I've been curious to know what new roles are being created in the generative AI film space. Are they net new roles? Are they a combination of roles into one? Or is there new labelling and nomenclature?
So I went looking.
I pulled thirty months of them. Over 500 job postings scraped off the public web between 2024 and 2026, across more than 400 different employers. I sorted every one by what the title claimed and what the description actually asked for, then read the whole pile.
One number really stood out.
Prompt writing appears in one posting out of five. And where it does appear, it is usually a single line buried in a much longer list of duties.
Prompting is assumed, not hired for.
The webmaster problem, again
Fifty-seven percent of these titles do the same lazy thing. They bolt “AI” onto a craft noun that already existed.
AI Artist. AI Filmmaker. AI Animator. AI Video Editor.
The interesting information isn't the prefix. It's the noun that follows.
A handful of genuinely new names have shown up too. Creative Technologist, Agentic Workflow Specialist, AI Worldbuilder, GenAI Wizard.
GenAI Wizard is apparently a job now.
But not one of these new titles has really caught on. The market isn't inventing new professions so much as re-labelling old ones and hoping the label survives contact with a recruiter's search box.
We've seen this movie before. Around 2001, there was a job called webmaster, and one person did it all. The HTML, the server, the graphics, the copy, and the thing your uncle called “the internet.”
Nobody killed the webmaster.
The work just got big enough to need names. Inside a few years, it had split into front-end, back-end, information architecture, SEO, DevOps, and a dozen more specialties. The title didn't die of embarrassment. It died of scale.
“AI Artist” is the webmaster of 2026.
And it's already coming apart.
The operator is being replaced by the author
You can watch it happening in the numbers. The generic AI Artist bucket was 36% of postings in 2025. In 2026 it's 25%.
Meanwhile, Filmmaker and Director roles more than doubled their share, from 7.7% to 16.1%. Producing rose alongside them. Team-leadership language like mentoring, managing and running a crew nearly tripled, from 6% of listings to 16%.
The language underneath the titles surprised me. Film-grammar words like shot, framing, coverage, lens and composition went from 17% of listings to 41% in a single year.
And traditional post-production tools get named more often than any generative model. Premiere, After Effects and Resolve show up in 27% of the pile.
The market started by hiring people who could drive the tools. It's now hiring people who can be held responsible for the film.
Driving the tools is table stakes now. Somebody still has to answer for whether the thing is any good.
The one word that stood out the most for me
Here's the pain point that keeps surfacing in these postings.
Consistency.
Character and visual consistency is named in 23% of all postings, and it climbed from 21% to 35% year over year. It is, by some distance, the thing employers write into a job spec because they've already been burned by it.
Anyone can get a nice shot. They're paying for the same character across forty shots. Same jawline. Same coat. Same light.
And they're paying for that on the Thursday afternoon the client rewrites the third act and eleven of those shots have to come back by Monday.
The difference between a talented home cook and a restaurant isn't the best single plate. It's the four hundredth identical plate on a Saturday night, at nine o'clock, with two people out sick.
Consistency is a workflow problem, not a prompting problem. Which is exactly why it separates a studio with real pipeline discipline from a very good freelancer with a Kling subscription.
The market is renting, not hiring
This is a freelance market wearing a staff-job costume. Fifty-one percent of listings are contract, freelance or per-project. Twenty-eight percent are salaried full-time. And 382 of the 434 employers posted exactly once.
The exceptions tell you where companies have conviction. Engineering and ML roles are 92% full-time. Design and graphics, 68%. Where the work is genuinely infrastructural, companies hire staff. Where it's generative content, they buy it by the project and move on.
Only 57 of the 531 listings disclose pay, so I wouldn't build a business plan around the salary data.
The salaried roles cluster in a $99K to $176K band, with a median of $120K. Hourly contract work sits at a median of $37.50 an hour, in a range of $20 to $65.
Run the hourly number out at full utilization, and you get about $78K a year.
Against a $120K median salary with benefits attached, the same skill is worth substantially more inside a company than outside it, while the freelancer carries the risk of everything between the projects.
Reading 500 job posts so you don't have to is one dispatch of many. The Signal ships what we're learning — monthly, short, no filler.
Tool fluency depreciates in months
The stack is churning so fast it's almost funny.
Kling went from 13% of tool mentions to 30% in a year. Seedance went from 2% to 19%. Nano Banana and Higgsfield went from statistically nothing to double digits.
And in the same twelve months, Premiere grew.
Resolve and Photoshop grew.
I think that's the part people should pay attention to. Every hot model on that list will be replaced. Probably by something that doesn't exist yet as I write this.
The talent underneath it won't be.
Coverage. Pacing. Performance. Composition. Knowing what to do when the client changes their mind. The discipline to deliver on a date.
Those things have survived every format change the industry has thrown at them since sound.
A studio that sells consistency, coverage and delivery is selling something else entirely.
It's selling the thing over 500 employers keep asking for and mostly can't find.
That's the bet we've made at Power Shifter Studios, and this data is the closest thing to a second opinion I've been able to get.
The models are going to change.
The job is becoming much clearer.
If you're building a generative capability inside your own company and trying to work out which of these roles you actually need to hire, I'm happy to compare notes.
---
Method. 500+ job listings collected from publicly posted roles between 2024 and 2026, including archived posts. We assigned role families by keyword precedence in the title, falling back to the description. Responsibility and tool figures are keyword prevalence across full description text, so they undercount employers who describe a duty without using its common name.
Caveats. This is the AI-filmmaking talent market, not the AI job market. One slice, one community. A small number of large employers account for outsized blocks of postings, which pulls the full-time and salaried figures upward.