience does to a skill that has not been run yet.

The comparison is useful, but it undersells one thing. A theater audience arrives on schedule and sits quietly in the dark. A live audience talks back, in writing, at the exact moment you are trying to talk, and it can leave with a swipe. "Speaking while reading a scrolling chat and deciding what to answer" is a skill with no real stage equivalent, and nobody is born with it.

The turn: courage is the wrong diagnosis

If going live were mainly a courage problem, trying harder would fix it, and the people who flinch would simply be the timid ones. The research describes something else. The audience amplifies whatever is already there. Arriving with a new skill and an audience, you get the new skill's worst version, however brave you are.

So the job is not to summon nerve. The job is to make talking to a live chat your dominant response before strangers are watching. That is a rehearsal job, and it needs a rehearsal room.

Nobody is naturally good at talking to a scrolling wall of strangers. It is a skill, and skills get practiced somewhere.

Talking to people who were never there

In the summer of 1998, three computer scientists at University College London, Mel Slater, David-Paul Pertaub and Anthony Steed, set out to answer a question that sounds almost silly. What happens when someone gives a talk to an audience they know is entirely computer-generated?

They built a virtual copy of the real seminar room where their volunteers filled in questionnaires, and they seated eight avatars in a semicircle facing the speaker. The avatars blinked, twitched and nodded so they would not look frozen. They could follow the speaker with their eyes, show six facial expressions, clap, yawn loudly, fall asleep, and stand up and walk out across the speaker's line of sight. Volunteers gave short talks to this audience in two moods, attentive and appreciative or bored and hostile, with the order varied.

The answer, published in IEEE Computer Graphics and Applications in 1999, was that people reacted to these obviously fake listeners much as they would to real ones. Speakers who felt the audience was interested rated their own performance higher. The sense of being with the audience, which the team called co-presence, amplified the effect in both directions, making a bad audience feel worse and a good one feel better. The authors' conclusion was that human subjects "do respond appropriately to negative or positive audiences, even when these are entirely virtual." They also reported, honestly, that in that small pilot of ten people they saw little evidence of a rehearsal effect across successive talks.

A larger follow-up, published in 2002 in the journal Presence with 40 participants, found that the hostile audience produced a strong emotional response in speakers whether or not they normally feared public speaking. The bored avatars got under people's skin even though everyone in the room knew they were not people.

That discovery opened a clinical line of research. In 2013 a team led by Page Anderson at Georgia State University ran what it described as the first randomized trial comparing virtual reality exposure therapy with in-person exposure for social anxiety. Its 97 participants all named public speaking as their main fear. After eight sessions, both treatments beat a waitlist, the two did equally well, and the gains held at a 12-month follow-up. A 2022 meta-analysis in Behavior Modification, by Rachel Reeves and colleagues, pooled 11 studies with 508 participants and found large reductions in public speaking anxiety from both virtual and in-person exposure, with in-person marginally ahead.

Those are clinical studies of structured treatment delivered with professionals, and their results do not transfer automatically to any app. What they establish is narrower and still important: an audience that is not real can still be a real enough rehearsal for the nervous system to take it seriously.

How the platforms raised the stakes

If rehearsal is the answer, the obvious question is where a beginner is supposed to do it. Since the summer of 2025, the honest answer has become: not in the places that used to work.

Start with Instagram. In early August 2025, Instagram began telling users that "only public accounts with 1,000 followers or more will be able to create live videos," as Engadget, TechCrunch and others reported at the time. Meta said the requirement applies to private accounts too, which removes the most natural practice room there was: a private account, a handful of friends, a first wobbly broadcast nobody else would ever see. Engadget noted that the change could also end the feature, added in 2024, that let users go live for just a few close friends. Meta framed the rule as improving the Live experience and did not say it was temporary, and we have not seen it reversed. For a new account, the path to Live now runs through earning a real audience first, the slow way our guide to getting real Instagram followers describes, with no shortcut worth taking.

YouTube's threshold is lower. Its help center lists at least 50 subscribers for mobile live streaming, along with a verified channel. It also offers something genuinely useful: a Practice Mode that puts you in a private session, in YouTube's words, with "no real audience." That is a technical rehearsal, and a good one, for framing, lighting, sound and connection. It is not a dress rehearsal, because there is nobody to talk to. TikTok's LIVE access is widely reported to require being 18 and having 1,000 followers in most regions, with variation by country, one of several differences we covered in Instagram vs TikTok.

Meanwhile, what happens to a stream after it starts has become more public. On October 20, Twitch relaunched Stream Recaps, which turn clips from your last broadcast into a reel shown in a shelf at the top of followers' mobile feeds and kept on your channel page. Net Influencer reported that recaps can publish automatically about an hour after a stream ends, unless the streamer turns that off, and a companion program called Recaps Spotlight, starting October 27, promotes selected clips on Instagram and YouTube. Two days later, on October 22, YouTube began rolling out a new top bar on its mobile Home tab that includes a Live feed of recommended streams alongside streams from subscribed channels, according to its own help pages as reported by PPC Land and Upstream.

None of this is sinister. Platforms want live video to be found, and they want creators to stay, the same logic behind Instagram's push into long-form video and the deals that came with it. But from the beginner's side of the screen, the direction is clear. The first live is more likely to reach strangers, more likely to be clipped and replayed, and harder to do in a small private room. The stakes went up and the rehearsal space went down at the same time. If the ranking systems behind all this are a mystery, our running notes on how the Instagram algorithm changed in 2026 are a reasonable place to start.

The industry has built a bigger stage for beginners and quietly taken away the rehearsal room.

Three tricks that survived the lab

Rehearsal is the main fix, but psychology has turned up a few smaller ones that cost nothing and have decent evidence behind them.

The first is knowing about the spotlight effect. In a series of studies published in 2000 in the Journal of Personality and Social Psychology, Thomas Gilovich of Cornell, with Victoria Medvec and Kenneth Savitsky, asked students to walk into a room of peers wearing a T-shirt with an embarrassing image (in the best-known version, the singer Barry Manilow). By most accounts of the study, the wearers guessed that about half the room would notice; far fewer, roughly a quarter, actually could say what was on the shirt. The paper's explanation is that people anchor on their own vivid experience of a moment and do not adjust enough for everyone else's much duller one. The stumble you will replay for a week is, for most viewers, a blur.

The second is the illusion of transparency. In a 2003 study in the Journal of Experimental Social Psychology, Savitsky and Gilovich found that speakers rated themselves as visibly more nervous than their audiences rated them. Then they tried an intervention. Speakers who were simply told about this bias, that nervousness is much less visible than it feels, went on to give speeches that both they and observers rated as better than those of speakers who got no explanation or plain reassurance. Knowing the audience cannot see inside you turns out to be a useful thing to know.

The third is the most counterintuitive. Alison Wood Brooks of Harvard Business School found that most people believe calming down is the best way to handle pre-performance nerves, then tested the alternative of relabeling the arousal as excitement. In her 2014 paper in the Journal of Experimental Psychology: General, 113 people sang a rock song on a karaoke video game after saying a single sentence out loud. Those who said "I am excited" scored about 80 percent on the game's own measure of pitch, rhythm and volume, against about 53 percent for those who said "I am anxious," with the calm, angry and sad groups averaging around 69 percent. In a separate experiment, 140 people giving a persuasive speech on camera were rated more persuasive, competent and relaxed when they first said "I am excited" rather than "I am calm." Her reasoning is that anxiety and excitement are both high-arousal states, so switching the label is easier than forcing the body down to calm.

The racing heart before a live is not a warning to stop. It can be read as fuel, and reading it that way measurably helps.

Myths worth retiring

"People fear public speaking more than death." It is a great stand-up line and a weak statistic. The best large survey data, the US national survey above, puts a lifetime fear of public speaking or performing at about one adult in five. That makes it the most common social fear in that survey, which is plenty. It is not universal, and treating it as universal can make a beginner feel the fear is a permanent fact rather than a skill gap.

"A tiny stream is low stakes, because nobody is watching." It used to be closer to true. With auto-published recaps, live feeds on home screens and clips promoted on other networks, a small stream can travel further than its live viewer count suggests. Low viewer numbers lower the stakes at the moment you are live; they no longer guarantee the moment stays small.

"You get used to it by doing it, so just go live." Exposure does work, and it is the backbone of the clinical results above. But the exposure in those studies is structured, repeated and gradual, starting with situations that are hard but survivable. A first live that lands in a pile-on is not gradual exposure. For some people it is the reason there is never a second one.

Case study: a crowd that exists only for you

Put those findings together and they describe a specific kind of room. It would be private, so no stranger can see the first attempts. It would talk back, so the chat-reading skill gets practiced, not just the camera framing. It would start gently and stay kind, because the research says a hostile audience bites even when it is fake. And it would be there again the next day, because one rehearsal does not change a dominant response.

That is close to what FansFeel sets out to build. It is an Android app, listed as coming soon on Google Play, that gives you a private crowd of AI fans to go live to. Its own pitch names the problem this piece has been circling: "An empty room, judging eyes, trolls: going live can be scary," and "most people never press the button. Not because they have nothing to say, but because of who might be watching." The answer it offers is a room where, as the site puts it, "nobody else is watching."

The design choices map onto the research in ways that are worth spelling out. The site says nobody else can join, watch or see a replay, and that your video and audio are not stored. The fans are a cast of 50 fictional characters, each with a backstory and a typing style, from a hard-to-impress retired gallery owner in Lyon to a one-thumb typist in Porto. They are built to start quietly, which addresses the empty-room moment from a different angle: the site says there is no wall of messages the second you go live, the fans wait for your hello, then the room warms up. They respond to what you say and show, so the chat is about you. They remember things you mention, such as a pet, an exam or an open-mic night, and ask about them next time, which turns a single rehearsal into a run of them with the same regulars. The site says you can see what they know, pause it or wipe it.

There are smaller touches aimed at habit rather than performance. Now and then a fan suggests a small, kind "quest," like drinking water or texting a friend, which the site says are "never dares, never money, never about your body." Hearts and gifts from fans, the site says, are part of the game and have no cash value, and purchases never buy leaderboard rank. The fans chat in 13 languages. The free tier gives up to 10 random fans and 15 live minutes a day; a Premium tier opens all 50 fans with unlimited live time, billed through Google Play, and the site says you see the price before you pay. It is meant for anyone 13 or older, and there is no iPhone version yet.

The honest critique matters as much as the description, and the most important point is one the product makes itself: "The fans are AI. Always." A friendly AI crowd is a dress rehearsal, not the opening night. It cannot reproduce a real hostile chat, and by design it does not try to. The Slater experiments suggest that a kind audience raises a speaker's self-rating, which is the point of a rehearsal, but a rehearsal that only ever feels good could also leave someone underprepared for a rude comment from a real stranger. The memory feature is warm, and it also means the app holds personal details you share, which is why the controls to view, pause and delete them matter. And nothing here is therapy. The clinical studies above involved structured treatment with professionals; an app that offers a kind stage to practice on is a different thing, and the product does not claim otherwise. Someone whose fear stops them doing ordinary daily things should talk to a qualified professional, not a crowd of characters.

The fair test for a tool like this is not whether people enjoy it. It is whether they eventually stop needing it, and carry the practiced skill to a real audience. Since the app has not launched, there is no user data yet to say either way. It is also part of a wider shift toward AI inside creator tools, from AI fans to Instagram's own Effect Builder for AI Stories effects, and each of those deserves the same question: does it build a skill, or replace one?

What we know, and what we don't

Some limits are worth stating plainly. No published study we could find tests an AI-generated text-chat audience as rehearsal for livestreaming specifically. The virtual-audience evidence comes from public speaking, mostly in laboratories or clinics, and often in virtual reality headsets rather than on a phone. The social facilitation literature is old, its effects are modest, and its most famous animal study only partly survived a careful replication. Viewer-count figures for Twitch come from analytics firms rather than the platform, and they vary with method. Instagram's live rule has not been officially described as permanent, and platform rules change often.

What is solid is narrower and still useful. Public speaking is the most common social fear in large survey data. An audience tends to make unpracticed performance worse, and that effect has held up better than its opposite. People respond to virtual audiences as if they matter. Structured, repeated exposure to the feared situation, including virtual versions, reduces public speaking anxiety in controlled trials. And since mid-2025 the main platforms have made first streams both harder to start privately and easier for strangers to find.

A rehearsal ladder for your first live

Here is the practical version, built from the research above rather than from motivation. Each rung is meant to be repeated until it feels boring before moving up.

1. Run the technical rehearsal alone. Framing, light, sound and connection are a separate job from performing, and doing them in advance removes a whole class of live mistakes. YouTube's Practice Mode exists for exactly this, and a phone's front camera recording to your own gallery works anywhere.
2. Script the first 90 seconds, not the whole stream. The opening is where the empty-room feeling is strongest. A rehearsed hello, a one-line reason for being live and a first thing to show give the nervous system a well-worn path at the worst moment. Planning lives around a regular slot, the way our Instagram content calendar guide plans posts, helps the same way.
3. Practice talking into silence. Low viewer counts are normal. Rehearse a few minutes of narrating what you are doing, so a quiet chat stops feeling like a verdict.
4. Rehearse with a responsive room before a public one. The specific skill of reading comments while speaking only develops with comments to read. A couple of trusted friends on a call can do this; so can a private crowd of AI fans, as long as it is treated as a stepping stone rather than a destination.
5. Write your rude-comment script in advance. Decide now what you will do with a hostile message: ignore, delete, block, or one calm line and move on. Set up word filters and, if you can, a trusted moderator before the first public stream. The approach in our piece on a comment strategy that works applies to live chat too.
6. Relabel the nerves out loud. Right before going live, the Brooks findings suggest saying "I am excited" instead of telling yourself to calm down. It sounds trivial. In her experiments it was worth a lot of karaoke points.
7. Count reps, not viewers. For the first month, the number that matters is how many times you went live, not how many people came. Viewer counts are a lagging measure; what to track once they start to matter is covered in our guide to Instagram analytics.

None of these steps asks for courage. Each one turns part of the performance into a practiced response before strangers are watching, which is what the audience research points to as the thing that works in your favor. It also connects to the broader engagement habits that grow an account between lives, and to the slower, saves-driven reach we tracked in our Reels data.

Opening night is not the first night

The trophies on Thursday and the crowds in San Diego this weekend will tell the familiar story of streaming: personality, stamina, nerve. It is a true story about the people who made it. It is a misleading story about how they got there, because it skips the part that happened before anyone was watching.

Every performer the audience research would recognize as good has run the show many times before opening night. What looks like courage on stage is mostly rehearsal that the audience never saw. The platforms have made that harder for beginners just as they have made the stage bigger, and that is worth naming, because it moves the blame. If your thumb has hovered over Go Live and pulled back, it does not mean you lack nerve or have nothing to say. It means you were about to give an opening-night performance of a show you had never run.

So run it. Somewhere private, somewhere kind, as many times as it takes for the hard parts to become the automatic ones. The audience will still raise the stakes when it finally arrives. The difference is that by then, it will be raising the stakes on something you already know how to do.

FAQ

Why do I freeze on live video when I can talk fine with friends?

An audience raises arousal, and arousal tends to amplify whatever response is already dominant. With friends, relaxed conversation is a practiced skill. On a live stream, talking to a camera while reading a scrolling chat is usually a new skill, so the audience tends to bring out its least practiced version. Rehearsal, not willpower, is what changes that.

How many followers do I need to go live on Instagram?

Since August 2025, Instagram has limited Live to public accounts with at least 1,000 followers, and Meta has said the rule applies to private accounts as well. YouTube lists at least 50 subscribers for mobile live streaming. TikTok's LIVE access is widely reported to require being 18 or older with 1,000 followers in most regions. Rules change, so check each platform's help pages before planning around them.

Does practicing in front of an AI or virtual audience actually help?

Lab research shows people react to virtual audiences much as they do to real ones, and clinical trials of virtual reality exposure have reduced public speaking anxiety. No study we found tests AI chat audiences for livestreaming specifically. A private AI crowd is best treated as a dress rehearsal for the chat-reading part of live video, not a substitute for a real audience or for professional help.

Is it normal to feel nervous before every live?

Yes. Public speaking or performing is the most common social fear in large US survey data, reported by about one adult in five. Research by Alison Wood Brooks suggests that relabeling the feeling as excitement works better than trying to calm down, and the spotlight effect means viewers notice your slips far less than you do.