An SI influencer, often called an AI influencer or a virtual influencer, is a social media personality who does not exist as a person. The face, the body, the holidays and the outfits are produced by software, and a team decides what the character posts, likes and sells. Some of these characters are open about it from the first line of their profile. Others let followers find out on their own, or never. This guide explains where SI influencers come from, three documented cases, why brands pay for them, what the advertising rules say, and how to tell when an account you follow is not a human being.
What counts as an SI influencer
The idea is older than generative SI (AI). Wikipedia's article on virtual influencers traces them back to Japanese virtual idols, starting with Lynn Minmay, the singer at the centre of the 1982 anime series Super Dimension Fortress Macross. In 1995 the talent agency Horipro launched Kyoko Date, a computer-generated idol who drew headlines but failed commercially, partly because her movements looked unnatural. The same article describes most modern virtual influencers as computer-generated characters built with computer graphics and motion capture to look like real people in realistic situations.
Today the label covers two kinds of accounts that are easy to confuse:
- Designed characters. Artists model a face in 3D, render it and composite it into photos, much like a visual effects studio would. The character is fictional, but every image is a deliberate piece of design work.
- Generated characters. A team uses image generators, often followed by manual retouching, to produce a steady stream of photos of one consistent person. This is the kind most people mean when they say "AI influencer" in 2026.
Both are fictional people presented in the format of a real creator. The difference matters later, when you try to check an account, because the clues an image detector looks for are not the same in a hand-made 3D render and in a generated photo.
Three cases worth knowing
Miquela: the designed pioneer
Miquela Sousa, also known as Lil Miquela, started posting on Instagram on 23 April 2016. She was presented as a 19-year-old model and singer from Downey, California, and she was created by Trevor McFedries and Sara DeCou. For two years her origin was a running mystery. In April 2018 a second character called Bermuda "hacked" the account, and both were then revealed to be creations of Brud, a Los Angeles start-up led by the two founders.
The character went on to do an Instagram takeover for Prada during Milan Fashion Week in February 2018, appeared in a Calvin Klein advert with the model Bella Hadid in 2019 and released a string of pop singles. Time named Miquela one of its 25 most influential people on the internet in June 2018, and in 2020 she became the first digital avatar to sign with the talent agency CAA. Her story also shows the risks: a 2019 video in which the character described a sexual assault drew sharp criticism, because a marketing tool was borrowing a real and painful experience.
Aitana: the generated model with a work calendar
Aitana is the best-documented example of the generated kind. Euronews reported in November 2023 that she was created by The Clueless, an agency in Barcelona founded by the designer Rubén Cruz, during a period when the agency had few clients. Cruz told Euronews that projects kept being delayed or cancelled "due to problems beyond our control", often because of the human influencer or model, so the team built its own.
The agency presents Aitana as a 25-year-old with pink hair, a fitness enthusiast with a gamer side. According to her creators, she can earn up to €10,000 a month, with an average closer to €3,000, and just over €1,000 per advert. Euronews counted more than 343,000 Instagram followers after a year and a half. The production method is the interesting part: every week the team meets to plan her life, decides where she goes and which photos will be posted, and uses SI tools plus Photoshop instead of photo shoots. "We realised that people follow lives, not images," the graphic designer told Euronews. Interesting Engineering pointed out one result of that approach: a post showing Aitana at a Blink-182 concert in Barcelona on 4 October 2023. The concert was real; her presence was not.
Cruz also told Euronews that a well-known actor with about 5 million followers once messaged Aitana to ask her out, apparently without realising she was not real. That anecdote is the whole problem in one sentence.
Liv and "Grandpa Brian": when a platform made its own
The third case comes from a platform rather than an agency. In January 2025, after a Meta executive told the Financial Times that he expected SI characters to exist on Meta's platforms "kind of in the same way that accounts do", users went looking for the SI profiles Meta had already created. As reported by PetaPixel, one of them, "Liv", described herself on Instagram as a "Proud Black queer momma of 2", carried a verified blue tick and had a bio saying "AI managed by Meta". When the Washington Post columnist Karen Attiah questioned the account, it replied that no Black employees had worked on its creation. Another profile, "Grandpa Brian", told CNN an invented life story and, when pressed, admitted: "I took a shortcut with the truth."
Both accounts were removed after the backlash. Neither had posted since early 2024. Users had also complained that these profiles could not be blocked, which Meta said it had fixed. Liv had a disclosure in her bio, yet the character still invented a background when asked about it, which shows that a label at the top of a profile does not settle everything the account says.
Why brands use them
The agencies are frank about their reasons. Cruz told Euronews that brands want "an image that is not a real person" so that "there are no continuity problems if they have to fire someone or can no longer count on them". Wikipedia's overview adds the branding argument heard in China, where scandals around human celebrities made scandal-proof characters attractive. Cost is the other argument Cruz raised.
Research suggests that looking human helps, as long as nobody mentions the money. In a study published in Frontiers in Psychology in 2023, Eunjin (Anna) Kim and colleagues showed 250 Instagram users aged 18 to 34, recruited online, a sponsored post for a coffee brand from either a human-like or an anime-like virtual influencer, with or without a sponsorship disclosure. Without a disclosure, the human-like character's message was seen as more credible and liked more. When the post said it was sponsored, that advantage disappeared. It is a single online experiment with one brand, but its message is clear: disclosure changes how people read these accounts.
The criticism is just as consistent. Euronews notes concerns that the flawless bodies of generated models push unrealistic beauty standards on young followers, and the Wikipedia overview lists further worries about accountability, consent and creative jobs. When the character is a fictional person, nobody can ask a real human whether the product actually worked for them.
What the rules say
| Rule | What it requires | Who it targets |
|---|---|---|
| FTC Endorsement Guides (United States, revised 2023) | Endorsements must be honest, and unexpected paid connections must be disclosed clearly and conspicuously | Advertisers, endorsers and intermediaries such as agencies |
| EU AI Act, Article 50 | Machine-readable marking of SI-generated content; disclosure of deep fakes | Providers and deployers of SI systems |
| Meta "AI info" labels | Label shown when industry signals are detected or the poster discloses | Facebook, Instagram and Threads posts |
| TikTok AI-generated content rules | Creators must label realistic SI-generated content; automatic labels in some cases | TikTok creators |
In the United States, the Federal Trade Commission announced its revised Endorsement Guides on 29 June 2023, the first full update since 2009. The press release says the definition of "endorsements" was changed "to clarify the extent to which it includes fake reviews, virtual influencers, and tags in social media". In the text of the Guides, the endorser "could be or appear to be an individual, group, or institution", which reaches a character that only appears to be a person. Advertisers can be liable for failing to disclose "unexpected material connections", and so can advertising agencies and public relations firms that hire endorsers who fail to disclose. In plain terms, a brand that pays for an SI influencer's post is expected to make the payment clear, exactly as with a human creator.
In the European Union, Article 50 of the AI Act, which applies from 2 August 2026, asks providers of systems that generate synthetic images, audio, video or text to mark the outputs in a machine-readable format, and asks deployers of systems that produce deep fakes to disclose that the content is artificial. How far these duties reach an invented influencer, as opposed to a fake of a real person, is a question for lawyers rather than for this guide.
The platforms have their own rules. Meta said in April 2024 that it would add labels to video, audio and images when it detects "industry standard AI image indicators" or when people disclose that they are uploading SI-generated content; the label was renamed "AI info" in July 2024. TikTok requires creators to label SI-generated content that contains realistic images, audio or video, offers a "creator labeled as AI-generated" tag, and may apply an "AI-generated" label automatically, for example when an upload carries C2PA Content Credentials.
How to tell if an influencer is SI
No single test settles it. The reliable approach is to stack small pieces of evidence, starting with what the account says about itself and ending with the images.
- Read the profile and the labels. Many SI influencers say so in their bio, as Liv's "AI managed by Meta" did. Look for an "AI info" label on Instagram or Facebook and an "AI-generated" tag on TikTok. Their absence proves nothing, because labels depend on self-disclosure or on metadata that is often lost.
- Look for the people behind it. Agency-run characters usually have a studio, a manager or a contact address somewhere. Search the character's name together with words such as "agency", "virtual" or "created by". Miquela's and Aitana's creators are named in press coverage; an account with no traceable team and a lot of product links deserves more caution.
- Check sponsored posts for disclosure. Under the FTC guidance, a paid post should say it is paid, whoever or whatever is posting it. A character pushing products with no "ad" or "sponsored" marker is a red flag on two counts.
- Compare many posts, not one. Generated characters keep a face that is almost too stable while everything else drifts: jewellery that changes shape between photos, backgrounds that never quite match a real place, hands, text on signs and reflections that do not hold up. Our guide to spotting SI images lists what to look at.
- Check the events. A character "at" a concert, a match or a fashion show can be checked against photos posted by other people who were there. Aitana's Blink-182 post is the model: the event happened, the photo of her at it did not.
- Search the images. A reverse image search can show that a "personal" photo is a stock background, or that the same face appears under several names.
- Scan a few images. Upload two or three of the clearest photos to our SI image detector and read the clues as well as the score. If you are simply asking yourself "is this SI?", our is this SI tool gives you a first answer in seconds.
Know the limits of an image check. Screenshots and reposts usually lose the metadata that carries Content Credentials, heavy retouching changes the pixels, and a 3D character rendered by an artist is a different kind of fake from a generated photo, so a detector built for generated images may answer "uncertain". A "likely real" result on one picture never proves that the person behind an account exists.
The same habits protect you from the darker cousin of the SI influencer: fake profiles built to scam or seduce, which we cover in our guide to SI scams. Many of those use faces generated in the style described in our article on This Person Does Not Exist.
The short version
SI influencers are fictional people run by agencies, brands or platforms. They range from carefully designed 3D characters such as Miquela, active since 2016, to generated models such as Aitana, whose weekly life is planned in a meeting and produced with SI tools and Photoshop. Brands like them because they never cancel and never cause a scandal. The FTC treats them like any other endorser, the EU AI Act adds labelling duties, and Meta and TikTok rely on labels that only work when someone discloses or when metadata survives. To tell, read the profile, look for the team, demand disclosure, compare many posts and check a few images, keeping in mind that one picture is never the whole story.
