The skeptic's question deserves a straight answer: do people actually engage with accounts they know are synthetic — or is the whole category operators selling tools to other operators? The evidence is a decade deep now, and the answer is yes, with conditions worth understanding before you build.
The base evidence
Humans demonstrably form attachments to characters they know aren't real — that's not an AI claim, it's the entire history of media. Anime characters have million-follower fan accounts. VTubers — explicitly animated personas — built one of the most monetized creator categories on earth, with superchat revenues that embarrass most human streamers. Mascots sell cereal. The first virtual influencers accumulated millions of followers and major brand campaigns while openly disclosing they were CGI. Parasocial attachment binds to character, continuity, and voice — not to a birth certificate.
What the engagement looks like
Patterns visible across successful virtual accounts:
- Comments skew conversational and playful. Audiences treat the character as a character — they banter, ask questions in-fiction, and reward the operator's reply game. The comment section is where the account actually lives.
- Novelty spikes, character retains. "Whoa, this is AI?" drives early discovery; what keeps followers at month three is the same thing that keeps them on a human account — a niche worth following and a personality worth hearing from.
- Saves and shares track content utility, not the poster's biology. A great outfit reference or workout gets saved regardless.
Where synthetic accounts genuinely underperform
Honesty section, because the conditions matter:
- Lived-experience niches. Parenting, health journeys, finance credibility — where the testimony is the content, a synthetic narrator has nothing real to testify. Avoid these lanes entirely.
- Trust-gated actions. A character can inspire and entertain; recommendations carry less weight than a trusted human's. Affiliate conversion runs lower per follower; volume and fit make up for it.
- The deception discount. Engagement collapses in exactly one scenario: the audience discovers the account pretended to be human. Every unmasking case study reads the same. Disclosure isn't just compliance — it's the inoculation against the only true engagement killer.
The variable that actually predicts engagement
It isn't render quality, and it isn't even disclosure (disclosed accounts grow fine). It's operational liveness: does the account reply within the hour, run recurring bits, post on rhythm, participate in its niche? Algorithms weight early interaction; audiences mirror the energy they receive. A gorgeous gallery with a dead comment section reads as a billboard — and billboard engagement is what it gets. This is the 80% of the job that no generator does for you.
One technical footnote: visual consistency is an engagement variable too. Followers bond with a face; an account whose character subtly morphs between posts feels uncanny in a way audiences can't articulate but reliably unfollow.
The verdict
Real engagement: yes — conditional on a clear niche, an alive operator, honest disclosure, and a character that stays the same character. The synthetic part sets your production costs ($0.25 an image); the human part sets your engagement. Build the character in ten minutes, $19 — then go be the personality behind her.