The Token Economy Behind Live Streaming

Who Actually Gets Paid

Somewhere right now a viewer is tapping a small animated rose onto a live stream. It costs about a penny. The performer on screen claps, smiles, maybe reads the sender's name out loud. It looks like affection. It is also a financial transaction that has already passed through several currency conversions and a platform's commission before a single cent reaches the person who earned it. That rose is the smallest visible unit of one of the internet's largest and least examined systems of exchange: the token economy of live streaming.

Not cryptocurrency tokens, though the mix-up is fair enough. These are the older, stranger kind: platform credits that viewers buy with real money and then spend on people, in the transaction the industry politely calls virtual gifting. Chaturbate calls them tokens. TikTok calls them coins on the way in and diamonds on the way out. Twitch calls them Bits. Bigo, Kick, YouTube and a dozen others each mint their own branded chips. The names differ. The machine underneath is the same. You turn your money into a private currency and spend it inside a closed system, and the platform sets the exchange rate at both ends.

The scale is easy to miss, because so much of it is built to be glanced at and forgotten. Analysts put the global live streaming market at roughly $87.5 billion in 2023. In China, where the format matured earliest, virtual gifts alone account for around 85% of streaming revenue, and generated some $6.4 billion as far back as 2017 from a few hundred million users. A single cartoon animation can carry a $500 price tag. Most carry a fraction of a penny. Both kinds are flying up the same screens, all day, everywhere.

The model was not invented in Silicon Valley but in the adult webcam industry, which has run per-minute private shows and tipped public rooms since the mid 2000s. Cam4 has been paying performers since 2007. A 2014 study of camgirls carried the title, with perfect economy, "500 Tokens to Go Private." Everything TikTok and Twitch later wrapped in cartoon animation, cam sites had already worked out: buy credits in bulk, tip to make the next thing happen, pay by the minute, chase a goal bar that fills as the money lands.

The interaction is engineered to feel like a game and a relationship at once. A public room is free to enter, which drops the barrier to zero. Inside, a goal sits at the top of the screen, 500 tokens for the next thing to happen, a bar slowly filling. Performers get coached to open the bar part-filled rather than empty, because a goal at 180 of 500 reads as something about to happen while a goal at zero reads as a dead room. A tip menu prices each act like a diner breakfast. Some rooms wire the tips to a vibrating device the performer is wearing, so a viewer's spending produces a visible physical reaction in real time. Leaderboards rank the top spenders by name, turning generosity into a public contest. Tips trigger sounds and a name spoken within seconds of arrival. Private shows meter by the minute.

The spread

The number that matters most in this business is the one no viewer sees and few performers can explain cleanly. It is the gap between what a token costs to buy and what it is worth to cash out. On Chaturbate, a viewer pays somewhere between 8 and 14 cents per token depending on the bundle. The performer receives 5 cents. Roughly half the money disappears into the platform before anyone has done anything at all. TikTok runs the same trick with an extra layer. Viewers buy coins, spend them on gifts, and each gift converts into diamonds in the creator's account. A diamond is worth about half a cent on withdrawal, 200 of them to the dollar, and the conversion is set so the creator ends up with roughly half of what the viewer paid. On an iPhone it is worse, because Apple has already taken close to 30% on the way in through its app store fee, which can drop the creator's share of the original spend to around a third.

Twitch, the respectable face of the industry, is no gentler. Its standard split on subscriptions is 50/50. A Bit is worth exactly one cent to the streamer, while Twitch keeps a margin of up to 40% on the sale to the viewer. Only the platform's biggest names negotiate 70/30 deals, and even those revert to the standard split past a yearly threshold. Twitch paid out more than $1 billion to streamers in 2022, which reads as generosity until you remember it kept a matching sum for itself. Whatever the branding, the industry has settled on a strikingly consistent answer: the house takes about half. Angela Jones, whose 2020 book "Camming" is the most serious economic study the sector has produced, found cam platforms taking between 20% and 60% of performer earnings. The Chinese market is harsher still, once the talent agencies have taken their cut on top.

Where a viewer's dollar ends up across the major platforms.

The other structural fact is that this money does not come from the crowd. It comes from a handful of people. Industry data suggests the top 1% of gifters in a given room routinely account for 60% to 80% of its gift revenue. The free public tier is not charity. It is an audience against which a small number of heavy spenders perform their spending. The person dropping a $500 "Universe" gift on TikTok, or a Chinese streamer's most devoted patron buying a digital sports car, is the real customer. Platforms know exactly who these people are, and they build ranking perks, badges and private access to keep them spending. Everyone else is there to make the room look alive.

There is a reason platforms reach for tokens rather than simply letting viewers pay performers directly. OnlyFans, the best-known holdout, runs on plain subscriptions and tips in real dollars. It takes a flat 20% and pays out the rest, and it works, but it carries the full weight of being a payment relationship, which is why in 2021 it briefly announced a ban on explicit content under pressure from its banking partners before reversing within days. Tokens put a buffer between the bank and the act. The viewer's card statement shows a purchase of credits from a platform, not a payment to a named sex worker, which keeps Visa and Mastercard comfortable and keeps the money moving. After the American FOSTA and SESTA laws made platforms legally liable for sex work conducted through them, that buffer stopped being a convenience and became a survival mechanism. The token is a compliance instrument as much as a spending one.

Here the token stops being a quirk of branding and becomes the point. Behavioural economists have a phrase, the pain of paying, for the small sting of handing over money. Drazen Prelec and George Loewenstein described in 1998 how that pain gets dulled by separating the moment of payment from the moment of pleasure, and a token does precisely this. You feel the cost once, when you buy a bundle. After that the spending is close to painless, because the leftover credits no longer read as money. They cannot be spent anywhere else. They already left your account. Casinos learned the same thing with chips a century ago, and mobile games rebuilt whole business models on it.

The details are tuned to push in the same direction. Bundles and gifts are priced in odd, non-round numbers, so working out what any gift actually costs in real dollars takes arithmetic nobody does mid-stream. Loss aversion handles the rest: the work economists trace back to Daniel Kahneman and Amos Tversky finds that losing $5 hurts about twice as much as gaining $5 feels good, and a bundle bought once and drained slowly sidesteps the sharpest part of that sting. Balances are sized so they rarely reach zero. That leaves a few stranded credits, which nag to be spent or topped up. A viewer who cannot easily price a gift will spend more on it. That is the design brief, not a side effect.

The people at the bottom of the screen

Strip away the animation and the token economy is a labor market, and a punishing one.

What performers sell is not a rose or a private show but attention that feels personal. Sociologists call it emotional labor, and researchers who study camming call this particular version commercial intimacy: closeness produced on demand and priced by the minute. Jones documented performers sustaining what she termed manufactured identities that stayed warm and present for hours, remembering regulars' names and moods, because the income depends on the feeling of a real relationship even when both sides know the meter is running. A cam performer who greets a returning viewer by name within ten seconds is doing skilled and tiring work. The platform captures the value of it and hands back a fraction.

The work does not clock off. Performers describe streaming six, eight, twelve hours at a stretch. They hold their energy the whole time because the algorithm feeds viewers to whoever is already busy and buries the quiet ones. The rules of that algorithm are never published. Academics who have tried to study cam platforms note that the companies protect their ranking systems as trade secrets, which leaves performers reverse-engineering a machine that can drop them overnight for reasons no one will explain. Building a base of loyal regulars is the only real insurance, and that base is itself unpaid relationship work, done off the clock, to protect income that was never guaranteed in the first place.

The rewards are savagely uneven, and by design. A 2020 survey put the average American webcam model's income near $4,500 a month for 18 hours of work a week, with full-timers earning more and the highest self-reported earners clearing $300,000 a year. Those numbers flatter the middle, because the market is winner-take-all. Tipping concentrates on a few rooms, the algorithm rewards whoever is already winning, and everyone else competes for scraps while doing unpaid relational work simply to stay visible. Nobody here is an employee. There is no minimum wage, no sick pay, and no protection from a platform quietly changing its conversion rate, which platforms do, without notice. The performer carries the risk and the platform carries almost none.

Between the streamer and the platform sits a layer most viewers never think about. In China the business is organised around guilds, and across the wider creator economy around multi-channel networks: agencies that recruit, train, house and manage streamers in return for a slice of the take. One Chinese agency profiled by reporters paid its streamers a $1,500 monthly salary and let them keep roughly 20% of their earnings after the agency and the platform had both been paid. Some arranged bank loans for cosmetic surgery. Some bought gifts through fake accounts to push a streamer up the rankings. Seen from the inside, the token economy looks less like a tip jar and more like piecework with better lighting.

For a long time the whole system escaped the scrutiny its scale deserved, partly because much of it is adult and easy to wave away, and partly because tokens are genuinely hard to trace. That is changing, and the pressure is arriving through the courts. In 2024 the state of Utah sued TikTok, and the unsealed filings are ugly. They describe an internal TikTok study, "Project Meramec," which found that 112,000 users aged 13 to 15 managed to access TikTok LIVE in a single month of 2022, despite an age floor of 18. The same filings quote the company's own researchers concluding that the feature enabled the "exploitation of live hosts" and that TikTok profited from "transactional gifting" involving nudity and sexual activity. Utah's attorney general called it a virtual strip club that TikTok kept running because it made too much money to shut down. A separate internal review, "Project Jupiter," had already flagged the gifting feature as a money laundering channel. Dirty money buys the gifts, and clean creator income cashes them out.

The harm is not confined to the adult corners. In 2025 The Observer documented livestreams in which vulnerable children in Pakistan, Syria, Afghanistan and elsewhere were placed in front of cameras to beg for virtual gifts for hours at a stretch, with organised handlers pocketing the proceeds while TikTok's algorithm promoted the content and the platform took a commission reported as high as 70%. Olivier de Schutter, the United Nations special rapporteur on extreme poverty, called it profiting from people's misery. The Financial Action Task Force has warned separately that gifting and micro-donation systems are being used to move money across borders with little traceability. A cartoon rose, it turns out, is an almost ideal instrument for both washing money and hiding who is really paying whom.

Regulators have barely started to catch up, and where they have, the responses show how differently the world reads this. China treats virtual gifting as a genuine financial risk. It caps how much a viewer can spend on gifts in a single day and requires real-name verification for tippers, blunt instruments aimed squarely at spending harm and money flows. The caps followed widely reported cases of viewers giving away their savings to chase a streamer's attention. The European Union mostly treats tips as digital-service transactions and charges VAT on them. The United States has no federal rule specific to tipping at all, which is why the Utah case had to reach for consumer-protection and child-safety law to get any grip on the thing. My own read is that the token is doing the job it was built for, and that this is the problem, not an abuse of it. The abstraction exists to make spending feel like generosity, to keep the platform's share out of sight, and to hold a workforce of millions in a category that is not quite "worker." A credit that costs a penny to send, returns a third of a penny to the person who earned it, and can obscure the identity of both parties is not a glitch.

None of this makes the affection fake. The lonely viewer tipping his favourite streamer at two in the morning is feeling something, and the performer thanking him by name may be too. What the token did was find a way to meter that feeling, take half of it at the door, and file the rest under "gifts," so that tax authorities, money-laundering investigators and labor regulators all end up looking at the same transaction and seeing three different things, or nothing at all. That ambiguity is not a defect the industry is racing to fix. It is the most valuable thing the token does.