Answer 32 of 50
How AI Tools Are Transforming Web Design Processes and Designer Roles
AI tools are already changing specific steps of the web design process: generating images and variations, drafting layouts from a prompt, and turning designs or descriptions into front-end code. They have not removed the steps where a person decides what the site should do, checks the output against brand, accessibility, and performance requirements, and answers for the rights in what is delivered. The designer’s role is shifting toward those steps, and the shift is uneven: the tasks tools now draft are production tasks.
Which steps of the process change, and with what
| Process step | What AI tools do today | Example tools and dates | What still needs a person |
|---|---|---|---|
| Visual assets | Generate images, extend or fill images, produce variations | Adobe Firefly, commercially available since September 2023, built into Photoshop and other Adobe apps | Brand fit, rights, factual accuracy of what is depicted |
| Layout exploration | Turn a text prompt or sketch into UI designs | Google Stitch, announced May 20, 2025 as an experiment from Google Labs; on May 19, 2026 Google added a real-time Stitch Agent for global users | Whether the layout serves the page's goal |
| Design to prototype or site | Generate prototypes and code inside the design tool | Figma Make and Figma Sites, introduced at Config on May 7, 2025 | Structure, content, accessibility |
| Code | Write front-end components from a description | Vercel's v0, which Vercel described in October 2023 as making website creation "as simple as describing your ideas" | Design system match, accessibility, performance |
| Alt text and copy | Draft descriptions and microcopy | General AI writing assistants | Meaning in context (see below) |
The sources: Adobe’s September 13, 2023 Firefly announcement, Google’s Stitch announcement, which describes turning “simple prompt and image inputs into complex UI designs and frontend code in minutes”, Figma’s Config 2025 recap, and Vercel’s v0 post, which says v0 “produces code using open-source tools like React, Tailwind CSS, and Shadcn UI”. A year after the first Stitch announcement, Google’s May 19, 2026 Stitch update introduced a Stitch Agent that streams its work to the canvas in real time and said the updates “are available to global users starting today”.
Firefly became commercially available in 2023 and the layout and code tools above were all announced by mid-2025, so “brief to concept” and “design to code” are products you can try now, not forecasts. What they share is that each output is a first draft.
Where human review is not optional
Alt text. AI can draft descriptions, but WCAG 2.2 success criterion 1.1.1, Non-text Content, asks for a text alternative “that serves the equivalent purpose” of the image. W3C’s own examples show the same photograph needing different alternatives in different contexts. A tool cannot know the purpose an image serves on your page, so every generated alt text needs a person to check it.
Brand consistency. A model does not remember what it made for you last month unless you anchor it to references. Keep an approved set of reference images, colors, and examples, and review every new asset against it.
Accessibility and performance of generated code. Generated components still have to meet your design system, keyboard and screen reader requirements, and performance budget. Treat them like code from a new contractor: review before merge.
Factual and cultural accuracy. Generated images can depict things that are wrong or inappropriate for a market. Someone who knows the subject and the audience has to look.
Rights in what you deliver
In the United States, the Copyright Office’s Part 2 report on AI, released January 29, 2025, concludes that “the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements.” That can include “a human makes creative arrangements or modifications of the output, but not the mere provision of prompts.” The Supreme Court denied review in Thaler v. Perlmutter on March 2, 2026 (docket 25-449), a case about a work claimed to be authored by an AI system.
For client work this means:
- Copyright can cover your human contribution, such as your selection, arrangement, and edits, not the AI-generated material on its own.
- Before promising a client exclusive ownership of every asset, tell them which deliverables contain AI-generated material.
- Check the commercial-use and indemnity terms of the specific model that produced each asset. Adobe said in 2023 that its first Firefly model “was trained on Adobe Stock images, openly licensed content and public domain content where copyright has expired”, and that Firefly for Enterprise offered an opportunity to obtain IP indemnification “for content generated by most Firefly-powered workflows.” Terms differ by tool and by model, and they change, so read the current terms for the model you used.
- Rules differ outside the US; if the work will be used in other countries, check there too.
This is general information, not legal advice; a contract that assigns rights in AI-assisted work is worth a lawyer’s review.
How roles are shifting
Without reliable data specific to design hiring, this page describes tasks rather than headcounts:
| Kind of work | Direction | Why |
|---|---|---|
| Production: resizing, variants, exports, first-pass layouts | Shrinking as manual work | These are the steps tools now draft |
| Review: checking generated output for brand, accessibility, rights | Growing | Every draft has to be checked by someone accountable |
| Direction: deciding what should exist and why | Unchanged or growing | Tools need a clear brief to produce anything useful |
| Client relationship and communication | Unchanged | Clients still need someone to explain trade-offs and take responsibility |
Where entry-level roles learn craft through production tasks, the change reaches them first. A junior designer who can direct tools well and review their output critically has a different value than one who only executes.
Small studios use the same public tools as large firms. Large firms can negotiate custom terms and integrate tools into their own systems; small studios compete on judgment, speed, and client relationships, which the tools do not supply.
What to do this quarter
- Pick one repeated production task and run it through an AI tool for a month, tracking time saved and corrections needed.
- Write down a review checklist for generated work: brand references, WCAG checks, performance, rights.
- Update your contracts and proposals to say how AI tools are used and what that means for ownership.
- Keep a record of your decisions on AI-assisted projects: which options you rejected and why. That record shows clients the judgment they are paying for.
How these roles may develop further, and how to plan for change without betting on predictions, is covered in what roles AI will play in the future of web design. Whether no-code builders with AI features can replace a professional site is covered in can no-code tools replace professional web design.
Use AI tools for drafts in the steps they handle well, put a named person on review for brand, accessibility, and rights, and tell clients plainly where AI-generated material is in what you deliver.