Contra’s 2026 Creative Industry Report: AI Makes Production Easier. How Do Creators Turn It into Value?

AI is already part of many creators’ work, but using a tool and earning more money are two different things. Some use it to expand their services; others see no change in income, or even face pressure on their jobs. Contra Labs’ Creative Intelligence Report 2026 brings these different outcomes together.

The report combines questionnaires, Contra platform project records, and creator interviews. The publisher says the survey covers more than 150,000 working creative professionals. The most useful questions to ask while reading it are: where does AI help, which work still requires human judgment, and how do these changes translate into outcomes clients are willing to pay for?

1. Creators are using AI. What do project records tell us?

Junior, mid-level, and senior creators report using AI at work at rates of 71.1%, 70.7%, and 69.8% respectively—all around seven in ten. AI use is not limited to experienced practitioners.

Contra’s paid project records show that the share of projects marked as using AI tools rose from around 7% at the start to around 19% at the end. AI-tagged projects in portfolios also reached around 17% at the end of the period.

The roughly seven-in-ten figure counts respondents; 19% counts paid platform projects. A creator may use AI on some projects and not on others. Dividing the two figures does not produce a “commercial conversion rate.” The next question is what work the tools actually participate in.

View the original portfolio project trend chart

2. Generation is easier; creative direction and judgment still need to come from you

When production costs fall, the most immediate change is seeing what an idea looks like sooner. Frank Bach at Anthropic says his UX process no longer always moves step by step from low to high fidelity. He can try a more complete proposal directly, or even use a generated podcast to communicate a brief. He feels this leaves more room for exploration, and imagines that the traditional longer process could eventually become a choice that values the process itself, much like woodworking by hand.

Generating more candidates does not mean you have found a better solution. Gui Seiz at Figma distinguishes “making things with AI” from “exploring things with AI.” Personal judgment, references, experiences, and constraints give exploration its direction. For him, doing less pixel pushing and writing less boilerplate code frees up time to think.

The way a tool operates also affects that exploration. Paper founder Stephen Haney says waiting for generation after entering a prompt can interrupt his thinking. He hopes future tools will let designers manipulate and guide results more directly, making judgments continuously as they explore.

Clarify the goal before asking a model for options

Worklife Ventures founder Brianne Kimmel criticizes a specific practice: entering prompts before writing a clear creative brief or thinking through the goal, then accepting the first one or two ideas the model offers. This can produce many similar proposals. She places greater value on the point of view a creator has formed before calling the model.

Daniel G Bright at ByBrightStudios says some clients initially thought they no longer needed a designer, then returned to professionals after seeing similar-looking results. He also uses models to build small tools that support design. Mark Vasilevskiy believes that as companies look increasingly alike, people will still pay for distinctive brands, websites, and motion. Both emphasize the same point: tools generate candidates, but what you select and how you revise it still affect the work’s appearance.

Designjoy founder Brett Williams calls AI an amplifier of taste. He uses ChatGPT for image exploration and Claude for mockups and developing ideas. His practice shows how tools participate in his work; it does not mean AI has solved aesthetic judgment.

How AI supports ideation, mockups, and checking

The report also measures respondents’ AI use across three stages: ideation, mockup, and refinement.

The figures are 87% for ideation, 69% for mockup, and 79% for refinement. They count people using AI at each stage, not how much work AI completes.

The report gives several examples: organizing competitor information from a client questionnaire; using Claude or ChatGPT to challenge positioning ideas and check a brand story; and comparing a finished advertising design with the original brief for missing requirements. AI provides candidates and review suggestions in these stages. The creator must still decide which suggestions to accept.

3. From completing a task to taking responsibility for a whole project

Another change is that creators can communicate ideas through unfamiliar media. Gui Seiz says that even without spending years learning 3D or motion, tools help him express what he imagines so people from different disciplines understand one another sooner. This expands the scope of expression and exploration; it does not mean someone has mastered another profession.

He also compares the change with the move from metal typesetting to desktop publishing: tools handle some operations for people, allowing them to focus more on what they want to express.

In team responsibilities, Paper places more value on people who can own a complete project. Stephen Haney says they need to check generated code and take responsibility for product decisions and whether the design suits users. Apple Product Director Jonathan Pimento uses AI to identify edge cases and unknown problems, and expects movement between design and code to change how roles are divided. Dann Petty at QuiverAI sees promise in designers building more independently.

Two interviewees go further in imagining future roles:

• Michael Riddering sees promise in the “Brand Engineer”: someone who can contribute from product strategy through implementation details and create a strong brand experience. He uses “five or six” and “ten” to describe ordinary output and distinctive experiences. This is his assessment of the hiring market; the scores are metaphors.
• Jamey Gannon imagines creative work consolidating around “product people” and “brand people.” The former would handle products people click and interact with; the latter would handle brand content people browse. Strategy, production, implementation, and testing would become more closely connected.

Business administration is entering tool-supported workflows too

AI is also used for business administration outside creative work. Mark Vasilevskiy says his team uses an internal Agent to track business details, reducing management interruptions for its main creators. He also reports that AI has expanded his business and helped increase revenue.

Justine Moore at a16z sees promise in similar uses. Small studios lack the support teams of large companies; research, outreach, invoices, and contracts all take time away from creative work. She hopes Agents can help them handle more administration. The useful lesson is which administrative tasks might benefit from tools, rather than treating future automation as already achieved.

4. How does being able to do more become a commercial return?

Jamey Gannon’s personal experience is that being able to take on tasks she could not do before also lets her charge higher fees. To understand whether that change is widespread, we need to look separately at project value and total personal earnings.

Contra tracked two groups of creators—AI adopters and non-adopters—and compared changes in their project value before and after:

The platform first calculates the change in each creator’s median project value, then takes the median of those changes within each group: +20% for adopters and +7% for non-adopters. This is an observational comparison; it alone does not establish that AI caused the gap.

In the questionnaire, surveyed creators report changes in their total earnings after adopting AI:

The income questionnaire shows a different distribution: 49.1% of AI adopters report higher total earnings, 37.0% unchanged earnings, and 13.9% a decline.

One chart tracks project value; the other asks about total earnings. They answer different questions. When the payment per project rises, the number of projects a creator can secure still matters. Evaluating whether a tool is worth continuing to use also requires accounting for subscription fees and rework time.

Comparing tools, earnings, and deliverables across disciplines

Different disciplines can be compared side by side:

View the complete original charts for all eight disciplines

For example, the rough median net paid for an “MVP / full product build” in web development is $2,122, compared with $331 for print and promotional collateral in graphic design. These involve different scopes of work. The payment gap should not be read as a premium obtainable simply by learning a tool. Looking at what you can deliver is more useful than choosing a profession from a payment ranking.

5. What new outcomes are clients starting to buy?

Clients on the platform still buy conventional creative services, while new types of AI deliverables are emerging. Putting demand changes beside project payment makes it easier to distinguish two commercial signals:

View the original chart of demand and payment by work type

Switching the order shows that AI automation, the fastest-growing demand category, does not have the highest project payment. Demand for RAG / Custom GPTs is 1.5 times the baseline, yet its rough median net paid per project reaches $3,960. Growth multipliers and project payments answer different questions.

(RAG retrieves sources before generating answers; Custom GPTs are assistants configured for particular tasks. The chart’s dollar amounts are rough median net paid per project, not profit or hourly earnings.)

Creators are also trying new services on the supply side. The questionnaire asks what they already offer or plan to offer. The three highest shares are AI Workflow Design (60.0%), AI Creative Direction (50.0%), and AI Landing Page Generation (49.8%). These options involve building workflows, setting direction, or completing a page; they are not subscription counts for a tool.

6. Why can opportunities and job pressure coexist?

Jitter founder Sébastien Robaszkiewicz sees promise in AI helping designers explore more directions, while believing that some execution tasks will shrink. He emphasizes judging what users need and choosing ideas worth pursuing.

Flora founder Weber Wong is more direct: he mentions creators already losing jobs and leaders hesitating over which positions to retain. He also predicts smaller creative organizations and sees opportunities for independent studios.

These views can coexist. Being able to do more may expand an individual’s business, and may also change a team’s existing division of labor. The report does not measure net job gains or losses across the whole industry, so some people’s stories of higher earnings cannot answer whether AI has eliminated creative jobs.

Cost, quality, and trust remain delivery challenges

Once tools participate in work, creators still face costs, output quality issues, and client requirements:

Cost is the concern selected most often (70.1%), followed by quality (44.6%), privacy and security (43.5%), and copyright (31.3%). Below, compare selections across experience levels and explore related anonymous feedback.

View the original charts of experience-level comparisons and anonymous feedback

This feedback makes the work remaining after AI use more concrete. Some respondents mention basic $20 subscriptions and too many tools. Others get no useful result from repeated prompts and still need to clean up generated content. Some worry about unreleased materials or encounter clients who do not want the final work to look AI-generated. These are different respondents’ experiences and concerns; they should not be combined into a situation shared by everyone.

Some also worry that relying on AI will weaken their skills. After reducing repetitive operations, creators still need to consider how they will practice selection, revision, and judgment in their work.

How to experiment and review the next real project

To apply the report to your next project, start with a few small steps:

1. Try a tool on one real task. Record which steps become faster, which require rework, and whether the client ultimately accepts the result. Look beyond the number of candidates generated.
2. Treat the model’s opinions as suggestions to assess. Let it organize competitor information, challenge positioning, or check a brief for omissions, then check the facts and finished work yourself.
3. Agree on the use of materials and tools in advance. Explain to the client where AI will be used and which unreleased materials may be processed by external tools.
4. Keep subscriptions based on actual use. Consider cost, usage frequency, and cleanup time together, reducing tools that are rarely used or repeatedly create rework.

These project-review suggestions draw on the practices in the report. Whether AI brings a return ultimately depends on what it helps you complete and whether the client values the result.

Supplement: Which tools creators use, and which tools clients mention

Different stages of work call for different tools. The report’s tool data is retained below for reference when choosing tools.

Portfolio tags, recent activity, and questionnaire adoption

ChatGPT is used by 29.3% of recently active platform creators, yet its adoption is 78.6% in the questionnaire’s LLM category. The populations and questions differ, so both can be true. The 13,165 figure in portfolio tags counts posts: one person can contribute several posts, so it is not a user count.

View the three original charts: post tags, active usage, and the LLM questionnaire

The interaction lets you compare each tool or switch between the three rankings. The complete figures are retained in the expandable section.

What does the tool-awareness chart compare?

The tool-awareness chart distinguishes reach, awareness, and adopter counts using Reach, Awareness, and Adopters. Bubble size represents the number of adopters. It does not provide a common denominator for calculating a conversion rate.

View questionnaire shares for six tool categories

The table follows the report’s tool categories. The category labels simply help readers find tools; they do not mean the questionnaire measured each tool’s effectiveness on specific tasks.

Category	Main tools in the questionnaire (self-reported share)
Text and language models	ChatGPT 78.6%, Claude 59.0%, Gemini 37.8%, Notion AI 18.9%, Perplexity 18.9%, DeepSeek 6.8%, Kimi 3.1%
Image generation	ChatGPT 38.4%, Midjourney 25.8%, Krea 17.2%, Flux 15.6%, Nano Banana 14.8%, Flora 6.3%
Video generation	Google Veo 27.3%, Kling 22.1%, Higgsfield 19.2%, Runway 17.6%, HeyGen 11.1%, Pika 6.5%, Opus Clip 5.6%
Audio and voice	Suno 43.1%, ElevenLabs 25.3%, Descript 5.0%, Renoise 2.9%
Coding and app building	Claude Code 61.5%, Cursor 27.0%, Lovable 25.4%, Replit 13.6%, v0 11.3%, Bolt 10.6%, Builder.io 7.7%
Native creative software	Figma 59.0%, Canva 55.3%, CapCut 46.1%, After Effects 36.7%, Framer 29.8%, Photoshop 20.7%, DaVinci 18.9%, Paper 12.4%

View quarterly changes in tool adoption

Comparing the second quarter of 2026 with the fourth quarter of 2025, platform creators’ usage shares reached these multiples of the baseline: Claude 32×, Higgsfield AI 14×, ElevenLabs 8×, and ChatGPT and Google AI Studio 6× each. Google Gemini, Kling AI, and Runway reached 5× each; Lovable, Figma Make, Cursor, and Leonardo AI 3× each; and Midjourney and Replit 2× each. Google Flow is marked New. These multipliers describe relative changes between two quarters, not absolute user counts.

Tool mentions in client requirements

The frequency of tool names in client project specifications is also changing. Comparing H1 2026 with H1 2025, Higgsfield reaches 25 times the baseline, Replit 10.8 times, Claude 9.9 times, and Cursor 9 times. They indicate which tools appear more often in client requirements, but cannot be used directly to compare tool performance.