AI companion ARPU and LTV benchmarks for 2026: CPM, CPA and payback

AI companion ARPU and LTV benchmarks for 2026 start with three public numbers: companion apps earned about $1.18 per download in 2025 (Appfigures), subscription apps realize a median $23 per payer in their first year (RevenueCat), and adult network display CPMs run roughly $2 to $10 in tier-1 countries. Use them as floors, then replace them with your cohort data.

Every external number below links to where it came from, every worked example labels its assumptions, and the last section shows how to build a benchmark sheet from your own traffic, the only benchmark a media buyer should scale against.

What do ARPU, ARPPU and LTV mean for an AI companion site?

Half the confusion in companion economics comes from mixing denominators. Before comparing any two numbers, check what they divide by.

• ARPU (average revenue per user): revenue divided by all users, paying or not. App-store data usually reports this per download, which makes it look tiny.
• ARPPU (average revenue per paying user): revenue divided by paying users only. This is the number that matters when you price traffic, because you buy payers, not visitors.
• LTV (lifetime value): total revenue a payer generates before they churn. Realized LTV counts what was actually collected; projected LTV extrapolates from a retention curve.
• CPM, CPC and CPA: what you pay per thousand impressions, per click, and per paying customer. CPM is what networks quote; CPA is what decides whether you make money.
• CAC payback: how many months of a payer's revenue (your share of it) it takes to earn back what you spent to acquire them.

A $1.18 revenue-per-download figure and a $30 monthly ARPPU can describe the same business. You just cannot set a CPA bid from the first one.

AI companion ARPU benchmarks: what the public data says

The best public data on companion spending comes from app intelligence firm Appfigures. In its August 2025 analysis reported by TechCrunch, Appfigures counted 337 active, revenue-generating companion apps that had been downloaded 220 million times and had driven $221 million in lifetime consumer spending. The category earned $82 million in the first half of 2025, and revenue per download rose from $0.52 in 2024 to $1.18 in 2025. Spending was extremely concentrated: the top 10% of apps took 89% of revenue.

A year later, Appfigures split the market by content. According to Decrypt, 214 romantic and NSFW companion apps generated $162.8 million in the first half of 2026, almost exactly matching the $164.8 million earned by general, PG companion apps, even though the PG apps had nearly twice as many downloads. Put simply, romantic companion users spent roughly twice as much per download. Those figures are gross, before the app stores take their fee (typically 30%), and they exclude web-only platforms such as Candy AI and SpicyChat.

The creator economy gives a second reference point. Euronews reported OnlyFans processed $7.2 billion in fan payments in 2024 across 377,456,000 fan accounts, with the platform keeping a 20% fee and booking $1.4 billion in net revenue. Divide gross payments by fan accounts and you get about $19 per account per year. Most of those accounts are inactive, so this is an ARPU floor, not what a paying fan spends.

What none of these sources publish is ARPPU for a web-based adult companion site, because the biggest ones are private and web-only. That gap is why operators should treat app-store figures as a lower bound. A web site keeps the store fee, can sell tokens and content unlocks without platform review, and can show adult content the stores would reject. For planning, the earnings calculator on our AI companion site earnings page uses $30.23 in average monthly revenue per paying member. Treat any single number like that as an input to test, not a promise.

AI companion LTV benchmarks: how much is a payer worth?

The largest public dataset on subscription lifetime value is RevenueCat's State of Subscription Apps 2026, built from more than 115,000 apps and over $16 billion in revenue. It is not companion-specific, but it is the cleanest base rate available.

Two findings from the same report matter directly for companion operators. First, AI-powered apps generate 41% more revenue per payer but churn 30% faster than other apps. That is the companion business in one sentence: users spend heavily while the novelty lasts, then leave unless something keeps them engaged. Second, look at the high price point row. Month 1 LTV of $35.89 rising to only $62.19 by the end of the year means the typical high-priced payer delivers more than half of their first-year value in the first 30 days.

The practical takeaway: your LTV is decided early. If you cannot see a retention curve flattening by month three, longer projections are guesses. Our AI fan site retention playbook covers the levers that move that curve.

How to calculate LTV for your own site

The simple formula is monthly ARPPU divided by monthly churn. A site with $30 ARPPU and 25% monthly churn has a projected LTV of $120 ($30 / 0.25). The formula breaks down when churn is front-loaded, which it usually is: many payers cancel after month one, then the survivors stick for a long time. A cohort table is more honest. Track each month's acquired payers, record how many are still paying in month two, three and six, and add up the revenue they actually produced.

Conversion benchmarks: how many visitors become payers?

Your CPA is set by your conversion rate far more than by your CPM. RevenueCat's 2026 data shows how much paywall design changes that rate:

• Hard paywall (pay before using the product): median 10.7% of downloads become paying by day 35, with the top quarter above 20%.
• Freemium (free tier with upgrades): median 2.1% by day 35, top quarter above 4.5%.
• Revenue per install by day 60: $3.09 median for hard paywalls versus $0.38 for freemium.
• Trial length: trials of 17 to 32 days convert at a 42.5% median, 5 to 9 days at 37.4%, and 4 days or less at 25.5%.
• Day-zero cancellations: 55.4% of 3-day trial cancellations happen on day zero, versus 31.1% for 30-day trials.

Companion sites rarely use a pure hard paywall, because a free first conversation is the product demo. The pattern that works is closer to a metered model: a few free messages, then a paywall at the moment the user is most invested. If your landing page to paid rate sits near the freemium median, your traffic or your first conversation is the problem, not your price.

One more RevenueCat figure is worth noting: web purchases are only 3.2% of subscription app revenue globally. Companion operators run the opposite model. Most adult companion revenue happens on the web, which is why app-store benchmarks undercount it.

CPM and CPA benchmarks for AI companion traffic

Mainstream platforms such as Meta and TikTok do not allow ads for adult content, so most adult companion offers are bought on adult ad networks. The networks do not publish rate cards, and prices move with geo, device, format and site tier. The most detailed public compilation is from affiliate publication AFFNinja, last updated December 2025. Treat it as a starting range and confirm with each network's bid estimator before you budget.

CPM only tells you the price of attention. To turn it into a CPA, use this formula: CPA = CPM / (1,000 x click-through rate x visitor-to-payer rate). The table below runs it for a $5 CPM, a typical mid-range tier-1 display price, using illustrative click and conversion rates rather than measured ones.

Doubling your landing page conversion rate cuts CPA in half at the same CPM. That is why experienced buyers obsess over the first chat screen and the paywall moment rather than shaving a dollar off their bids. Our paid traffic playbook covers creatives and landing pages in detail.

What is a good CAC payback for an AI companion site?

Payback combines everything above: CPA, ARPPU, retention and the share of revenue you keep. Here is a worked example with every assumption stated. It is a model, not a reported result.

• Monthly revenue per active payer: $30 (subscription plus tokens and unlocks).
• Share of active payers remaining: 100% in month 1, 55% in month 2, 42% in month 3, 35% in month 4, 30% in month 5, 27% in month 6.
• Partner economics on a white-label platform: a revenue share of up to 60% of total site revenue, so up to $18 per active payer per month.

Read the last column against your CPA. At a $25 CPA you are paid back in month 2. At $35 it takes until month 3. At $50 you need five months, and if month-two retention slips from 55% to 40% you may never get there inside six months. A common rule of thumb is to recover CAC within about three months on paid traffic and to want lifetime value of at least three times CAC. Neither is a law, but both force you to look at retention before you scale.

If you build your own stack instead of using a revenue share, you keep 100% of gross, but you also pay for model inference, image generation, hosting, moderation and high-risk card processing, which carries higher fees and rolling reserves than standard processing. Our guide to high-risk payment processing for AI companion sites explains those costs. Get real quotes before you compare the two models.

How to build your own AI companion benchmark sheet

Public benchmarks get you to a first bid. After about two weeks of traffic, your own numbers should replace them. Track these per traffic source, per landing page and per companion niche:

1. Cost per click and click-through rate, straight from the network.
2. Visitor to first message: the share of landing page visitors who start a conversation.
3. First message to payer: the share of chatters who pay within 24 hours and within 7 days.
4. First-week ARPPU: subscription plus tokens and unlocks bought in days 0 to 7.
5. Month-two retention: the share of a month's payers still paying 30 days later. This single number predicts most of your LTV.
6. Refunds and chargebacks as a percentage of transactions. Processors watch this closely, and a high rate can cost you the account.
7. Cumulative revenue per acquired payer at day 30, 60 and 90, compared with the CPA you paid for that cohort.

Cut each metric by geo too. RevenueCat's regional gap ($32 year-one LTV per payer in North America versus $14 in India and Southeast Asia) shows how much a geo mix can move a blended number. A campaign that looks unprofitable in aggregate is often one profitable tier-1 geo dragged down by cheap tier-3 clicks.

Log the date, source, CPM, CPA, first-week ARPPU and month-two retention for every test. Within a few months you will have benchmarks for your own traffic, niches and funnel. If you would rather spend that time on traffic than on infrastructure, WhiteLabelFans runs the site, the AI companions and billing under your brand. See how the white-label AI companion platform works.

Frequently asked questions

What is a good ARPU for an AI companion site?

It depends on the denominator. Across all downloads, companion apps averaged $1.18 in 2025 according to Appfigures. Per paying user, subscription apps realize a median $23 in their first year and $32 in North America, according to RevenueCat. Web-based adult companion sites that sell tokens and unlocks on top of a subscription should aim above those floors, and confirm it with their own cohorts.

How do you calculate LTV for an AI companion site?

The quick version is monthly ARPPU divided by monthly churn, so $30 with 25% churn gives $120. Because companion churn is front-loaded, a cohort table is more reliable. Group payers by the month they joined, track how many still pay each month after, and add up the revenue each cohort actually produced. Use realized revenue, not projections, when you raise bids.

What CPM should I expect on adult ad networks?

Third-party compilations put typical CPMs at roughly $2 to $8 on ExoClick, $3 to $10 on TrafficJunky and $1.50 to $6 on TrafficStars, with US desktop and video inventory at the top of those ranges and tier-3 geos far below. Networks do not publish official rate cards, so check each network's bid estimator for your geo, device and format before budgeting.

What is a good LTV to CAC ratio for AI companion traffic?

Many subscription businesses use three to one as a rough target: a payer should be worth at least three times what it cost to acquire them. For paid traffic, payback speed matters as much as the ratio. If your share of a cohort's revenue covers its CPA within about three months, you can recycle cash into more traffic instead of waiting on long-tail renewals.

Why do app store benchmarks understate web companion sites?

App-store data counts every download, reports revenue before the stores take their fee, and excludes web-only platforms such as Candy AI and SpicyChat. App stores also restrict explicit content, which caps what romantic apps can sell. A web site keeps the store fee, can sell tokens and unlocks freely, and serves the adult audience directly, so its per-payer numbers tend to run higher.

Sources

• TechCrunch: AI companion apps on track to pull in $120M in 2025 (Appfigures data)
• Decrypt: This is how much your boyfriend is spending on AI girlfriends (Appfigures H1 2026 data)
• RevenueCat: State of Subscription Apps 2026
• Euronews: OnlyFans takes $7.2bn from subscribers in 2024
• AFFNinja: Top adult ad networks and platforms 2026