Consistent criteria
Every app is measured against one published model, so a 7.0 means the same thing on every listing.
Categories
AI girlfriend apps 80 AI roleplay apps 84 NSFW AI chatbots 23 Adult AI video 15 NSFW AI games 12 Image generators 4Best lists
All 8 curated lists Anonymous apps Voice-enabledDecide
How we score Candy AI Shorts New AI Gallery New Secrets AI Shorts New Live Cams NewCompany
Advertise Submit a tool Report a tool ContactOur review process
Every NSFWRating score comes from a published weighted model: 11 criteria for AI companions, and a matching set of 11 for image and video generators, so 173 apps can be compared on identical terms.
What guides every review
The same rules apply to every product, regardless of popularity, commission rate or commercial relationship.
Every app is measured against one published model, so a 7.0 means the same thing on every listing.
Claims must trace back to hands-on testing, the vendor's own policy pages, or reputable public sources.
Scores reflect our own evaluation, not unverified star averages, and not paid placement.
Pricing, limits, features and privacy terms are re-verified, and each listing records when it was last checked.
The weighted scoring model
The overall score is a weighted average of 11 criteria. This is the model for AI companions; generators have their own below. Higher-impact areas such as chat quality, quality and privacy carry more weight than smaller factors. The weights below are the live values used by the scoring engine and total 100%. A criterion we have not scored yet is left out of that app's result and the remaining weights are re-spread to 100%, so a gap never counts as a zero.
What we test
Each category has a defined set of checks. Reviewers record evidence, note limitations, and assign a criterion score before the overall result is calculated.
We assess coherence, repetition, tone control, refusal behaviour, realism and response consistency across repeated sessions.
We review account and chat deletion, what signup demands, data-training rules, and how clearly the privacy policy states them.
We judge the overall quality of what the app gives you, whether that is images, video, streams or text, across repeated sessions.
We test how well an app holds on to preferences, character detail and conversation context between messages and sessions.
We review character builders, persona controls, scenarios, saved presets and how deep the available adjustments go.
We verify daily limits, starter credits, which features are locked, trial length, signup friction and any hidden restrictions.
We examine age gates, consent rules, fictional-only boundaries, moderation, and the policy on uploading images of real people.
We count what the platform actually offers against what its category typically offers, and whether each feature is usable rather than a checkbox.
We weigh what you get for the money against how clearly plans and credits are priced, plus refund rules and renewal behaviour.
We time the path from signup to first result and judge interface clarity, mobile web and app quality, speed and friction.
We test responsive behaviour, mobile-web usability, app availability, navigation speed and interaction stability.
AI image and video generators
Chat quality and memory say nothing about an image generator, so generators swap them for image quality, video quality and prompt control, and customization drops out. Everything else is checked the same way as for companions. These are the 11 generator criteria and their share of the score.
We judge still images for realism, anatomy, skin and hands, detail at full size, and how consistent results are across repeated runs of the same prompt.
We judge clips for smooth motion, faces that hold together frame to frame, clip length, resolution and how often a generation fails or needs a retry.
We test how precisely prompts steer the result: negative prompts, seeds, style presets, model choice, img2img and inpainting, and whether the tool ignores what you asked for.
Reviews scored before this model keep their previous ten criteria (chat quality, memory, media quality, customization, free tier, pricing clarity, privacy, mobile experience, safety controls, user rating) until they are re-tested. The score panel on each review lists exactly which criteria were used, and the same rule applies: anything not scored is left out rather than counted as zero.
How we gather data
Hands-on evaluation is combined with vendor documentation and reputable public sources. Every material claim should connect to something observable.
Confirm product scope, plans, policies, supported platforms and current claims.
Use the product across repeated sessions, recording results against each criterion.
Pricing, privacy, feature and safety claims are checked against the vendor's own pages.
Criterion scores are weighted and converted into the final score out of ten.
Listings are revisited when plans, limits, features, policies or quality change.
Editorial independence
Featured placements and affiliate relationships never change a product's editorial score. Disclosures stay visible, and criticism is not removed for commercial reasons.
The evaluation is completed against the criteria above, independently of advertising, commissions or partnership terms.
Read the affiliate disclosure →Featured positions are identified wherever they appear and are never blended into a score calculation.
How featured placement works →Limitations, policy concerns and pricing weaknesses stay in the review even when a brand advertises with us.
Report a correction →How often we update
Prices, free-plan limits, features and policies move constantly in this category. High-impact listings get priority, and every listing shows the date it was last checked so you can judge how fresh it is.
Common questions
How the scores are produced, what they include, and what they deliberately do not.
No. Scores are editorial, calculated from the weighted criteria above. A moderated public-sentiment rating can be added as an optional extra criterion, and most reviews leave it out.
No. Scores are determined independently of commissions, sponsorships, coupon agreements and advertising. Paid placements are labelled and kept separate from the editorial score.
Hands-on testing is the standard for a full review. Some listings are scored from the vendor's own documentation and observable public evidence instead; where that is the case the review says so rather than implying a deeper test.
Rankings move when pricing, free-plan limits, features, privacy terms or safety rules change - or when a competing product improves. Each listing records when it was last checked.
It is a weighted average out of 10 across 11 criteria. Across the 173 apps scored today the range runs from 4.6 to 8.8, with an average of 6.6 - so a 7 is genuinely strong rather than average.
It is left out of that app's result and the other weights are re-spread to 100%. A gap never counts as a zero, and the score panel on each review lists exactly which criteria were used.
Send a factual correction with supporting evidence through the report form. Corrections are reviewed editorially and never require a commercial relationship.
Browse every reviewed product, check category rankings, and inspect privacy and pricing before you choose.