■ CRITICAL RISK ■ Creative Arts
For catalog and e-commerce work, it's already happening — brands are generating models who never existed and licensing digital twins of ones who do. Celebrity faces and runway presence resist longer, but the volume work that paid most working models' rent is evaporating.
“AI-generated faces don't age, complain, or need a makeup artist.”
Our AI replacement risk score — how we score jobs
Modeling looks glamorous from outside and logistical from inside: castings, fittings, hours under studio lights holding positions that look effortless and aren't, reshoots because the sample garment arrived late, and travel schedules that eat your life. The economics were always brutal — a thin layer of supermodels atop thousands of working models grinding out e-commerce shoots, where a brand might need the same hoodie photographed on six body types in fourteen colorways. That volume work is the industry's payroll backbone.
It's also the part AI eats first. Generative systems now produce photorealistic fashion imagery: brands like Mango and Levi's have publicly experimented with AI-generated or AI-augmented models, and services exist specifically to render garments on synthetic bodies of any size, age, and ethnicity — diversity without booking anyone. Digital twin arrangements let a model scan themselves once and license the likeness in perpetuity, which sounds like passive income until you notice one scan replaces hundreds of shoot days across the industry. Virtual influencers with millions of followers already book brand deals no human competed for.
What resists: physical presence. Runway shows need bodies in rooms. Celebrity and top-tier editorial modeling trades on a real person's fame and story, which a synthetic face doesn't have. Fit modeling — where garments are constructed on real bodies — retains a technical function. And there's a growing authenticity countercurrent, with some brands loudly advertising real, unretouched humans. But those niches employ a fraction of the people that catalog work does, and a 76 risk score is what it looks like when an industry's volume tier gets synthesized.
Automatability: our editorial assessment of current and near-term AI capability
Already biting. AI-generated fashion imagery moved from novelty to production tool within the last two years, and e-commerce — the volume end of modeling — is converting fastest because nobody emotionally bonds with a hoodie photo. Expect steep contraction in catalog bookings through the late 2020s. Runway, celebrity campaigns, and personality-driven work hold out well into the 2030s, but they were always the smallest slice of paid modeling.
Yes — this is not speculative. Major brands including Mango and Levi's have publicly used or piloted AI-generated model imagery, virtual influencers hold real brand contracts, and multiple services now exist purely to render clothing on synthetic bodies. It's concentrated in e-commerce and social content so far, which unfortunately is where most working models earned most of their income.
Partially, through contracts. Read every release for likeness-licensing language — some agreements now quietly grant rights to scan, replicate, or train on your image. Negotiate for per-use licensing fees on any digital twin, refuse perpetual buyouts, and involve your agency or a lawyer before signing scan agreements. Legislation on likeness rights is evolving, but for now your contract is your main defense.
As a sole income, it's riskier than it's ever been — and it was never safe. The realistic path now treats modeling as one component of a personal brand: build an audience, develop adjacent skills like content creation or styling, and target the physical-presence work AI can't do. The models who thrive will be small media businesses that happen to model, not faces waiting for bookings.
Anything where your specific, real identity is the product: celebrity campaigns, runway shows with live audiences, brand ambassador roles, and personality-driven editorial. Fit modeling also persists because garments must be constructed on actual bodies. The vulnerable work is anonymous-body imagery — catalog, e-commerce, stock — where the client needs a look, not a person, and a render is cheaper than a shoot day.