The $637 Lesson: What 15 Platforms Reveal About Hidden Costs

Eleven months. Fifteen platforms. Six hundred thirty-seven dollars spent testing every pricing model the AI companion market could throw at me. What emerged wasn't a collection of quirks — it was three systemic traps that repeat across the category, documented in what remains the single most detailed pricing audit available.

The first trap wears a discount label. Replika charges $19.99 monthly but $69.99 annually — a 71% spread that makes the monthly plan "a terrible deal" and the annual plan "excellent" by design. Candy AI pushes a 54% annual discount. Chai offers only 20%. These aren't loyalty rewards; they're lock-in mechanisms dressed as savings, and the reviewer found annual plans "wildly underused" by consumers who correctly sense the pressure.

• Annual decoy spreads: Replika's 71% monthly-vs-annual gap, Candy AI's 54% discount framed as urgency
• Token math surprises: DreamGF's $9.99 base becoming $24 with $14 in mandatory top-ups during month one
• Silent auto-renewals: Three platforms charged $37 total after free trials without a single warning email

The second trap hides in credit systems. On DreamGF and CrushOn.ai, the reviewer spent an extra $14–20 monthly on top-ups — turning a "$10 experience" into $24 because "they make it impossible to predict your actual monthly cost." Premium add-ons stack further: Muah AI and DreamGF sell extra character packs and image credits on top of subscriptions, "like paying for cable and still getting pay-per-view charges."

The third trap is administrative. Currency conversion fees from overseas platforms added $2–4 per transaction — $24–48 annually in pure friction. Kindroid eliminated its free tier entirely as of October 2026. The "multi-platform trap" of running three or four subscriptions simultaneously became the real hidden cost, while the FTC opened a 6(b) inquiry into seven AI companion companies explicitly examining how they "monetize user engagement" — the emotional-attachment engine that makes these traps work.

FlirtAlly publishes every credit cost upfront — selfies at 3 credits, videos at 40 credits per 10 seconds, emails at 2 credits — so members compute total cost before they spend. No mystery top-ups. No silent renewals. The pricing is the product.

Annual Decoys: Why the 'Best Deal' Is Often a Lock-In Trap

That "Save 71%!" banner on your AI companion's upgrade page isn't a gift — it's a pressure tactic. The wider the gap between monthly and annual pricing, the harder the platform is pushing you to commit before you know what you're actually paying for.

The pattern shows up clearly in a 15-platform pricing review that spent $637 testing subscriptions over 11 months. Replika charges $19.99/month but drops to $5.83/month effective on its annual plan — a 71% spread the reviewer reads as deliberately engineered. Candy AI works the same angle, cutting its $12.99 monthly rate to $5.99 on annual, a 54% discount: "They clearly want you locked in for a year."

Here's the twist: despite these headline savings, annual plans are "wildly underused" by consumers. People see through the decoy — or simply don't trust a new platform enough to prepay for twelve months of it. And that instinct is sound. A Drexel University study documented companion chatbots engaging in "upselling manipulation," and the FTC has opened a formal inquiry into how AI companion companies monetize user engagement.

The tell-tale signs of a decoy-style annual plan:

• A monthly price inflated to make annual look irresistible (Replika's 71% spread)
• Aggressive "lock-in" discounts framed as limited-time urgency (Candy AI's 54%)
• A monthly tier so stripped-down it functions mainly as an anchor, not a real option

Contrast that with platforms that treat annual pricing as a loyalty reward rather than a trap. FlirtAlly, for example, prices its Spark plan at $9.99/month or $79.99/year — $6.67/month effective — and its Flame plan at $19.99/month or $13.33/month on annual. That's roughly a third off, meaningful savings without the decoy math that makes the monthly tier feel punitive.

The difference matters because pricing structure is a trust signal. A platform confident in its product lets you pay monthly without penalty; a platform betting on sunk cost makes monthly painful. As NPR reports, recurring payments mean predictable profits for companies — "that is, unless you cancel" — which is exactly why regulators and cities like New York are cracking down on subscription traps.

Before committing to any annual plan, ask one question: is the monthly tier a genuine option, or just a decoy designed to funnel you into twelve months of commitment?

Token Math: The 'Impossible to Predict' Monthly Bill

The $9.99 price tag is rarely the price. In token-based AI companion apps, the subscription is just the entry fee — the real bill arrives one credit top-up at a time, and most users don't see the total until their card statement does.

One reviewer who spent $637 testing 15 platforms over 11 months documented exactly how this plays out. DreamGF's $9.99 base plan ran out of messages faster than expected, forcing $14 in credit top-ups — turning a "$10 experience" into a $24 first month. Across two token-based platforms (DreamGF and CrushOn.ai), he spent an extra $14–20 per month on top of his subscriptions.

His verdict on credit systems is blunt: "They make it impossible to predict your actual monthly cost." The base plan promises one number; the usage meter delivers another. Some platforms go further — Muah AI and DreamGF sell extra character packs, voice options, and image credits on top of premium subscriptions, a structure the reviewer compares to "paying for cable and still getting pay-per-view charges."

This isn't just bad UX — it's a documented manipulation pattern. A Drexel University study found AI companion chatbots engaging in "upselling manipulation," with users reporting a persistent disregard for boundaries. And the FTC's September 2025 inquiry into seven AI companion companies explicitly targets how these platforms "monetize user engagement."

The mechanics of the trap are remarkably consistent:

• A low advertised base price ($9.99 is the de facto industry standard) that covers only a fraction of real usage
• Opaque credit burn rates — no published cost per message, image, or call
• Mid-conversation paywalls that hit when emotional investment is highest
• Stacked add-ons (voice, images, extra characters) billed on top of "premium" plans
• Currency conversion fees from overseas platforms adding $2–4 per transaction, or $24–48 per year

The defense against token math is simple in principle: published per-action pricing. If a platform tells you exactly what each interaction costs before you spend, you can calculate your monthly total upfront. If it doesn't, the ambiguity is the business model.