AI-Assisted Or AI-Generated?

There is a big difference between asking AI to create something for you and asking it to help you create something, yet increasingly we describe both in exactly the same way. Someone could give an AI a prompt and publish the resulting article almost untouched, while someone else could spend hours researching and writing their own piece before using AI to improve the structure, tidy a paragraph or challenge an argument.

Both might be described as having “used AI”, despite the human contribution being completely different.

AI Use Isn’t Binary

The language around AI often suggests a fairly simple choice: something was either created by a human or generated by AI. In practice, there is an enormous amount of activity sitting somewhere between the two, particularly as AI becomes embedded within tools people already use.

A photographer might use AI to remove something distracting from an image, while a designer could generate individual elements before assembling the finished work themselves. A writer might use it to brainstorm, edit or interrogate something they’ve already written, and someone preparing a presentation could use AI to turn their own research into a first draft of a slide. All of those involve AI, but describing the finished work as simply “AI-generated” doesn’t necessarily tell us very much about how it was created.

The difficult question is where assistance becomes generation, and there probably isn’t a universally useful line. The answer depends partly on what AI contributed and partly on what we expected the human to contribute in the first place.

What Did The Human Actually Do?

Looking at what the human actually contributed may be more useful than simply asking whether AI was involved. If the ideas, judgement and decisions came from a person, AI may have played a substantial role in producing the finished output without necessarily replacing the thing we valued about the human contribution. At the other end of the spectrum, someone can remain technically involved while contributing little beyond the original prompt and approving whatever appears, which is where the rather horrible term “meat proxy” starts to become relevant.

That’s why labels such as AI-generated can sometimes feel inadequate. They tell us that technology was involved without telling us whether it provided the thinking, the execution or simply some assistance along the way, and that distinction can matter considerably depending on what we’re evaluating.

We may care very differently about AI assistance in an internal email than we would in an academic essay, piece of journalism, artwork or job application. In some situations, all we really care about is whether the finished result is useful and accurate; in others, the human contribution is part of what we’re evaluating in the first place.

A Better Question

As AI becomes embedded within everyday tools, separating work neatly into “human” and “AI” categories is only going to become harder. We already accept spellcheck correcting our writing, cameras computationally improving photographs and design software automating decisions that once required considerably more manual work, while generative AI expands that assistance much further.

Rather than simply asking “Was AI used?”, the more useful question may be “What did AI actually do?”