AI could accelerate the gender revolution in marketing – if we get it right

This summer, I found myself wrestling with machine learning algorithms and data bias theory as part of an AI Business Transformation course at Saïd Business School, Oxford University. While my classmates came from varied backgrounds – tech entrepreneurs, consultants, seasoned executives – we were all united by the same humbling realisation: in the world of AI, everyone’s a beginner.

But it was precisely this fresh perspective that got me thinking about how artificial intelligence might reshape one of marketing’s most significant transformations: the ongoing gender revolution in advertising.

As I struggled through coursework on algorithmic bias (with considerably more enthusiasm than natural aptitude), a fascinating evolution in my thinking emerged.

Earlier this year, I wrote about the risks of AI undoing marketing’s progress on gender representation. But my deeper dive into the technology has revealed something more nuanced: while the same technology that risks entrenching gender stereotypes could indeed set us back, it could also be the catalyst for accelerating gender progress beyond anything we’ve achieved through traditional means.

How far we’ve come (and haven’t)

The evolution of women’s representation in advertising tells a story of progress punctuated by setbacks. We’ve moved from the patronising ‘Because you’re worth it’ era to Nike’s empowering ‘Dream Crazier’ campaign, celebrating female athletes who dared to break boundaries.

Dove’s ‘Real Beauty’ revolution challenged beauty standards, while brands like Always transformed ‘like a girl’ from an insult into a rallying cry.

Yet for every breakthrough, there are reminders of how entrenched gender stereotypes remain. Bumble’s 2024 ‘Anti-celibacy’ campaign, featuring billboards declaring ‘You know full well a vow of celibacy is not the answer’, sparked widespread outrage for undermining women’s bodily autonomy and choice. Even well-intentioned campaigns sometimes miss the mark – remember Bic’s tone-deaf ‘For Her’ pens that sparked worldwide mockery?

More insidiously, subtle biases persist. Women are still disproportionately shown in domestic settings, defined by relationships to others, or portrayed as primarily concerned with appearance. The Geena Davis Institute’s research consistently shows that even in progressive campaigns, women get less speaking time and are more likely to be depicted in passive roles.

Where AI could change everything

This is where artificial intelligence becomes genuinely exciting for gender representation. Imagine market research that can analyse sentiment and behaviour patterns across diverse female demographics that traditional focus groups miss entirely. Picture content creation processes that can identify and flag unconscious gender bias before campaigns launch.

AI could help us move beyond the “strong woman in a power suit” trope to something far more nuanced. Machine learning algorithms can analyse how different female audiences respond to messaging, imagery, and cultural references – revealing insights that might escape even the most gender-aware creative teams.

Consider the possibilities: AI-powered analysis of social listening data that captures authentic conversations among women across different communities, ages, and life stages – going far beyond the narrow demographic slices that traditional research typically captures. Predictive models that can test how diverse female audiences will respond to creative concepts before we even enter production. Even AI-generated imagery that helps creative teams visualise more inclusive representation during early concepting phases – though as I’ve written before, this requires extreme caution given current training data biases.

Most importantly, AI could help identify the subtle microaggressions and unconscious biases that human reviewers miss. An algorithm trained to spot gender stereotypes might catch that women in a campaign speak 30% less than men, or that female characters consistently appear in supporting rather than leading roles – the kind of systematic patterns that individually seem minor but collectively reinforce problematic narratives.

The critical caveat

But here’s where my beginner’s status in that Oxford classroom becomes crucial, and where my earlier warnings about AI’s risks remain absolutely valid. Every lesson hammered home the same truth: AI systems inherit the biases of their training data. If we feed algorithms decades of advertising that portrayed women as domestic goddesses or sexual objects, we’ll get more of the same – potentially at unprecedented scale.

We’ve already seen this play out. AI image generators consistently produce sexualised or stereotypical depictions of women. Recruitment algorithms have been caught discriminating against female candidates. The risk of amplifying gender bias through AI is real and immediate.

Getting it right

The brands that will lead the next phase of the gender revolution will be those that approach AI with both ambition and humility. They’ll invest in diverse teams to build and audit their AI systems. They’ll regularly test algorithms across different female demographics to ensure fair representation. Most crucially, they’ll use AI to augment human cultural understanding, not replace it.

The opportunity is enormous. AI could help us discover and authentically represent the full spectrum of women’s experiences – not just the narrow slice that traditional research captures. It could reveal cultural nuances and preferences that well-intentioned but homogeneous creative teams haven’t recognised.

But realising this potential requires constant vigilance. The most dangerous outcome would be surrendering the hard-won progress of the past decades to algorithms that perpetuate yesterday’s biases at tomorrow’s speed.

The future of gender representation in marketing will be built by combining AI’s analytical power with human wisdom about culture and equality. If we get it right, artificial intelligence could accelerate the gender revolution beyond anything we’ve achieved so far.