AI companion cost watch: July 2026 checks before you pay

This month

• Sponsored linksNomi, HeyGF, and Candy AI may appear as disclosed outbound links where relevant.
• ScopeOnly apps with current Companion Cost pages and source coverage are linked from this watch page.
• Main riskReal cost still depends on media use, credits, coins, billing platform, refund limits, and privacy terms.
• Testing statusBased on linked official pricing, help, terms, and privacy pages, plus public signals. Paid hands-on test: not completed yet.

Six checks to run now

• AI identityDoes the product clearly tell you that you are interacting with AI?
• Generated mediaCan you tell when an image, voice, or video is synthetic?
• Data controlCan you obtain and delete chat or interaction history without hunting for a contact address?
• Training consentIs sensitive interaction data kept out of model training unless you separately agree?
• Exit controlCan you end the interaction directly, without the companion prolonging the conversation?
• Usage remindersDoes extended use trigger a clear reminder instead of silently continuing?

Research signal · Accepted July 16

A disclosure alone is not the whole safeguard.

A newly accepted JMIR paper argues that responsible companion design also needs transparent limits, escalation paths, and ongoing evaluation. That supports checking what the product lets you do after the first AI label.

Why it matters

• Faster comparisonSix routes are shown side by side before one checkout page frames the whole decision.
• Better external hookThe table is easier to cite in forum answers, directories, newsletters, and short posts than a long guide.
• Safer buying pathThe CTA points to cost and risk checks before any official pricing page.