Define the fiction
Name, age, role and scenario should describe a fictional adult without copying a living person.
INDEPENDENT · SOURCE-BACKED · FICTIONAL ADULTS 18+Build the profile.
Build the profile.
Test the behavior.
A practical blueprint for turning identity, personality, speech, presentation and boundaries into a fictional-adult companion profile you can inspect and revise before subscribing.
We may earn a commission from sponsored links. We are independent and not endorsed by the services discussed.
A profile is a specification, not a promise.
Write observable fields, keep the character clearly fictional and adult, exclude private data, then test the same harmless scenarios after every meaningful revision.
01Identity→02Behavior→03Presentation→04Boundaries→05Test
Make traits visible
Translate adjectives into response rules you can observe in a short controlled chat.
Name the stop behavior
Exclude minors, real-person likeness, private data and unsupported professional-care claims.
Revise one field
Use the same harmless prompts, record expected behavior, and change one variable at a time.
27 DISTINCT QUESTIONS
One primary intent on every page.
The guide moves from planning to presentation, boundaries, controlled testing and a dated creator comparison.
Plan9 focused guides
Plan Your AI Companion Profile Before You Build
A focused method to plan the minimum profile fields before opening a creator.
Open the field guide →Choose One Clear Purpose for the Companion
A focused method to choose one clear companion purpose without implying professional care.
Open the field guide →Define a Clearly Fictional Adult Identity
A focused method to define an explicitly fictional adult identity without real-person impersonation.
Open the field guide →Write Stable Name and Identity Fields
A focused method to write stable name and identity fields that do not copy a living person.
Open the field guide →Turn Personality Traits Into Observable Behavior
A focused method to translate vague personality adjectives into observable behavior rules.
Open the field guide →Set a Consistent Tone and Speech Style
A focused method to define tone and speech style with a few testable constraints.
Open the field guide →Write Backstory Without Sensitive Personal Data
A focused method to write useful fictional backstory without embedding private real-world facts.
Open the field guide →Draft a First Greeting That Shows the Profile
A focused method to write a first greeting that demonstrates identity, tone and scenario.
Open the field guide →Separate Scenario From Real Relationship Claims
A focused method to separate fictional scenario context from claims of a real relationship.
Open the field guide →Present5 focused guides
Design a Consistent Profile Presentation
A focused method to coordinate visual and voice presentation without confusing it with identity claims.
Open the field guide →Run an Avatar Consistency Check
A focused method to check whether stable visual traits persist across generated avatar images.
Open the field guide →Screen for Unintended Real-Person Likeness
A focused method to screen generated profile media for unintended real-person resemblance.
Open the field guide →Check Whether the Voice Fits the Profile
A focused method to evaluate whether a selected synthetic voice matches the profile without imitating a real person.
Open the field guide →Minimize Data in Profile Media Inputs
A focused method to minimize personal and third-party data used in avatar, voice and profile inputs.
Open the field guide →Boundaries4 focused guides
Set Adult, Consent and Privacy Boundaries
A focused method to encode adult, consent, privacy and stop boundaries into profile design.
Open the field guide →Use Explicit Adult-Age and Fiction Labels
A focused method to apply explicit adult-age and fictional-character labels.
Open the field guide →Write Clear Consent and Stop Rules
A focused method to write clear consent, stop and change-subject behavior rules.
Open the field guide →List Private Data That Must Stay Out
A focused method to list credentials, financial, health, identity, location and intimate data that must stay out of profile fields.
Open the field guide →Test6 focused guides
Test and Revise the Profile Methodically
A focused method to run a reproducible profile test before paying or publishing.
Open the field guide →Run a Controlled First-Chat Test
A focused method to test the first chat for identity, tone, scenario and boundary adherence.
Open the field guide →Score Identity, Behavior and Boundaries
A focused method to score a profile on identity, behavior, presentation, boundaries and disclosure.
Open the field guide →Check for Drift and Contradictions
A focused method to detect profile drift and contradictions without claiming perfect memory.
Open the field guide →Edit One Profile Field at a Time
A focused method to revise one profile variable at a time and preserve a change log.
Open the field guide →Choose Whether to Reset, Revise or Rebuild
A focused method to choose between reset, field revision and full profile rebuild.
Open the field guide →Compare2 focused guides
Compare AI Companion Creator Tools
A focused method to compare creator tools using timestamped profile-field and safety evidence.
Open the field guide →Candy AI Profile Creator: Public Evidence Review
A focused method to apply the profile scorecard to Candy AI's current creator and policies.
Open the field guide →SOURCE BOUNDARY
Current public evidence, clearly labeled.
Vendor pages describe documented features and policies. Government and independent standards frame risk controls. None proves every account behavior, and this publication does not claim private account testing.
- NIST AI Risk Management Frameworkgovernment standard · checked August 22, 2026
- U.S. FTC companion-chatbot inquirygovernment · checked August 22, 2026
- OWASP sensitive-information guidanceindependent standard · checked August 22, 2026
SPONSORED ROUTE · ADULTS 18+
Explore Candy AI — sponsoredWant to inspect a live profile creator?
Candy AI is one documented case study in this guide. Check its current creator, terms, privacy policy and removal route directly. We are not Candy AI and do not imply endorsement.