top of page
Auditing the Bias Hiding in Your Brand's AI-Generated Imagery
The tradwife aesthetic has gone from niche subculture to global phenomenon — and it circulates a narrow, historically raced and classed ideal of the "good mother." My PhD research is now finding that some of this content isn't even human-made to begin with: it's AI-generated, trained on — and reproducing — exactly that same narrow ideal. This gallery is the working archive from that research: images generated from prompts like "tradwife," "yummy mummy," and "pregnant celebrity," first in MidJourney and now using Google's Gemini Pro and Nano Banana, run through Claude for a second reading and compared against my own. That's the same method I bring to brand work below.
The Problem
If you're generating brand imagery with AI, the model is drawing on the same training data, the same cultural defaults, that my research keeps finding bias in. A "family," a "professional," a "mother," a "leader" — AI has a narrow, statistically skewed idea of what each of these looks like. If nobody's checking, you're very likely reproducing it, whether or not you intended to.
My Method: The Refraction Method
I bring two decades in the image industry — including Getty Images and Amazon — together with doctoral training in visual discourse analysis. My approach centres on comparing what an AI system produces against the lived, diverse reality it claims to represent, and mapping exactly where — and why — the two diverge. That divergence is where the bias becomes visible.
What I Offer
AI Visual Bias Audit: A structured audit of your brand's AI-generated or AI-assisted visual content that identifies where imagery reinforces narrow or exclusionary representation. Deliverable: audit report, risk map, prioritized recommendations.
AI Ethics & Visual Strategy Advisory: Ongoing or project-based strategic input for teams building AI into their creative pipeline. Deliverable: strategy framework, internal playbook. Leadership Briefings For teams who need to understand the visual bias risk in AI-generated content — grounded in research, not alarmism.
Why Me
Twenty years photographing for Getty Images and Amazon. A track record of catching transformation early — from the newsroom's shift to digital in 1999, to running photography audits and training for SMB retailers across Scotland with proven ROI, to an ESRC-funded PhD studying algorithmic bias in real time. I've navigated three prior media disruptions. This is the fourth, and I'm already inside it.
Get In Touch
Ready to know what your AI visuals are really saying? Email Marsaili McGrath and Lets Talk.
bottom of page