Truenect, the AI companion from the studio that publishes SheAI News, opens its homepage with a single promise: "Someone who remembers you." The product is built around that line. Its personas keep a memory in three layers that holds for weeks, they may write first but not often, and they talk with users in text and voice.

Three layers of memory

The Truenect homepage describes the memory as "3-layer": recent messages, extracted facts and rolling summaries, a memory "that holds across days and weeks". Each layer does a different job. Recent messages are the live thread of the conversation, extracted facts are details lifted out of it, such as a name, and summaries condense what was said before.

The page gives an example. On day one, a user mentions that Rex, their dog, was waiting at home. Fourteen days later, after a rough day at work, the persona Luna asks whether Rex at least made them smile. "Weeks later they bring it up, the way a real friend would," the page says.

The memory is not meant to be a transcript. A persona "remembers what's important: your name, key moments, ongoing topics", according to the help center, while "trivial details fade, just like in real conversations". Forgetting is part of the design.

They write first, within a limit

Truenect's personas do not only answer. The homepage says they have "their own schedule, their own moods" and may send a message when something reminds them of the user: a sunset, or a question about how an interview on Tuesday went. The same page sets a ceiling: "Never spam: one message per 12 hours, at most." The help center adds that users who prefer silence can switch proactive messages to a quiet mode in Settings.

12 hours
the shortest gap between two messages that a persona starts on its own, as the Truenect homepage states (read on 1 October 2026)

Personas that say what they are

Every persona has a name and a backstory; Luna, Mia, Sakura and Eva are among those presented on the homepage, which says that "every persona tells you what they are". Conversations run in text and voice. A free tier includes a daily message allowance and text conversations, the help center says, and voice messages come with the paid plans.

The terms of service put the disclosure in writing: "All personas on TRUENECT are AI characters, not real people," and users "always know" they are talking to an AI. The same section says AI conversations are not a substitute for therapy, medical advice or human relationships.

In the European Union, this kind of disclosure is now a legal duty. Article 50 of the AI Act, which applies from 2 August 2026, requires AI systems that interact directly with people to be designed so that those people are informed they are interacting with an AI system, unless that is obvious.

Creator twins

A second part of the product is aimed at creators. On the creators page, Truenect offers a "creator twin": a persona built with a creator from their photos, voice and phrases, under hard limits the creator sets. The page states that creators keep 55 percent of what their fans spend with the twin, that one switch pauses it at any time, and that fans "always know it's an AI twin". Creators see aggregate insight such as top topics and message volume, "never private chats or personal fan data".

Why memory is a trust question

Memory is what makes a companion feel personal, and it is also what makes it sensitive. In a study published on 24 July 2026 in the Journal of Computer-Mediated Communication, Hsuen-Chi Chiu and Jeremy Foote interviewed 15 users of companion chatbots. They found that human-like design lowers the threshold for sharing, and that users protect themselves in advance, for example with fake names, because correcting what they have shared "feels futile". For some, losing the conversation's memory was the bigger worry: "a relational reset feels more threatening than a data breach".

AI companion chatbots feel like intimate confidants but exist as part of corporate data systems.
Hsuen-Chi Chiu and Jeremy Foote, Journal of Computer-Mediated Communication, July 2026

Earlier privacy reviews of the category were critical. When the Mozilla Foundation reviewed 11 relationship chatbots for its *Privacy Not Included guide in February 2024, every one of them received a warning label, ten failed its minimum security standards, and most did not let users opt out of having their intimate chats used to train AI models.

Truenect's privacy policy, last updated on 8 August 2026, sets out its own answers. It says conversations are not used to train AI models. Two rights work without writing to anyone: "Export my data" in Settings downloads everything held about the user as a JSON file, and "Delete account" erases the account, every conversation and "all three memory layers your personas keep about you", with permanent removal after a 30 day grace period. The help center says conversations are encrypted at rest and lists "no ads, no data sales" among the product's differences. These are the company's own commitments, as its pages state them.

How to try it

Truenect is at truenect.com. Choose "Start free" to meet a persona; the page says no credit card is required, and the plans are listed further down the same page.