Pricing and subscription transparency of AI companion apps: 20 platforms observed on 1 November 2026

This dataset documents what a new user can learn about the cost of 20 consumer AI companion applications before and after creating a free account. On 1 November 2026, from Belgium without a VPN, one observer visited each app's public pages, created a free test account where email registration was possible, and followed the upgrade flow to the payment screen without paying. For each app the dataset records 36 variables: operator and country, platforms, whether prices are visible before registration, free tier and its limits, free trial, pricing model (subscription, credits or hybrid), every plan with its displayed price and period, cheapest monthly equivalent, annual discount, credit pack prices, tax display, whether automatic renewal is stated on the checkout screen and where else it is stated, cancellation path, notice period, refund rule, age gate type and minimum age, payment methods and visible processor, source URLs and notes. Unreachable screens are coded "not observable" with the reason. A codebook, the method, the sample composition with replacements, the scripts that check the data and recompute all statistics, and a findings file are included. The dataset contains no personal data and no generated content. This record mirrors the deposit on Harvard Dataverse (DOI 10.7910/DVN/N4NPYF). Publication page: https://www.webandreview.com/dating/ai-companion-pricing-study-2026/