Copying a living artist's style with an AI image generator is generally not illegal in itself — copyright protects specific works, not styles — but it can still get you banned from a platform, dropped by a client, or sued if your output is close enough to a particular image.
The real risks are not one thing but three: copyright (only if you reproduce protected expression), platform policy (rules that go further than the law), and reputational or contractual fallout (clients and audiences who don't care whether it's legal).
Most beginners worry about the wrong one. They ask "can I be sued for the style?" when the thing that actually ends their account is a terms-of-service violation, and the thing that loses them a commission is a client recognising a dead ringer for another illustrator's work.
Start with what copyright does and does not cover. A style — loose brushwork, a pastel palette, big-eyed characters — is an idea or a method, and copyright does not protect ideas. What it protects is a specific fixed expression: one painting, one character design, one comic panel.
So a prompt like "moody oil portrait, thick impasto, muted teal and rust palette" is imitating a method, and methods are fair game. A prompt that names a living artist and a specific piece, or that produces an image recognisably close to one of their works, is a different situation.
The line is not "did you name them" — it's "did the output reproduce protected expression." That's why the practical question isn't whether style imitation is legal, but how close your result lands to a specific work. Adobe's Firefly tools inside Photoshop, for example, are marketed around being trained on licensed and public-domain content, which changes the risk profile of what you generate — but it doesn't change what you do with the output afterward.
According to our AI tool database, Photoshop 2026 (v27.5) ships with Firefly AI Generative Fill and Firefly Boards, so a lot of beginners are already generating images inside a tool whose provenance story is more defensible than a scraped-from-the-web model.
Here's a concrete worked example. Say you want a book cover in the style of a well-known living illustrator. In Midjourney, you could write "cover art in the style of [artist name]" — and that's the version most likely to get you a takedown, a refund demand, or a public callout, because you've made the imitation the point.
A safer version uses the same visual ingredients without the name: "editorial book cover, gouache texture, limited palette of burnt orange and deep navy, flat shapes, heavy grain." Same vibe, no borrowed identity. If you want to steer the model visually rather than with words, Midjourney's style reference feature (--sref) lets you feed it an image and inherit its look.
That's the sharper risk: feeding it a living artist's actual painting pulls you much closer to reproducing protected expression than feeding it a generic texture, a public-domain painting, or your own work. According to our AI tool database, Midjourney V7 includes Omni Reference and Personalization v2 among its features, and both let you push output toward a reference — which means both raise the same question: what exactly are you referencing?
A style descriptor is a method. A specific artwork is expression. The tool doesn't know the difference; you have to.
Now the limits, because this is where most advice goes soft. First, platform rules are stricter than copyright and they change often. Some services restrict prompts naming living artists; others don't enforce it until someone complains.
Pricing and policy details shift constantly, so the vendor's own page is the only reliable source — don't trust a blog post from last year. Second, "legal" is not "safe." You can win a copyright argument and still lose the client, the publisher, or the audience.
Book covers, game art, and ad campaigns are commissioned with contracts that often include originality warranties — meaning you've promised the work isn't infringing, and that promise can be breached even if a court would side with you. Third, this advice does not cover trademarks or character designs, which are protected differently and more aggressively; a recognisable mascot or logo is a much bigger problem than a brushstroke.
Fourth, if you're imitating a dead artist whose work is still in copyright, the estate can and does act. The honest summary: generic style descriptors are low-risk, naming a living artist is high-risk, and referencing a specific artwork is the highest-risk move of all. If you want the look without the exposure, describe the ingredients — medium, palette, texture, composition — and keep your references generic, public-domain, or your own.
That's not a legal guarantee. It's just the version that keeps you out of the fights most people actually get into.