What “AI-integrated” actually means on a creator’s portfolio in 2026

A designer we know shipped a brand identity last spring. The client loved it, paid, and posted it everywhere. Then a teammate mentioned the early concepts were generated with AI, and the mood changed overnight. The work had not gotten worse. The trust had.

That gap, between what you made and what the client thought you made, is the real subject of portfolio AI disclosure in 2026. The deadlines and the platform rules are the easy part, and they are already covered: if you want the pre-pitch checklist, see our guide on what to update before you pitch H2 clients, and for the legal steps, the EU AI Act compliance checklist for creative teams. This piece is about the harder question underneath all of it: how do you show AI-integrated work so it reads as craft, not as a confession?

Start with the carve-out most guides skip

First, a reframe that takes the pressure off. The EU AI Act does not order you to stamp “made with AI” on every portfolio piece.

Read Article 50 closely and you will find a carve-out that most disclosure advice ignores. The deployer disclosure duty is aimed at realistic deepfakes of real people and at text published to inform the public on matters of public interest. On top of that, there is an explicit exemption for “evidently artistic, creative, satirical or fictional” work. A portfolio is mostly promotional, and most of what sits in it is creative work rather than a deepfake of a real person. So for ordinary AI-assisted design, illustration, code, or copy, the strict legal labeling requirement probably does not bite.

That matters because it changes your motive. You are not labeling your work to avoid a fine. You are labeling it to win a client who is quietly wondering how much of it was really you. This is general guidance rather than legal advice, and the rules read differently depending on where you and your client sit, so check your own jurisdiction. But the headline holds: the law asks for less than you fear, while clients ask for more.

How much more? Around 90% of consumers globally want to know whether an image was created using AI, according to Getty Images research. Meanwhile, 58% of creative professionals already use AI without telling the client, per Envato’s 2026 State of AI in Creative Work report. That space, high expectations meeting rare disclosure, is exactly where a clear label becomes a competitive edge rather than a liability.

A four-category framework for portfolio AI disclosure

Vague labels create the very suspicion you are trying to avoid. “Some AI was used here” tells a client nothing, so they assume the worst. A tiered label fixes that, because it says precisely how much of the work was yours. Use these four categories as a spectrum, not as rigid boxes.

Label	What it means	What a client reads into it
Human-made	Conceived and executed without generative AI. Ordinary tools like Figma or Photoshop do not count as AI here.	Premium craft, and the safe choice where authenticity or likeness matters.
AI-assisted (human-led)	You own the concept, direction, and final call. AI helped with drafts, variations, or speed.	“Fluent and in control.” Usually the trust sweet spot.
AI-generated	AI produced the core artifact. You curated and edited, but did not author it by hand.	Wants to know which elements, and how much you steered.
AI-curated (directed)	You art-directed: prompting, selecting, and refining AI outputs into one coherent result.	Values the editorial eye, but dislikes it when it is dressed up as hand-craft.

The point of four labels rather than one is honesty with range. A client who sees you distinguish “AI-assisted” from “AI-generated” reads you as someone in command of the tools, not someone hiding behind them. We notice this from both sides of the table, since we brief creators on behalf of clients too. The creator who names the category up front almost always reads as the safer hire.

Where the disclosure line actually sits

So when does a piece cross from “human-made” into label territory? Honestly, the line blurs, and reasonable people disagree on the exact point.

As a working trigger, minor tasks rarely need a label. Fixing grammar, nudging tone, or cleaning up a layout sits inside normal craft, the same way using a spell-checker does. The label earns its place when AI did something a client would want to know about: generating ideas, drafting whole sections, or producing core visual elements. When you are genuinely unsure, label up rather than down. Over-disclosing costs you almost nothing, while under-disclosing is the thing that detonates trust later.

How to narrate each label

Here is the part that separates a confident portfolio from a nervous one. A label tells a client what happened. Narration tells them why to trust you anyway, and it is where most portfolios go quiet. Each category needs a slightly different story, so match the narration to the label.

For AI-assisted work, lead with your judgment. Show the volume AI gave you, then the taste you applied:

“I generated 30+ layout directions with AI in an afternoon, then built the final system by hand in Figma. The speed let me test more ideas before committing. The structure is mine.”

For AI-generated work, be specific about which elements are AI and what you did to them. Vagueness here is what clients punish:

“The base imagery is AI-generated. I retouched every frame for consistency, fixed the hands and type, and color-graded the set to match the brand. The concept and final edit are mine.”

For AI-curated work, own the art direction openly. Do not dress prompting up as hand-craft, because clients can tell, and the mismatch costs more than the honesty would:

“I art-directed this series: writing the prompts, running hundreds of generations, and selecting and sequencing the final twelve. The editorial eye is the work here; the raw renders are AI.”

Notice the shared pattern. Each version names the AI role, names your decisions, and ends on the human judgment. One more move makes it land: place an early AI draft next to your finished piece. Seeing the gap between a raw generation and your polished result proves your taste far better than any disclaimer, because it shows the work instead of asserting it.

The label isn’t the trust. The narration is.

It helps to remember what a label actually does. By naming what is AI-made, you also protect everything that is not. That is the quiet advantage here. In a market where most creators stay silent, the one who explains their process with confidence reads as the safer choice, not the riskier one.

The deadline on 2 August 2026 will pass, and the legal questions will keep getting refined. Done well, portfolio AI disclosure is not a compliance chore but a positioning move. What will not change is the thing underneath all of it. Clients hire people they trust, and trust is built by saying the true thing before anyone has to ask.

The Blue Mango is a creative co-op that connects creators with client work. If you want your AI-integrated work in front of clients who value how it is made, work with us.