In late September 2025, a 20-something British actress with a warm smile and a film trailer reel became the most argued-about performer in Hollywood. She had never been to Hollywood. She had never been anywhere, because she does not exist. Tilly Norwood is a character generated entirely by AI, and the news that talent agencies were circling her to sign a representation deal set off a backlash that ran through actors, three national unions, and eventually into a real contract negotiation.
For a marketing team, it is tempting to file this under celebrity drama. That would be a mistake. The Tilly Norwood fight is the clearest preview yet of what happens when a brand puts a synthetic person in front of customers. The talent objections, the union response, the consent questions, and the new disclosure laws are not Hollywood's problem alone. They are the operating conditions for anyone considering an AI spokesperson, an AI brand ambassador, or an AI actor in an ad. This post is about what that fight actually decided, and what it means before you greenlight a face that was never born.
Where Tilly Norwood came from
Tilly Norwood is a product of Xicoia, the AI talent division of Particle6 Group, a London production company. Particle6 was founded in 2015 by Eline Van der Velden, a Dutch actress and producer. Xicoia itself was set up in February 2025, and Van der Velden has described it as the world's first AI talent studio.
The character was built quietly through the first half of 2025. An Instagram account went live on 6 May 2025. A comedy sketch called "AI Commissioner" appeared on 30 July 2025, made with around ten different AI tools and a script generated by ChatGPT. By early October the Instagram account had roughly 50,000 followers and the sketch had passed 700,000 views on YouTube.
None of that caused a firestorm. The firestorm came from a single trade story. On 27 September 2025, around a presentation at the Zurich Summit, Deadline published a piece reporting that talent agents were interested in representing Norwood. That framing, an AI character being courted like a working actor, is what detonated. Within days, performers including Emily Blunt, Natasha Lyonne, Mara Wilson, Toni Collette, and Sophie Turner had spoken out. Whoopi Goldberg weighed in. Some actors suggested boycotting any agency that signed her. Two major agencies, WME and Gersh, publicly said they would not.
Three actors' unions condemned the project: SAG-AFTRA in the United States, Equity in the United Kingdom, and ACTRA in Canada. SAG-AFTRA's statement was blunt. It said creativity should remain human-centered and that the union opposes replacing human performers with synthetics. The same statement made a sharper point that matters for brands: a synthetic performer has no life experience to draw from and no emotion to convey, and audiences are not asking to have human performers replaced.
Van der Velden pushed back. She framed Norwood as a creative work rather than a replacement for a person, compared building the character to a writer inventing a character, and said the project employed real human artists doing a different kind of craft. She also said she had no interest in copying any real individual, and that creating a digital twin of an actual person would require their consent and fair compensation. Particle6 has claimed that working with a synthetic performer could cut production costs by as much as 90 percent. By November 2025, Van der Velden was telling Deadline that 40 more AI characters were in the pipeline, an expanding roster the press began calling the Tillyverse.
Where the synthetic actor question stands now
Strip away the celebrity quotes and three concrete issues remain. Each one lands directly on any brand considering synthetic talent.
The first is the consent and likeness question, and it is the one Particle6 cannot fully wave away. Even granting that Norwood was not copied from one identifiable person, the model behind her learned from images and performances of real people. Mara Wilson put the objection in plain terms, asking what about the women whose faces were effectively composited into this one. A Scottish actress, Briony Monroe, said she believed her own likeness and mannerisms had been used and consulted Equity about it. This is the part of the debate that has actual legal teeth. The body of law here is the right of publicity, sometimes called name, image, and likeness, or NIL. It gives a person control over the commercial use of their identity. A purely synthetic character was supposed to sidestep it. But if a synthetic face is recognizably close to a real performer, or if it was trained on a specific person's work without permission, the right of publicity is back in play. Lawyers writing about the Tilly Norwood moment have called it a new frontier for NIL law precisely because it tests whether a composite can still misappropriate an identity.
That legal exposure is widening fast. Tennessee's ELVIS Act, in force since 1 July 2024, made it unlawful to use a person's voice or likeness without consent and even to distribute tools whose main purpose is generating unauthorized replicas. At the federal level, the NO FAKES Act was reintroduced in the Senate in 2025; if it passes, it would create a federal digital replication right for the first time. The direction of travel is unmistakable. A synthetic spokesperson built on murky training data is not a clever shortcut. It is an unbounded liability that grows as the law fills in.
The second issue is union economics, and this is where 2026 moved the story from outrage to policy. Through the spring, as SAG-AFTRA prepared to negotiate a new contract with the major studios, the union floated an idea the press named the Tilly tax. The mechanics: studios using a synthetic performer, whether a digital replica of a real actor or a fully AI-generated character, would pay a fee into union health and pension funds. The point of the fee was to erase the cost advantage, to make a synthetic performer cost roughly what a human one does and remove the financial reason to swap them. Even inside the union the idea was contested. One AI task force member, Brendan Bradley, called it the best bad idea on the table for 2026. A former technology committee co-chair, Erik Passoja, said no three times over, then conceded that if it had to exist the money should go to pension and health.
Negotiations opened on 9 February 2026, with the existing contract set to expire on 30 June. On 2 May 2026, SAG-AFTRA and the studios reached a tentative four-year agreement. A standalone Tilly tax line item did not survive as such, but the principle behind it did. The deal carries roughly a dozen AI provisions. A company now needs an articulable business reason to scan a performer at all. It must obtain consent to use an actor's digital replica, and there is a minimum payment rate plus residuals when a replica is used, including replicas the company did not create itself. Digital replicas cannot stand in for striking union members. And studios must commit to a principle strongly favoring human performances: a synthetic cannot take a role a human would otherwise play unless it brings significant additional value. The plain reading is that Hollywood, for now, has decided real performers still matter and has made the synthetic shortcut expensive and paperwork-heavy.
The third issue is reputation, and it is the one a marketer should weigh most heavily, because it does not wait for a lawsuit or a contract. The audience is already unhappy. Luminate research, cited by EMARKETER in its coverage of the Tilly tax debate, found that only about 22 percent of US film and TV viewers are comfortable with entirely synthetic actors while 48 percent are uncomfortable with them. The advertising-specific data is just as pointed. IAB and Sonata Insights research found that 82 percent of ad executives believe consumers view AI-generated ads positively, while only 45 percent of Gen Z and millennial consumers actually do, a perception gap that widened from 32 points in 2024 to 37 points in 2026. The brand graveyard of 2025 makes the abstract numbers concrete. McDonald's pulled an AI holiday ad in the Netherlands after viewers called it creepy and tone-deaf. Coca-Cola was mocked for glitchy AI-generated trucks. AI-generated models in Vogue and Guess print ads drew social backlash. A Valentino AI campaign was called cheap and tacky.
The reaction has been sharp enough that brands are now positioning against AI as a selling point. Aerie and Dove have pledged not to use AI in advertising. Equinox, Aerie, and Almond Breeze ran 2026 campaigns that openly mocked AI slop. Gartner has predicted that around 20 percent of brands will market themselves on the absence of AI, a kind of no-filter movement for the AI era. If a meaningful share of your competitors are about to advertise that they do not use synthetic people, putting one at the center of your own campaign is a strategic choice, not a neutral production decision.
Where this is heading, and what a brand should do
The synthetic actor question is not going to resolve into a simple yes or no. It is going to resolve into rules, and the rules are arriving now.
Disclosure is becoming law. New York signed a synthetic performer disclosure law on 11 December 2025, effective 9 June 2026. It requires advertisers to conspicuously disclose when a synthetic performer appears in an ad, where the synthetic performer is a digitally created asset that looks like a human performance but is not recognizable as any real performer. Penalties start at 1,000 dollars for a first violation and 5,000 dollars for each one after. New York is the first state to do this, and first is rarely last. The honest planning assumption is that within a couple of years, using a synthetic spokesperson will come with a mandatory on-screen label in major markets. The quiet substitution, where an AI face simply stands in and nobody is told, is closing as an option.
The cost case is also weaker than the brochure says. Synthetic talent looks cheap when the comparison is a clean AI render against a day rate plus residuals. But the real comparison now includes rights clearance on training data, likeness indemnification, disclosure compliance across states, and the reputational cost of a campaign that lands as slop. EMARKETER's read on the Tilly tax debate was direct: marketers should expect a more regulated, rights-heavy AI content world and should rethink the assumption that synthetic talent is simply cheaper. The cost advantage is real for some uses and largely imaginary for others once the full bill arrives.
That points to where synthetic performers will actually settle in brand work, which is not as a replacement for a star. The roles under genuine pressure are the templated ones: background presence, generic stock-style commercial talent, and product demonstration. A synthetic presenter reading the same localized script in 30 markets is a defensible use. A fully synthetic brand ambassador carrying your reputation, with a fabricated personality and an implied relationship to your customers, is a far riskier bet, both legally and reputationally.
If a brand still wants to move, a few rules separate a defensible synthetic talent project from a costly one. Know exactly what the model was trained on, and if a vendor cannot tell you, treat that as a liability rather than a detail. Avoid any synthetic face that resembles a recognizable real person, because resemblance is where right-of-publicity claims live. If you use digital replicas of real people, get explicit, specific, time-bound consent and pay fairly, which is now the contractual standard in Hollywood and a reasonable floor everywhere else. Disclose the synthetic performer plainly, ahead of the law rather than behind it. And run the reputational test before the production test: ask whether your audience would feel misled, because the cost of getting that wrong is not a fine, it is trust.
The deeper signal in the Tilly Norwood story is that synthetic talent is no longer a technical capability waiting for permission. It is a contested space with unions, lawmakers, lawyers, and audiences all actively setting terms. The brands that come out of this well will be the ones that treated a synthetic spokesperson as a serious decision with legal and reputational weight, not as a line item that happened to be cheaper than a person.
Council summary
This post argues that the Tilly Norwood controversy is not Hollywood gossip but a working preview of the legal, contractual, and reputational conditions any brand will face if it puts a synthetic person in front of customers. The council verified the core facts: Xicoia as Particle6's AI division founded February 2025, the Zurich Summit and Deadline story that triggered the backlash in late September 2025, the SAG-AFTRA tentative deal reached 2 May 2026 with roughly a dozen AI provisions, the ELVIS Act and the reintroduced NO FAKES Act, and New York's synthetic performer disclosure law effective 9 June 2026. Two corrections were made: the 22 percent and 48 percent viewer-comfort figures were reattributed to Luminate, the original source EMARKETER cited, and unverifiable specifics about the deal's contract language were rewritten to match what the union and trade press actually confirmed. The reader takeaway is concrete: treat synthetic talent as a real decision with rights clearance, disclosure duties, and trust at stake, reserve it for templated roles, and never build a brand face on training data a vendor cannot explain.
Comments