AI Guides › Creative & Design
The Difference Between AI-Assisted And AI-Generated, And Why It Matters To Say Which
By Nigel Guy · 2 min read
Most people asked "did you use AI on this?" answer yes or no, as if it's one
question. It isn't. A logo you sketched, refined by hand, then had a tool
help you vectorise cleanly is a different thing from a logo a tool produced
end to end from a text prompt — and collapsing both into "I used AI" tells
the person asking almost nothing about what they're actually looking at.
The rule: say what AI touched and what a person touched, specifically —
"AI-assisted" and "AI-generated" aren't a spectrum of honesty, they're two
different descriptions of two different processes, and the vague version of
either one is the dishonest one.
The distinction, made concrete
- AI-generated means the tool produced the substance of the output —
the words, the image, the composition — from a prompt, with a person
selecting and perhaps lightly editing the result. The creative decisions
mostly happened inside the generation.
- AI-assisted means a person made the creative decisions — the
composition, the argument, the structure — and used a tool for part of
the execution: cleaning up a draft, generating a variation to react
against, handling a mechanical step like resizing or transcription.
- Mixed is real and common, and worth naming as mixed rather than
rounding to one label. "I generated the base image, then hand-painted
over sixty per cent of it" is a sentence, not a confession — say it.
The mechanism: label at the point of substance
Ask one question about the specific piece in front of you: if you removed
the AI step, would the core creative decision still exist, or would there be
nothing there? If the composition, the joke, the argument, the likeness
only exists because a model produced it — that's generated. If the
composition, the joke, the argument was already yours and the tool did
mechanical or editorial work on top — that's assisted. Label accordingly,
and do it per-piece, not once for your whole practice.
What to skip
Skip the instinct to pick whichever label sounds better for the situation —
"assisted" to sound more skilled, "generated" to sound more efficient — and
label based on what actually happened instead. And skip inventing a third,
softer term to avoid the question entirely; if a client, platform, or
competition asks directly, a vague non-answer reads worse than either
honest label.
Guardrails
- Different platforms and competitions define these terms differently, and
some define them narrowly enough to matter for eligibility — check the
specific rules of wherever the work is going, rather than assuming your
own definition is the one that counts.
- This distinction is about disclosure, not about quality — plenty of
AI-generated work is good, and plenty of AI-assisted work is bad. Neither
label is a quality claim.
- If you're not sure which label fits your own process, that's usually a
sign the process wasn't deliberate enough to describe cleanly yet — worth
fixing before the next piece, not just before the next answer.
All 751 AI guides · JulieMango plans from £17/mo