# How We Review Adult AI Chat: Evidence and Scorecard Canonical: https://nsfw-ai-chat.org/how-we-review/ Language: en Published: 2026-10-07 Updated: 2026-10-07 Publisher: NSFW AI Chat Read our eight-criterion adult AI chat scorecard, source standards and evidence limits. Public-source comparison without fabricated private-session ratings. 18+ only. AI-generated fictional adults. Public-source editorial comparison; no paid or private chat sessions. # How we review adult AI chat 18+ only. AI-generated fictional adults; no real people depicted. Portraits are editorial illustrations. We compare publicly accessible product, help and policy pages. We did not create accounts, buy plans, send private chat messages, audit security or measure response performance. Our ranked reading order evaluates the usefulness of the documentation for an adult’s decision, rather than awarding fabricated conversation scores. ## Eight consistent criteria Identity is a prerequisite. We apply these eight criteria to every identifiable subject including JoiAI’s documented guided-character workflow. Weight 2 gives greater decision importance than weight 1; these weights guide the qualitative reading order and are not converted into a numerical performance rating. ## Adult scope and boundaries A public 18+ rule and fictional-character/content constraints. Documented policy is distinct from tested enforcement. ## Conversation workflow Published text-chat entry point, character selection and roleplay mechanics. We do not measure reply quality. ## Character control Published appearance, personality or scenario controls and whether current access is clear. ## Continuity and memory Published history/context behavior plus disclosed forgetting or inconsistency; never a measured recall score. ## Voice and optional media Published voice, image or video interaction as an extension of chat, with tier limitations identified. ## Entry access Published free access, sign-in and message limits; the current account displays the applicable allowance. ## Billing clarity Published subscription/token mechanics, renewal/cancellation, and dated amount availability. No checkout verification. ## Data and deletion clarity Published processing, moderation access, retention/deletion and named geographic restrictions; a policy review is not a security audit. ## Evidence status, not star ratings - Official documentation: Dated provider pages support the described controls and billing rules. - Reported price: A named publication’s dated amount is attributed and kept separate from an official checkout quote. - Account-specific selection: The selected plan, region, promotion and optional purchases determine the final charge. If we ever publish numeric scores, an 8 must represent directly repeatable evidence across the defined task; a 5 must represent a materially limited or conditional result on that same task. We have not collected such session evidence for this release, so neither value is assigned. Strong documentation is not a substitute for measured output. ## What we checked and when Public information was checked October 7, 2026. CandyAI’s payment and privacy terms establish an entry threshold, renewal and token mechanics, and retention statements. LoveScape’s creation and limitations articles document controls and context limits. GoLove’s public payment page establishes term totals and Stars, while its current creator describes appearance, personality, voice and relationship settings. JoiAI’s official terms explain Premium, bot access and Neurons; its September cash-price report is attributed to AI Miracle. Our original screenshots show safe public documentation. Their captions identify the document and capture date. External product addresses are handled by the service-opening controls; this site’s article links stay within the publication. We preserve the underlying source record privately and never imply that a documentation capture is a paid conversation test. ## Why this reading order LoveScape comes first for detailed character and dialogue preferences alongside candid context limitations. CandyAI follows for character-first setup and specific entry mechanics. JoiAI follows for guided character creation with multiple payment layers. GoLove comes fourth for catalogue discovery and a three-step creator. This reading order follows documented workflows rather than a cross-currency price ranking. The comparison table uses the same criteria throughout. ## Adult AI chat categories require different evidence Adult AI chat comparisons need to distinguish character discovery, persistent companion setup and general AI workspaces. Their public controls can solve different tasks. The following category examples were checked October 7, 2026; they explain our criteria without joining the four-subject ranking or receiving performance recommendations. ### Character discovery and companion setup differ GirlfriendGPT combines community character discovery with own-character creation and separate image tools. OurDream AI separates appearance, personality, voice and relationship choices into setup steps. Those examples show why we distinguish selecting an existing persona from defining one: both can precede chat, but they require different controls and entitlement checks. SpicyChat documents personality and greeting fields, plus public, unlisted and private character visibility. A greeting supplies an opening; personality supplies continuing instructions. Visibility determines character distribution, so we do not treat a private-character setting as evidence of confidential conversation storage or audited data security. Kindroid describes personality and backstory setup with text, voice and video channels. Nomi describes custom backstories and preferences alongside individual conversations and group chats. Channel availability and multi-character format are separate from recall quality; none of these public descriptions proves that a long exchange preserves every instruction. ### General AI tools need a separate check Venice AI presents writing, code and multiple media-generation formats in a broader workspace. A flexible text interface is different from a curated companion catalogue. We therefore assess chat workflow separately from image tools, rather than count every generated-media format as a benefit for an adult character conversation. Janitor AI belongs to the wider category discovery set rather than this four-service shortlist. A category mention does not transfer another provider’s features, prices or performance into our ranking. The reviews focus on CandyAI, LoveScape, JoiAI and GoLove, each matched to its own public product information. We do not infer which LLM a reviewed service uses from a convincing reply. A system prompt can express character instructions, while content filters concern permitted output; neither label proves continuity or unrestricted access. Our scorecard requires product-specific evidence instead of transferring assumptions from another interface. ## Limitations and corrections Prices retain their source currency, date and status. The JoiAI amounts are a reported September snapshot; its current official selection opens after registration. Provider memory, security and content descriptions are attributed rather than treated as measured results. We do not create accounts or pay to produce evidence. Read the publication profile, the AI illustration policy or contact information for questions and corrections.