AI Girlfriend Apps Are Taking Fan Spend

TL;DR. AI girlfriend apps now take real money from the same wallet as OnlyFans, but only from one slice of it. App intelligence firm Appfigures counted 214 NSFW and romantic AI companion apps taking 162.8 million dollars in the first half of 2026, with general companion apps adding 164.8 million, reported by Decrypt on 13 July 2026. The whole companion category was tracking to roughly 120 million for all of 2025, so mobile companion spend has multiplied several times over in twelve months. It is still small against OnlyFans, whose parent Fenix booked 7.22 billion dollars of gross fan spend in FY2024, about 602 million a month, roughly 22 times the NSFW companion category's monthly run rate. The threat is not that a chatbot outbids your whale. It is that the cheapest, chattiest, lowest-margin fans in your funnel now have an unlimited alternative inside your own price band, and a new reference point for what conversation is worth. Defend on realness, scarcity and live surfaces, not message volume.

The damage lands on margin well before it lands on revenue.

How Big the AI Companion Category Got in Eighteen Months

Two Appfigures data points bracket it.

August 2025. TechCrunch reported 337 active, revenue-generating AI companion apps worldwide, 128 released that year, taking 82 million dollars in the first half of 2025 and on track for more than 120 million across the full year. Cumulative spend stood at 221 million as of July 2025, off 220 million downloads, with the top 10 percent of apps taking 89 percent of the money.

July 2026. Decrypt, citing Appfigures Intelligence, reported 214 NSFW and romantic companion apps taking 162.8 million dollars in the first half of 2026, with general companion apps taking a further 164.8 million. Cumulative spend for the tracked set reached 427.3 million across 165.3 million downloads since late 2022.

Total category spend therefore went from 82 million in the first half of 2025 to roughly 327 million in the first half of 2026. Call it four times and hold it loosely, because the two Appfigures cuts use different taxonomies: the 2025 read counted 337 apps under one label, the 2026 read splits a 214-app romantic set out from everything else. The direction is not in doubt, but the multiple compares two definitions, so treat it as directional rather than as a growth rate you can extrapolate.

Concentration is extreme, and it changes who your actual competitor is. In the first half of 2026, Zeta took 33 million dollars, Tipsy Chat 15.2 million, ChatBox 13 million, Crushie AI 8.8 million and Emochi 7.5 million: 77.5 million across five apps out of 214, close to half the category. Zeta is a Korean character roleplay app from Scatter Lab, launched April 2024, grown fastest in Japan and Korea and only now testing the United States. Its product is co-writing fiction with characters, not simulating a specific woman who is into you. Much of the money labelled NSFW and romantic there competes with fan fiction and mobile games, not with a subscription to a real creator.

Decrypt notes one more limitation: the figures cover mobile app stores only and exclude web-only platforms such as Joi AI, Candy AI and SpicyChat AI. Both stores restrict adult content, so web is where the genuinely explicit products live, and the mobile figure undercounts precisely the segment that overlaps with your business. Treat 162.8 million as a visible floor.

Who Is Actually Paying, and What They Buy

Category revenue tells you little. Per-user behaviour tells you more, and there is one good survey.

On 19 May 2026 the Wheatley Institute at Brigham Young University and the Institute for Family Studies published "Secret Soulmates," a survey of 2,431 United States adults aged 18 to 30 in committed relationships. One in seven, 15 percent, said they regularly interact with an AI chatbot simulating a romantic partner, and 20 to 30 percent had experimented at some point. Thirteen percent regularly roleplay romantically or sexually with one, and 11 percent often use one to generate sexually explicit content. The Institute for Family Studies separately cites Gallup work for the Walton Family Foundation putting monthly use among 18 to 28 year olds at 10 percent.

Then the finding that matters. Among partnered regular users, roughly 30 percent said their partner had no knowledge of the use at all, and 69 percent said keeping it from their partner was somewhat or extremely important.

That is the same psychological slot a paid fan subscription occupies. Discreet, recurring, emotionally loaded, billed to a card the buyer would rather nobody examined. It is a competitor for a budget line the fan already keeps hidden, which makes it a closer substitute than the raw spend number implies.

Now the money per head. Cumulative spend of 427.3 million dollars across 165.3 million downloads is about 2.59 dollars of lifetime revenue per install. Freemium conversion is low, commonly put under 5 percent, a practitioner figure rather than one Appfigures published. On that basis lifetime revenue per paying user lands in the low tens of dollars. Review sites that test these apps put advertised entry tiers roughly in the 5 to 17 dollar band, with token models for images, voice and video pushing effective spend higher.

Fenix's FY2024 accounts show 7.22 billion dollars of gross fan spend across 377.5 million registered fan accounts, about 19 dollars per account per year. Most are dormant, so the real distribution is a small paying core carrying an enormous tail, the pattern in our breakdown of who OnlyFans fans are and how they spend. Both businesses live off a thin layer of heavy users.

AI Girlfriend Apps vs OnlyFans: Where the Overlap Is Real

Run the scale arithmetic first. OnlyFans took 7.22 billion dollars of gross fan spend in the twelve months to 30 November 2024, roughly 602 million a month. The tracked NSFW and romantic companion category took 162.8 million in six months, roughly 27 million a month. One platform's monthly gross is about 22 times the whole visible mobile companion category. Annualise the H1 2026 NSFW figure and you get about 326 million, around 4.5 percent of what fans put through OnlyFans alone in FY2024.

So anyone telling you AI girlfriends are eating the creator economy is selling something; at visible scale they are not. They are growing several times faster than the platform, from a base small enough that growth is cheap, into a wallet that is not growing at all.

Three tests decide whether a given fan is exposed.

Is he buying volume of attention, or access to a specific person? A fan who wants someone to talk to for two hours at midnight is buying volume, and an app serves that better than any chat team, forever. A fan who wants this creator, whose face he recognises and whose schedule he knows, is buying access. No model substitutes for that.

Does the product survive being told it is not real? For a roleplay habit, nothing happens. For a girlfriend experience relationship built over eight months, the value goes to zero. California now forces that test into the opening line of the conversation, which we come back to below.

Is he buying reciprocity or performance? Reciprocity, the belief that a real person is reacting to him specifically, is the one thing a language model cannot supply honestly. Performance, meaning erotic text on demand, is now a commodity at 13 dollars a month.

Sort your paying base against those three and the exposed segment is usually smaller and lower value than owners assume: low-tier subscribers who consume enormous chat volume and buy almost nothing.

Why a Thirteen Dollar Companion Resets the Price Anchor

The subscription price is not where the damage happens. OnlyFans subscriptions run from a 4.99 dollar floor to a 49.99 dollar ceiling. Companion apps advertise roughly 5 to 17 dollars. The AI product sits inside your price band, not underneath it, so it does not undercut the subscription at all.

What it undercuts is everything sold on top. Consider what a fan gets for 13 dollars each way. From a companion app: instant replies, unlimited volume, no queue, perfect recall, no refusals, availability at four in the morning on a Tuesday, images or voice for a token top-up. From a managed creator at 12.99: a content feed, plus a thread answered by a chat team working a queue across many fans, latency in minutes to hours, and the substantive content behind unlocks priced at 15, 25 or 50 dollars.

No fan builds that comparison consciously. Anchoring does not require it, only repeated exposure to a different price for something adjacent, which quietly changes what feels expensive. The fan who spent three months getting unlimited responses for 13 dollars opens a 25 dollar unlock for a two-minute clip and, for the first time, hesitates. Nothing about your product got worse. His reference point moved.

At the top of the ladder the anchor is irrelevant. Whales are not buying message volume. They buy scarcity, status inside a small group, and the certainty that a specific real person noticed them. An AI companion has infinite supply and zero scarcity, which makes it structurally incapable of delivering what a whale pays for. Our guide to building and keeping whale accounts applies harder now that the commodity tier has a commodity competitor.

At the bottom of the ladder the anchor is decisive. The 10 to 60 dollar a month fan who mostly wanted conversation is the exposed cohort, and on most rosters he consumes the highest share of chat minutes per dollar of revenue. If offshore chat labour costs you somewhere in the practitioner range of 3 to 8 dollars an hour fully loaded, and a meaningful share of those minutes goes to fans who never buy, losing part of that cohort is a margin event rather than a revenue event, provided you resize the chat team instead of wasting the recovered hours. Measure exposure in chat minutes, not only subscription dollars, and the picture flips from threat to reallocation.

What the Substitution Does to Girlfriend-Experience Revenue

Girlfriend experience is the exposed product line, because its promise, she talks to you like she is yours, is word for word the companion app pitch. Anyone selling it as message volume is selling a commodity against a competitor with zero marginal cost.

Model it rather than guessing. Take a creator grossing 10,000 dollars a month on a split of roughly 30 percent subscription and 70 percent messaging, tips and unlocks. That split is a practitioner working assumption, not a published platform figure, so substitute your own dashboard numbers. Now assume 10 percent of her active subscribers migrate to a companion app over twelve months. The naive read is a 10 percent revenue hit. Three things make the real read different.

Churn is not random by spend band. Migrating fans skew heavily toward the non-buying tail, because those are the fans whose entire consumption is conversation. Losing 10 percent of subscribers may cost 3 or 4 percent of revenue.

Chat minutes freed exceed revenue lost. Those same fans consumed a disproportionate share of the queue. Recover 10 or 15 percent of chat hours, redirect them to the top spend decile, and net contribution can rise while gross revenue dips.

The unlock economy is where the real risk sits. The dangerous scenario is not churn, it is a quiet decline in unlock conversion across the whole base as the reference price shifts. That shows up as flat subscriber counts with falling revenue per fan, the number to instrument this quarter.

The defensible core of girlfriend experience is continuity with reality: today's date, an event that happened this week, her actual voice, a schedule she keeps. None of it is available to a model. Shift the product from unbounded message volume to fewer dated, verifiable, unmistakably human moments, which is where our guide to monetising the girlfriend experience was already pointing.

Defensive Moves: Scarcity, Realness Proof and Live Surfaces

Six moves, in the order they pay back.

1. Stop competing on latency. You cannot win a response-speed race against a model, and trying to has been quietly setting your chat team's headcount for years. Replace the implicit promise of instant replies with an explicit, reliable window: she reads messages between set hours, and she always does. Reliability is a human signal. Instantaneity now reads as a machine one.

2. Build realness proof into the paid product. Not claims, artefacts. A voice note naming the day and something that actually happened in it. A clip filmed with a visible current date. A first-name mention agreed in advance. Content shot the same day it sells. The test: could a model have produced this a week early? If yes, it is not proof.

3. Introduce hard scarcity and publish it. Cap girlfriend-experience slots per creator per month at a number you can service, eight or twelve, and say so. Run a waitlist. Scarcity is the one attribute infinite compute cannot copy, and it is what whales pay the largest premiums for.

4. Move margin onto live surfaces. Scheduled video calls, streams with tip goals, real-time requests on camera. Live is where a fan asks for something specific right now and watches it happen, the cleanest demonstration of a real person that currently exists.

5. Use the disclosure asymmetry. California's SB 243, signed 13 October 2025 and effective 1 January 2026, requires companion chatbot operators to give a clear and conspicuous notification that the chatbot is artificially generated and not human wherever a reasonable person could be misled into thinking they are talking to a person, backed by a private right of action for the greater of actual damages or 1,000 dollars per violation plus injunctive relief and attorney's fees. Regulation now makes your competitor announce that it is not real, while OnlyFans requires identity verification for every creator account. The law is doing your differentiation for free.

6. Do not answer AI with undisclosed AI. The temptation to close the volume gap with a bot is obvious and the risk is asymmetric: platform disclosure rules are tightening, state chatbot statutes are proliferating, and the value of a verified-human product dies the day a fan learns he was talking to a script. That make-or-buy call is worked through in our comparison of AI versus human OnlyFans chatters. Wherever you land, document the decision and keep supervision current.

Roster and Revenue-Mix Decisions for the Next Twelve Months

Four decisions, all of which you can start on Monday.

One: re-weight the roster toward creators who can go live. A creator whose entire product is text and stills competes with a commodity. One who can hold a camera, take a call, record a voice note and improvise does not. Add it to your intake criteria for every signing from this quarter, and be honest about which existing creators cannot clear the bar.

Two: reprice deliberately rather than defensively. If unlock conversion is drifting down, cutting prices into a competitor with zero marginal cost is the wrong move, because that race ends at free. The right response is usually fewer, higher-value, more clearly human offers, and a subscription price reflecting what only a real person supplies. If you are raising prices on an existing base, sequence it using our walkthrough of raising the subscription price on existing subscribers.

Three: treat AI-native supply as a separate business line, not a hedge. Sacra puts Fanvue at roughly 200 million dollars of annual recurring revenue as of May 2026, up from about 100 million at the end of 2025, with AI creators accounting for something like 15 percent of platform revenue. That is a real business, and a different one, with its own pipelines, disclosure duties and platform risk. It does nothing to protect your human roster. Our assessment of AI creators on Fanvue sets out the case both ways.

Four: instrument four metrics per creator, monthly. Revenue per active fan. Share of revenue from the top 1 percent. Share of chat minutes spent on fans who bought nothing in ninety days. Churn split by spend band. Without those you cannot tell AI substitution from seasonal softness or a chat team that got worse.

A note on pace, because the ground moves both ways. In October 2025 Sam Altman said OpenAI would allow erotica for age-verified adults from December 2025. It never shipped, and by late March 2026 it was reported as paused indefinitely. In July 2026 San Francisco City Attorney David Chiu sent Apple and Google cease-and-desist letters over 13 nudify apps: Apple said it had removed three and was contacting four more, and Google suspended all five Play apps named. And when xAI launched its Ani companion in July 2025, TechCrunch reported daily iOS downloads up 40 percent the next day to 171,000 while revenue rose only 9 percent, to 337,000 dollars. Companion features drive installs far more reliably than they drive money, distribution for explicit products stays fragile, and the largest potential entrant shelved its adult plan rather than ship it. Plan for a competitor that grows steadily, not one that arrives at once.

This is commercial analysis, not legal, tax or investment advice. Chatbot disclosure statutes, platform AI policies and age verification duties differ by state and change quickly, so take specific questions to a professional. To pressure-test your revenue mix, we are on Telegram at t.me/whalefindersupport.

Frequently Asked Questions About AI Companions and Fan Spend

How much do people actually spend on AI girlfriend apps?

Appfigures counted 162.8 million dollars across 214 NSFW and romantic AI companion apps in the first half of 2026, with general companion apps adding 164.8 million, as reported by Decrypt on 13 July 2026. Cumulative spend is 427.3 million across 165.3 million downloads since late 2022, roughly 2.59 dollars of lifetime revenue per install. Those figures cover mobile app stores only and exclude web-only platforms, so the real category is larger.

Are AI companions hurting OnlyFans creator earnings?

No published dataset shows a measured decline caused by AI companions, so anyone quoting a precise figure is guessing. What the numbers support is a scale comparison: OnlyFans took roughly 602 million dollars of gross fan spend a month in FY2024, against about 27 million for the whole visible NSFW companion category. Treat it as developing pressure on one segment, not a market shift.

Which fans are most likely to switch to an AI companion?

The low-tier subscriber who consumes large amounts of conversation and buys little else. He pays for volume of attention rather than access to a specific person, and an app serves that better and cheaper. Whales are least exposed, because they buy scarcity, status and reciprocity from a real person. That exposed cohort also consumes the most chat minutes per dollar of revenue, so losing part of it is a margin question first.

Should our agency launch its own AI companion or AI chatter?

Separate the two. An AI-native creator on a platform that permits and labels them is a distinct business line with its own pipeline and disclosure duties, and it does not hedge a human roster. Undisclosed AI answering messages on a verified-human account is different: the value collapses the day a fan finds out. Decide deliberately and get the compliance position checked.

Does the law require AI companions to disclose that they are not human?

In California, yes, from 1 January 2026. SB 243, signed 13 October 2025, requires companion chatbot operators to give a clear and conspicuous notification that the chatbot is artificially generated and not human wherever a reasonable person could be misled into believing they are talking to a person, backed by a private right of action for the greater of actual damages or 1,000 dollars per violation, plus injunctive relief and attorney's fees. Rules differ by state and are moving quickly, so check the position where your fans actually are.

Is this investment advice, and how does WhaleFinders fit in?

No. This is educational commercial analysis, built on published app intelligence, a university survey and company accounts, none of which describe your roster. WhaleFinders works white-label as the marketing direction arm inside OnlyFans agencies on flat monthly pricing, 349 dollars single channel, 529 dual, 679 triple and 799 omni per creator per month, and never chats with fans or handles creator money.