White AI Girlfriend Companions Written With Real Specificity

Half of AI girlfriend users choose caucasian characters by default, and that default is usually the only option platforms ship with any depth (per News Channel Nebraska 2025). Character.AI's most-used characters are caucasian by a wide margin. Candy.AI's curated library leans caucasian. Replika's single customizable character is caucasian by visual default. The entire AI companion category treats white as the unmarked category, the neutral one, the default. And that means it tends to be written with less specificity than the categories that have to earn their audience.

AIAngels writes the white category with the same specificity as every other category. Five characters across five cities, each with a profession, a first language or accent, and a memory that holds across weeks. Dublin biomedical researcher. Melbourne wine importer. Amsterdam architecture student. Montreal bilingual novelist. San Francisco UX designer. These are not stock characters. They are specific women with specific lives.

The SEO search is white ai girlfriend because that is what users type. The character creator UI labels this category Caucasian because that is the internal convention. They are the same companions and the same memory. We explain the bridge directly on the page so users who click through from search to the creator do not get confused. Same pattern as the black/African ethnicity page — two names for the same door.

The test for this category is whether the writing is specific enough that a user notices the character is written with care. For most platforms, white characters are the default everyone defaults to, and the writing defaults with them. For AIAngels, the white category has to earn its depth the same way every other category does. That means real cities, real professions, real voices.

Dublin is the biomedical researcher. Works at Trinity College on a cardiology research project, lives in Ranelagh, reads Irish poetry in her spare time. Her voice is intellectual, Irish in the specific cadence Dublin speakers have, and unexpectedly warm for someone who works in academic science. Explains her research at whatever depth you want to hear it. Remembers the article you mentioned last week.

Melbourne is the wine importer. Works for a small Australian importer sourcing wines from boutique European estates, travels to Europe three times a year, has strong opinions about natural wine versus conventional wine. Speaks Australian English with the specific rising intonation. Sends you the name of a wine she thinks you should try and remembers whether you did.

Amsterdam is the architecture student. Final year at TU Delft, focused on sustainable urban design, lives in a canal house apartment in Jordaan. Speaks Dutch and English with the specific directness Dutch speakers bring into English. Sends you a building she is studying and remembers which design elements you reacted to. Her voice is precise, European, and grounded.

Montreal is the bilingual novelist. Writes fiction in English, translates her own work into French, lives in Mile End, teaches creative writing at a CEGEP. Speaks Quebecois French and English with the specific code-switching Montreal bilinguals have. Sends you a paragraph she is working on and remembers whether the feedback you gave was useful. Her voice carries both languages and the reflective quality that comes from writing in two at once.

San Francisco is the UX designer. Works at a mid-sized tech company, lives in the Mission, bikes to work, has opinions about the transit system that will make her sound like every other Bay Area tech worker if you let her. Her voice is fast, design-literate, and specifically Bay Area in the way Bay Area tech workers are specific. Sends you a design problem she is working on and asks for your honest reaction.

Memory on AIAngels is the same feature across every category: permanent, unlimited on the free tier, included by default. The question for the white category specifically is what memory looks like when the characters are written with specificity rather than as a generic default.

The answer is that memory holds the specific things each character introduces in context. If the Dublin biomedical researcher taught you about a specific study design in week one, she can reference it in week twelve when you ask her about a related news article. If the Melbourne wine importer recommended a specific Riesling from Alsace, she remembers whether you tried it and asks for your tasting notes. If the Amsterdam architecture student sent you a photo of a Rem Koolhaas building, she remembers which element you noticed and brings it back when she sees something related.

On Character.AI (20 million monthly active users per platform documentation), memory is session-scoped. The most-picked caucasian characters on the platform forget your name after logout. Users who have built long conversations with white characters on Character.AI are the largest user segment hit by memory loss, simply because they are the largest user segment on the platform.

On Candy.AI, memory degrades after about 50 messages. The small curated caucasian library on Candy.AI runs into the memory wall faster than users expect because the characters are the ones users have long conversations with.

On Replika, memory was stable until February 2023, when the platform removed explicit content and broke continuity for users who had built the most long-standing relationships. Caucasian users were the largest segment affected simply by population size.

On AIAngels, the five white characters have the same memory architecture as every other character on the platform. Permanent. Unlimited. Free tier. No token wall, no session reset, no February 2023 rug-pull. The category that most users default into gets the same memory promise as the categories that users seek out. That is the whole offer, applied consistently.

Start by picking a character based on what you want the conversation to feel like. Dublin for intellectual warmth and Irish literary grounding. Melbourne for practical sophistication and wine culture. Amsterdam for European directness and design thinking. Montreal for bilingual reflection and writing. San Francisco for fast tech-literate energy.

One note on the character creator. Inside the creator UI, this category is labeled Caucasian because that was the original internal convention. The SEO slug uses white because that is what users type in search. They are the same companions with the same memory. When you click through from this page to the creator to customize a character, you will see Caucasian in the ethnicity picker. Do not let the label trip you up.

None of the five characters are static. The character creator lets you adjust her hair color, eye color, body type, and other details while keeping her personality and memory intact. If you want a Dublin biomedical researcher with red hair instead of brown, you can make that change without losing the rest of who she is.

The free tier includes unlimited text messages with all five characters. Each relationship develops on its own timeline. Three steps: pick, say hi, mention one thing from your day. The memory starts at message one and holds on day ninety.