AI for graphic designers: 5 tools with prices, 8 prompts, and the copyright line a client contract hits

Which AI tool does each design job, and what does it cost?

Image generation is where designers pay for AI, and every vendor counts it differently. We read each vendor's own pricing page on 25 October 2026. The prices below are in US dollars:

Privacy matters for client work because an NDA covers the concepts too. A logo concept for a brand that hasn't launched can't sit in a public gallery. On Midjourney's Basic and Standard plans, it isn't hidden. Midjourney also says a company making more than $1,000,000 a year in gross revenue must buy Pro or Mega, so a studio that size pays $60 a month from day one.

Canva is missing from the table on purpose. Its Magic Studio tools sit inside the editor, but Canva's pricing page wouldn't show us a US price on 25 October 2026, and we don't print a price we couldn't read. Check its plan page before you count on its AI for client work.

Who owns an AI image you hand to a client?

In the US, nobody owns the parts a model made from your prompt. The Copyright Office's report of 29 January 2025 put it this way: "prompts alone do not provide sufficient human control to make users of an AI system the authors of the output." That matters because most design contracts assign copyright to the client. If a logo came straight out of an image model, there's little or no copyright in it to assign.

Your own work on top of it is a different matter. The same report says a person may "select or arrange AI-generated material in a sufficiently creative way" that the whole work is original, and that edits which add new authorship are protected too. The protection covers what you added, not the generated pixels underneath. Keep the layered file. Keep the sketches too. Then say in the contract which parts a model made. A lawyer should word that clause; this page doesn't replace one.

Training data is the second question clients ask. Adobe says it trained its first Firefly model "on licensed images from Adobe Stock and public domain content", and that outputs from features without the beta label can be used commercially. Read that line closely. Firefly plans now include partner models from Google, OpenAI and Flux, and Adobe's training statement is about its own model, so pick the Firefly model by name when a client asks.

Prompts for the brief, the concepts and the copy

A chatbot is the cheapest AI a designer can use, and the words around a project are where it saves the most time. ChatGPT, Claude and Gemini all handle these on a free plan. Paste the client's own words, then make the model mark what's missing:

Here is a client's email about a new project: [paste]. Turn it into a one page creative brief: the goal, the audience, the deliverables with sizes and formats, the deadline, the brand rules they mentioned, and what success looks like to them. Use only what the email says. Write [ASK] wherever a brief needs an answer the email doesn't give, then list those questions at the end, most important first.

I'm designing [deliverable] for [brand], whose audience is [audience] and whose brand words are [three words]. Give me three concept directions that differ in idea, not just in color. For each: a one line idea, the visual approach in two sentences (type, image style, layout), a palette described in words, and the one risk of that direction. Don't suggest any font or image you can't describe.

Write 10 headline options for [deliverable] under [number] characters each, in the voice of [brand: paste two lines of their existing copy]. Mix three lengths: short, medium and the full character limit. No puns, no exclamation marks, no claims I haven't given you. Then pick the three that would work best at small size on a phone, and say why.

The character limit in the last prompt is the designer's rule, not the copywriter's. A headline that fits the layout saves a round of feedback.

Prompts for image models: moodboards, extensions and textures

Use image models for moodboards, backgrounds and fills, and keep the type and the logo in your own file. Exact brand colors and a client's real product belong there too. A generated layer is a starting point you paint over.

Moodboard image for [project]: [subject], [mood in three words], [lighting], [color mood described in words], photographed in the style of [editorial, product, documentary]. No text, no logos, no people's faces. Plain background area on the [left or right] third for type.

(Generative Fill in Photoshop, with the empty canvas area selected) Extend the background to match the existing scene: same lighting direction, same depth of field, no new objects, no text.

Tileable texture of [material, for example recycled kraft paper with fine fibers] that repeats with no visible edges, flat even lighting, no shadows, shot straight on.

The "plain background area" line is the one designers forget. An image made for a layout needs room for the type before it's generated, because cropping it in later loses the composition.

Prompts for critique and client feedback

A model can look at a draft and give you a second pair of eyes before the client does. Upload a flattened PNG or JPG of the draft (not the working file), then ask for a critique against the brief, not a redesign:

Attached is a draft of [deliverable]. The brief: [paste the goal and audience]. Critique it as a senior designer would, in this order: hierarchy (what the eye sees first, second, third), legibility at the final size, contrast, alignment and spacing, and whether it answers the brief. Give each point as a problem and one specific fix. Don't redesign it and don't comment on taste.

My client sent this feedback: [paste, for example "make it pop more, it feels a bit flat"]. Turn it into a list of concrete design changes I could test (contrast, scale, color, spacing, imagery), say which one most likely matches what they mean, and write me two short questions that would pin it down without sounding defensive.

The second prompt turns "make it pop" into a question the client can answer. It's the most common note designers get, and a guess costs a whole round of revisions.

Will AI replace graphic designers?

The US government expects fewer graphic design jobs, not none. The Bureau of Labor Statistics counts about 253,100 graphic designer jobs in 2025, with a median pay of $62,960 in May 2025, and projects employment to fall 2% from 2025 to 2035. It names AI directly: "Automated design tools, such as artificial intelligence (AI), are projected to make graphic designers more productive and reduce the need for these workers."

The same outlook still projects about 16,000 openings a year over the decade, because designers who leave the field need replacing. The prompts on this page keep four things away from the model: the idea, the type, the logo, and the call on whether the work answers the brief.