The AI tools landscape for designers moves fast enough that any list risks being outdated within months. But the underlying categories — research, ideation, layout, copy, code, client comms — are stable. What changes is which tools are actually worth using inside each one. This is our honest, opinionated assessment based on integrating these tools into live client projects over the past year.
This isn't a feature comparison. It's a practical guide to what actually earns a place in your workflow versus what sounds compelling in a Product Hunt launch and then never gets opened again.
Discovery & Research
Research is the phase where AI has made one of the most underappreciated improvements. Not in the depth of insight — that still requires human judgment — but in the speed of synthesis.
Perplexity is the tool we reach for first when starting a new project. Unlike raw search, it doesn't just surface links — it synthesises answers from multiple sources with citations you can actually check. For competitive landscape research, industry context, and audience understanding, it cuts the first pass of research time dramatically. The citations are important: it keeps you from building on hallucinated facts.
ChatGPT works well for open-ended research framing — generating hypotheses about what a market cares about, drafting research questions, or structuring a discovery session agenda. It doesn't replace primary research with actual humans, but it helps you go in better prepared.
Neither Perplexity nor ChatGPT should be your final source of truth on market facts or statistics. Use them to find the right questions and directions, then verify the specifics through primary sources. Errors in research compound throughout the rest of the project.
Visual Ideation
Image generation tools have matured significantly. The gap between what these tools produce and what a photographer or illustrator produces for specific commercial purposes is still real — but for the purposes of early-stage visual direction, they've become indispensable.
Midjourney remains the strongest option for photographic and painterly image generation. Version 6 improved photorealism significantly, and the ability to use style references means you can iterate on a visual direction consistently across multiple images. For client presentations and mood boards, it's changed the speed and quality of early visual communication completely.
DALL·E 3 (via ChatGPT) is more controllable in terms of following specific instructions, which makes it useful for generating custom diagrams, icons, and illustrative content. The trade-off is that it produces images that often look distinctly AI-generated in a way Midjourney has largely overcome.
Adobe Firefly is the commercially safe choice. All training data is licensed, which matters for clients who are sensitive to IP. Output quality lags behind Midjourney for photographic work, but for illustration and graphic elements it's genuinely strong — and the integration into Photoshop for generative fill is practically useful in real production work.
The best use of image generation tools in a studio workflow is not producing final assets. It's producing the visual vocabulary that anchors a shared understanding between designer and client before the real creative work begins.
Layout & Prototyping
This is the category with the most hype and the most nuance. AI layout tools have improved, but the gap between "AI-generated scaffold" and "finished design" remains significant.
Framer AI can generate a full landing page from a text prompt in under a minute. The output is structurally coherent — it understands the conventions of web layout, hero sections, feature grids, testimonials. What it produces is a starting point you'd be embarrassed to ship, but a genuinely useful starting point to react to and build from. The speed of getting to something tangible to critique is the real value here.
Figma AI is more useful for iterating on existing designs than generating from scratch. The rename layers function alone has saved real time. The "make component" features are quietly excellent for cleaning up messy design files. The generative features are less impressive than Framer's but the tool is integrated into the design environment most studios already live in.
Copy & Content
Copy generation is where the quality delta between tools is sharpest. Not all language models produce the same quality of web copy — the differences in tone, nuance, and strategic thinking are significant.
Claude is our preferred tool for almost all copy work. It writes with more nuance than ChatGPT, handles complex brand positioning more confidently, and is better at following specific tone guidelines when you give it examples. For hero copy, UX microcopy, long-form blog content, and anything where voice matters, it's the first tool we reach for.
ChatGPT is stronger for structured content — feature lists, FAQ generation, meta descriptions, structured data. It's also faster for high-volume output where nuance matters less than speed. For a/b testing copy variants across many options, it's more than sufficient.
- Claude for brand voice, strategic positioning, and nuanced writing
- ChatGPT for structured formats, bulk variants, and research synthesis
- Neither as a final editor — always read everything before it goes to a client
Code & Development
AI coding tools have changed the speed of front-end development more than any other category. If you're building in Framer, Webflow, or custom HTML/CSS/JS, the quality of AI assistance has reached a point where it genuinely accelerates delivery.
Cursor is the tool that most changed how we write code. It's an editor built around AI assistance — the difference from using a plugin in VS Code is significant. The ability to have a contextually aware conversation with the codebase, ask it to refactor sections, generate new components that match existing patterns, and debug errors in plain language reduces the time cost of custom code dramatically.
GitHub Copilot is the mature, stable choice. More conservative in its suggestions, better integrated into existing workflows. For studios that don't want to move their entire development setup, it adds meaningful productivity without disruption.
Client Communication
This category is less discussed but meaningfully impactful in a real studio workflow. The administrative and communication load of running client projects is significant — and AI has made a dent in it.
We use Claude regularly for drafting client feedback responses, particularly in situations where the feedback is emotionally charged or strategically complex. Having a first draft to react to is faster than composing from scratch, and the act of editing an AI draft often helps clarify your own thinking about what you actually want to say.
Meeting summaries, project status updates, and proposal drafts are all areas where AI assistance saves real time without sacrificing quality — provided you edit and verify the output rather than sending it directly.
How the Stack Works Together
The real leverage comes from how these tools connect within a project workflow, not from any single tool in isolation. A well-integrated AI stack for a typical web project might look like this:
- Perplexity for market and competitor research in discovery
- Claude for brief analysis, strategic positioning, and initial copy directions
- Midjourney for visual mood and direction in early client conversations
- Framer AI to scaffold initial layout structures for critique
- Claude again for refined page copy and UX writing during design
- Cursor for custom code components and interaction logic in build
- Claude for client communication drafts throughout
The tools that don't earn a place in this stack get cut. There are dozens of AI design tools launching every month, and most of them solve problems you don't actually have or solve real problems worse than existing tools do. Selectivity is part of the skill. A stack with seven well-integrated tools that you use fluently is worth more than a subscription to twenty things you half-remember exist.
The discipline of evaluating tools honestly — asking not "is this impressive?" but "does this save real time on real work?" — is what separates a studio with a useful AI workflow from one that's just collecting tool subscriptions.