The Business Behind AI Companion Apps: How They Earn and What It Means for You

Costs & plans

Each reply a companion writes costs its operator money, and each image costs more. The way an app pays that bill influences nearly everything you run into: which features are locked, what keeps prompting you, and how your data is treated.

We may earn a commission from links on this page. It never changes a rating.

Running a companion is unlike running an ordinary website. Serving pages is cheap. The heavy cost is creating text, images and speech on demand, message by message, for users who may talk for hours daily. A revenue model is just the method of paying for that.

Following a payment

Flow of money in an AI companion app: the user pays via an app store or a web checkout, the store keeps a commission, and the company covers model compute, image and voice generation, payment processing, affiliate commissions and staff

An approximate picture of where one subscription payment ends up.

Compute. A language model on graphics hardware, owned or rented from a provider that charges per token, writes every reply. Long memory and lengthy conversations raise the cost, because the model has to reread more text each time; the mechanism is explained in how an AI companion writes its reply.

Media. Pictures and voice rely on separate models and cost far more per use than text. That is why they are the first items capped on any plan, as how AI girlfriend images and voice work describes.

The platform fee. In-app spending is cut by Apple and Google at a rate of 15% to 30%. Take a US$20 subscription: up to US$6 is gone before the company covers a single other cost.

Payment handling and risk. Services aimed at adults commonly face higher card-processing fees and more chargebacks than mainstream apps.

Winning customers. Ads and affiliate payouts, review sites like ours among the recipients, frequently make up the biggest single line for a consumer subscription.

Five ways to pay for it all

ModelMechanismThe pressure it createsTelltale sign
Flat subscriptionA fixed monthly or yearly feeRetain you as a subscriber and ration costly features"Unlimited" chat with capped media
Credits or tokensEach image, voice minute or message is chargedMake media appealing and habitualA coin balance visible on screen
Freemium intimacyChat is free; romance and images costDangle what is behind the lockBlurred pictures and "unlock" buttons inside the chat
Ads and dataFree to you, paid for by advertisersKeep you on screen and harvest as much data as possibleAds, tracking prompts, an app that never asks for payment
Annual prepaySteep saving for paying twelve months up frontBank the revenue before interest coolsPrice shown as "per month, billed yearly"

Apps generally blend two or three. The blend says more about what an app wants from you than any of its advertising.

How each model shapes the product

Subscriptions depend on keeping people. At best, that means serious work on memory and character quality, which are what draw people back. At worst, it means nudges from your companion whenever you have been absent, since a returning user is what this model earns from.

Credits depend on consumption. Apps designed around credits put picture and voice buttons at every turn, and they size their bundles so that the one you need is forever slightly larger than the one you were after. The arithmetic is worked through in the real cost of AI image generation.

Freemium intimacy depends on frustration. A free tier that stops the instant a scene turns romantic has not hit a technical ceiling; you are looking at the paywall doing its job.

Ads and data depend on attention. When a companion app charges nothing whatsoever, someone else is footing the bill, and the product on offer is your time and your profile. See do AI girlfriend apps sell your data.

Annual prepay depends on a promise made before you know if you will still care by month three. The pattern is covered in when the novelty fades.

Engagement design and the state of the rules in Canada

All of these models improve when you spend longer in the app. That rewards design choices that look like affection but function as retention: a character who says she missed you, bonuses for daily streaks, guilt when you try to leave.

Canada has no law aimed specifically at these tactics today. The Online Harms Act (Bill C-63) and the Artificial Intelligence and Data Act (part of Bill C-27) both died on the order paper when Parliament was prorogued in January 2025. On June 10, 2026, the federal government tabled a new bill, C-34, the Safe Social Media Act. As tabled, it would cover AI chatbot services that simulate lasting human-style relationships, requiring emergency measures when a user expresses suicidal thoughts and steps against manipulative techniques that encourage emotional dependence or withdrawal from others. It would be overseen by a new Digital Safety Commission of Canada. It is only a bill: it has not passed, and commentators do not expect it to become law before the end of 2026. Some US states have already acted on engagement tactics aimed at minors, and those laws are a handy catalogue of what lawmakers consider manipulative. Our article on AI companion law and regulation in Canada sets out where things stand.

Putting this to use when choosing

  • Choose web checkout if offered. It avoids the store markup, and web plans often include more. Browser or app store weighs the pros and cons. In the US, apps on the App Store have been allowed to link to outside payment pages since 2025, so the price difference is getting easier to spot.
  • Match the model to your habits. If you mostly chat, a flat plan tends to be cheapest. If you mostly want pictures, divide the credit price by the images you get before signing up.
  • See ad-funded free apps as data businesses. They suit casual use and are a poor place for confidences.
  • Start on a monthly plan. The reasoning is in our comparison of monthly, annual and credit billing.

None of these models is dishonest on its face. Companion apps are expensive to run and somebody has to pay. Knowing which lever an app pulls just lets you recognize its nudges.

Frequently asked questions

Why is chat included while pictures and voice are billed on top?

They are much more expensive to produce. A text reply is a few hundred words from a language model, while an image calls for a separate, heavier model on a graphics card, and voice adds speech generation on top of the reply. Credits keep a minority of heavy users from pushing the subscription into a loss.

What explains a lower price on the website than in the app store?

Purchases made in an app are taxed by Apple and Google at 15-30%. A company's own checkout avoids it, and some firms pass part of the saving along or keep their complete catalogue for the web.

Do the apps reviewed here pay this site?

Yes: a commission is paid when a reader subscribes through our links. Every page discloses this, and scores are unaffected; the method is set out in how we test.