Should you disclose AI in your design portfolio?

What did the Pope's AI-art warning change for working designers?

Pope Leo XIV's criticism of AI-generated art, delivered to artists and cultural leaders at the Vatican on 2 October 2026, changed no hiring policy anywhere — but it named the suspicion creative candidates now walk into. He argued there's an ontological difference, prior even to an aesthetic one, between art and what a machine produces through statistical calculation based on millions of images created by others, and said algorithms lack the spark of humanity. He urged the Church to renew its alliance with artists, and warned that a culture of instant answers drifts toward sameness. Whatever you make of the theology, one phrase matters for your job search: it has become urgent to distinguish human work from machine output.

That question has already landed in hiring. Creative teams are reviewing submissions they suspect were generated rather than made — in this year's hiring surveys, close to three in ten managers say they've received AI-generated portfolio pieces or creative work, and a clear majority report AI-written CVs and cover letters. So your reviewer reads your case study with an extra question running underneath: did you actually do this? Silence doesn't register as modesty. It reads as a topic you'd rather avoid, and people fill that gap with their least generous guess. The opposite extreme costs you too: a portfolio that trumpets its tooling and never shows a single decision looks like a prompt library with a name attached.

The practical answer sits between those poles, and it's the same problem HRLens solves from the recruiter's side — how your wording and your claims read to a human reviewer and to an automated screen within the same thirty seconds. You're being assessed on provenance now, in your images and in your bullet points. That means deciding, deliberately, what you say about your tools, where you say it, and which piece of work carries the burden of proving you can still originate something. The rest of this piece is that decision, broken into the four places it shows up: contracts, case studies, your CV, and the interview.

How much of your AI use do you actually have to disclose?

Disclose at the process level: which tools, at which stage, on which deliverables, and what you decided yourself. You don't owe anyone your prompt history, and you shouldn't stamp "AI-assisted" on every thumbnail. Three forces make silence genuinely risky rather than merely awkward. Client contracts increasingly carry an AI clause. Copyright registration in the US now asks you to account for machine-made material. And the EU's transparency obligations for AI-generated content apply from 2 August 2026, which pushes disclosure duties down the chain to whoever actually made the asset — meaning the freelancer or the in-house designer. Vagueness used to be free. It now has a legal shape, and hiring teams in regulated sectors know it.

Those contract clauses come in three recognisable shapes, and knowing which one you're under tells you exactly what to put in a case study. A full prohibition bans AI anywhere in the making. An output prohibition — the common form in publishing and legal work — lets you use AI for research and rough drafts but requires the delivered work to be substantively human. A disclosure obligation permits AI and demands a written note naming the tool and the portions it touched. Illustration agencies have gone further, adding clauses requiring the illustrator's prior consent before a client uses technology capable of generating work in their style. Show a piece with no provenance and a cautious client sees risk.

Copyright gives you the second reason to be specific. The US Copyright Office holds that prompting alone — even elaborate prompting — doesn't make you an author, because the model determines the expressive elements. Where a work mixes human and machine contributions, only your contributions are protectable, and more than a trivial amount of AI material has to be disclosed on registration and disclaimed in the claim itself. Failing to disclose risks cancellation of the registration. So "I generated this and shipped it" isn't a flex in a portfolio review; it's a story about a deliverable the client may not fully own. "I generated a base, then redrew, recomposed and colour-corrected it" is a different conversation entirely.

Where should the disclosure live inside a case study?

Put it in the process section of the case study, as one plain line, and never as a disclaimer in the footer. One sentence does the whole job: the tool, the stage it entered, and the part that's unambiguously yours. "Firefly for texture passes and background variants; all type, layout and final colour set by hand" tells a reviewer more than three paragraphs of philosophy about creativity. The line that fails is the vague one — "AI-enhanced workflow" — because it hides the only thing anyone wants to know, which is where your judgement entered. Write it the way you'd answer a client who asked you directly, because sooner or later one will.

Then show the decisions, because that's the part a model can't hand you. Name what the tool proposed first, what you rejected and why, what you fixed locally versus regenerated from scratch, and what finally shipped alongside the number it moved. Include one honest failure: identity drift across a set, a palette that carried the wrong association in the client's market, type that fell apart at small sizes. A case study with a diagnosed miss beats a flawless gallery, because reviewers have learned that flawless galleries are now cheap to produce. The specificity is the credential — three days of ideation cut to half a day is a claim; explaining what you threw away is proof.

Here's the opinion I'll defend: keep at least one fully human piece among your first three case studies, and don't label every project with an AI badge. Reviewers rarely reach piece seven. If everything above the fold is model-assisted, the implicit message is that your taste runs downstream of a tool, however carefully you describe your edits. One project you drew, shot, wrote or art-directed end to end settles that question before it's asked. And blanket badging has a second cost — it invites a conversation about tooling when you'd rather be discussing the business problem you solved and the craft decision you'd defend in a room.

How do you word AI tools on a designer's CV without sounding replaceable?

List AI tools in your tools line, beside your craft software, and nowhere else. They belong in the same breath as Figma, Illustrator, After Effects or InDesign — because that's what they are to you, instruments you direct. What sinks candidates is promoting the tool into a job title ("AI-first designer") or into an achievement of its own ("used generative AI to create assets"). Both tell a hiring manager your contribution was the typing. A tools line like "Figma, Illustrator, After Effects, Firefly, Midjourney; motion, type systems, brand guidelines" reads as range. The same tools listed as skills in their own section read as a substitute for the skills underneath them.

Then rewrite the bullets so the tool is the verb's object, never its subject. Weak: "Used Midjourney for concept exploration." Strong: "Cut concept rounds from three days to half a day by directing Firefly variations, then redrawing final assets — bought a second round of stakeholder feedback before launch." Same workflow, entirely different candidate. One more structural point creatives get wrong constantly: your beautifully designed portfolio PDF is not your CV. Two-column layouts, icons and text boxes break parsing, so build the text document separately in a single column — a chat-based CV builder will produce the plain, parseable version while your portfolio stays as designed as you like.

There's a second trap worth naming. The CV you wrote with an assistant's help can itself read as machine-made — the giveaways are predictable, and reviewers in creative hiring are now primed to spot them. Phrases with no specifics attached, suspiciously even bullet lengths, adjectives doing work that numbers should do, and the same three verbs recycled through every role. If you've drafted or tidied your CV with ChatGPT or Claude, check whether it still reads as AI-written before you send it. It's a strange irony: the document arguing you bring human judgement to creative work is often the one that most obviously wasn't written by a human.

What will interviewers ask about AI, and how do you answer?

Expect two questions. The first is operational — "how do you use AI in your workflow?" — now close to standard in design interviews across industries. The second is a values probe: "we're moving from in-house design toward more AI-generated assets, how do you feel about that?" Answer both with a boundary map rather than enthusiasm or refusal. Say where AI speeds you up (variants, resizing, background work, early ideation), where you stay fully hands-on (concept direction, brand voice, final aesthetic judgement), and where it changes the client conversation (licensing, disclosure, managing expectations). Candidates who answer "I love AI" and candidates who answer "I don't touch it" lose the same point for the same reason: no judgement shown.

Structure each example like a brief: the business problem, the tool and why that one, what you did with the output, and the measurable result. Interviewers dig in with two follow-ups that catch out anyone performing fluency — "which prompt did you actually use?" and "tell me about a time it got it wrong." Have a real miss ready, with the diagnosis attached: the model gave you six near-identical compositions, so you blocked out thumbnails by hand and used it only for texture; or it drifted off the brand's illustration rules across a set, so you redrew the characters and kept the backgrounds. That answer reliably outperforms a polished success story.

Be equally clear about your limits, because that's what senior reviewers are listening for. You don't push client-confidential assets into public models. You check the contract before anything AI-touched goes into a deliverable. You don't claim ownership of material you can't actually assign. Say those out loud and the AI question stops being a threat to your candidacy and becomes evidence of professionalism. Before you apply, it's worth seeing how all of this reads from the other side: a free CV analysis scores your CV out of 100, rates ATS compatibility, impact and ownership, clarity and structure, and flags what your layout does to a parser. Paste the job ad and it lists the skills you're missing.

Frequently asked questions