Brandtech’s gen AI business manager Amy Arnell explores the ‘in-between zone’ of AI humans, and where brands are considering using synthetic people in their campaigns.
Conversations around brands using AI at scale in their marketing began with a legal lens: “What can we do?” Once those boundaries became clearer, the questions shifted to an ethical lens: “What should we do?” And one question remains constant: “Do we want to generate AI humans in our ads?”
A thorny terrain: ethically, legally, and in the glaring spotlight of AI’s well-documented bias problem.
Still, while the topic is polarizing, the answers don’t necessarily have to be. Rather, there’s nuance: namely when to generate AI humans (and, more importantly, when not to), but also how to generate them.
Brands are now beginning to rethink initial stances and navigate ‘the in-between’. There’s no one-size-fits-all answer, but here are a few thought-starters for grappling with the gray.
“Is that really AI?” Visually, I can hardly tell the difference between AI and real. And that gap is only going to shrink. However, just because AI can do amazing things, that doesn’t mean it always should.
There are red lines that simply shouldn’t be crossed when it comes to AI humans.
For me, two examples stand out as fundamentally non-negotiable:
There’s no universal right answer. Ultimately, what matters is being intentional. Here’s how I see the in-between zone shaping up right now.
No humans, full stop: If you’re an activist brand with a strong POV on real representation, synthetic humans might feel like a non-starter. A valid line in the sand.
Or, flip the model: if you save time and money using AI for product shoots, backgrounds etc, could you reinvest those savings into shooting more human models? I love this approach.
Body parts, no faces: I promise in practice this is less sinister than it sounds. Synthetic hands pouring drinks, swiping nail polish, scrubbing surface cleaner: it’s subtle, it’s scalable, and lots of brands are leaning in.
But, here’s the catch: if the cleaning hand is always female and the manicured hand is always flawless and 22, we’ve got a problem.
That’s why we built Bias Breaker, an AI prompt improver tool introducing intersectional representation in AI images.
Groups, blurred faces: Crowds at parties, blurred people in the background – non-identifiable AI figures (again, there’s no way to phrase this in a way that doesn’t sound terrifying!) are gaining acceptance, even from brands that initially said ‘no humans‘. Once again, representation can’t be an afterthought, even in the blur – don’t sleepwalk into stereotypes.
Fully synthetic humans: Face-to-camera AI faces? To summarize, I’d say the verdict’s still out. For some brands, the decision comes down to where the creative appears (display banner v hero campaign), or what the product is. A beauty brand might tread more carefully than, say, a dog food or furniture brand. That feels logical and sensible. And in healthcare, an argument could be made that using AI over real patients or a person depicting a patient is more ethical.
AI, but make it equitable? With digital twins, real people can license their likeness: we train an AI model, and the talent can keep earning every time their likeness is used, with no additional shoot days required. We are working on this approach now with some clients.
Remember the bigger picture – in front of and behind the ‘camera’ (or AI) lens. However you use AI humans (or don’t), some principles aren’t up for debate. This goes beyond creative execution. It speaks to the fundamentals of who you are as a company.
Consider these points when planning campaigns:
If your brand hasn’t meaningfully invested in representation in the real world (in front of and behind the camera), AI won’t be your shortcut. In fact, it’ll be a magnifying glass. And, your customers will notice.
2. Inclusive AI teams make inclusive AI work. Think about who is prompting with AI. Make sure the people generating AI content reflect the communities being represented.
3. Opt for a human x tech approach. Humans are biased, AI is biased. The only way to lay the groundwork for inclusive AI content is by combining thoughtful human oversight with the right tech tools. One without the other won’t cut it. Advocate for internal training on inclusive prompting, and opt for technologies that prioritize and build for bias mitigation.
4. Consider your stance on watermarking. This would make for a full article in its own right. To try and summarize in two sentences: Transparency is king. Your consumers will thank you for it.
There’s no fixed destination here when it comes to how, when, or if at all to use AI humans in marketing – each brand will chart its own course, and that course may well change over time.
What matters is that we keep the ‘in-between’ conversation evolving, while ensuring that progress does not outpace perspective.
