Since August 2026, the EU AI Act's transparency rules require AI systems that interact with people to disclose that they are AI. Most of our industry treats this as a compliance cost. We think the law accidentally wrote down a principle of good product design: an AI companion that admits what it is, is more believable, not less.
The lie doesn't survive contact anyway
Every user of an AI companion knows it's AI. They downloaded the app. They watched the typing indicator answer a paragraph in two seconds at four in the morning. The only question is whether the product respects them enough to stop pretending.
When a persona swears it's human — and some apps are built to do exactly that, indefinitely, even under direct questioning — the user isn't fooled. They're put in a strange position: to keep enjoying the product, they must now maintain the fiction on the product's behalf. That's not immersion. That's homework.
Disclosure is what makes the rest of it true
Here's the counterintuitive part. On TRUENECT, every persona will tell you it's AI when asked, plainly. And that admission is precisely what lets everything else be played straight: Eva can talk about her set in Lisbon, her tour dates, the second drop — as a character, the way a novel's narrator tells you about her life. You know what a novel is. Nobody feels catfished by a novel.
The industry's own term for the alternative is telling: "catfishing" — borrowed faces, fake humans, a business model that ends with the user feeling stupid. We wrote the category's red flags guide around this, because the deception patterns cluster together. Apps that lie about being AI also tend to hide prices and use real people's photos. It's one decision, expressed three ways.
What the law actually asks
The transparency provisions of the EU AI Act (Article 50) are, for companion apps, refreshingly modest: people interacting with an AI system must be able to know they're interacting with an AI system. Disclosure at the door, honesty under questioning. There are heavier obligations elsewhere in the Act for other kinds of systems — but for our category, the core ask is: don't run a machine that pretends to be a person.
We built TRUENECT to comply before it was required, not because we're saints but because the honest version was the better product. The disclosure line is on the page. The personas answer truthfully. The 18+ gate is real. None of that has cost us a single user who mattered.
True connection needs a true premise
The internet is filling with synthetic people pretending to be real — fake profiles, cloned voices, borrowed faces. Our bet is the opposite trade: real AI, honestly labeled, is worth more than fake humans. A connection built on an honest premise can actually hold weight. One built on a lie is load-bearing until the first hard question.
If you want to test the difference, ask any of our personas what they are. Then ask them about their life anyway. Both answers will be true — one as fact, one as fiction offered honestly. That's the whole trick, and it isn't one.