TL;DR: AI is making it dramatically easier for marketers to generate audience insights. But generic AI models can give marketers a polished, inherited version of an audience based on information that is incomplete, biased and shaped by what has already been published online. This is particularly problematic when using AI to understand Gen Z, where online commentary can differ significantly from how they actually think and behave. The answer isn't to stop using AI. It's to combine Gen Z and AI with real, relevant human data.
AI is now one of the first places marketers turn to when deciding who to target and how to reach them.
That's especially true when it comes to winning the attention and loyalty of newer generations like Gen Z, whose global spending power is forecast to quadruple by 2030.
The problem is that, as marketers increasingly use AI to understand Gen Z, they risk falling into a trap that makes marketing to this audience worse, not better.
AI is making audience analysis so much easier
Not long ago, understanding an audience's needs and wants could mean waiting months for bespoke research, stitching together disparate internal datasets, sifting through social media noise or paying for market reports that could be outdated by the time you opened them.
AI, especially through large language models (LLMs) such as ChatGPT and Claude, has shortened the path to audience insight.
Open any chatbot, ask it who Gen Z are and what they care about, and within milliseconds you can receive a confident response. Nearly half (44%) of marketers and advertisers did that last year for their audience targeting decisions, with only personal intuition and experience (51%) used more. AI use was also level with internal customer data (44%), ahead of custom research (31%), social media listening (32%) and one-off paid reports (22%).
But look closely at those quick, polished answers. Are they actually useful?
Within AI's answer to "Who is Gen Z?", you'll likely spot some familiar buzzwords and clichés: "they want authenticity," "they're socially conscious," "they're digitally native," and so on.
But what's often missing is the nuance, the surprising detail or the quantifiable evidence that might genuinely change your decisions for the better.
When recommendations are made based on those observations, they can also be very predictable. Gen Z values authenticity, so lead with values rather than features. Use UGC and build community. None of this is necessarily wrong. But it's hardly transformative for a marketer looking for the specific audience "aha" details that can elevate campaign resonance and performance to achieve cut-through.
And when you try to push the chatbots further to uncover more unique insights or interpretations, it's not unusual to see them take huge leaps from "what the information says" to "what it means," with numerous assumptions filling the gaps.
Ultimately, AI's ability to help you understand your audience is only as strong as the data it runs on, and in many cases, it's the Wild West, also known as the internet.
AI gives you the internet's version of Gen Z
Depending on the system, AI chatbots can draw on information learned during training, connected databases and, increasingly, live information retrieved from the internet.
The internet matters because it allows AI to go beyond what it learned during training and find more recent, specific information in response to your question. Ask it about Gen Z, for example, and it can pull together articles, reports, research summaries, social commentary and other publicly available sources to build an answer around what is already out there.
That's useful. But because the internet isn't neutral or a representative sample of your target audience, neither will the insights AI gives you about your audience.
Instead, it's shaped by what gets published, indexed and made accessible, which means some perspectives, markets, languages and sources are far more visible than others. Valuable academic or proprietary research may sit behind paywalls or in less-indexed sources, while widely published content can be easier for AI to find regardless of how strong the underlying evidence is.
And although some AI chatbot outputs now include sources, you still need to ask what sits behind the answer. Is the information credible? Is it current? Is it based on actual consumer data, or on the accumulated commentary and assumptions that already exist online?
Because when you ask an AI chatbot to describe Gen Z or any other generation, it isn't interviewing them. It's giving you an inherited version of that audience, shaped by what the internet has already said about them.
What's missing is the human data to evade the assumptions
We tested this ourselves. We asked widely used AI tools to describe Gen Z investors as a target audience, an audience that has already received plenty of column inches and media commentary about how they are influencing online finance communities and investment culture.
The answers across platforms consisted of some version of: digitally native, predominantly male, mobile-first, risk-tolerant, crypto-aware, socially influenced, values-led and hard to reach through conventional advertising.
But when we connected those same questions to GWI's Agent Spark platform, the picture changed. Instead of relying on what had already been said about Gen Z investors, the chatbot could draw on fresh consumer survey data from more than 2 million interviews a year across 53 markets, with data refreshed each quarter.
The results revealed far more intriguing traits and observations about Gen Z, including how they describe themselves and their lives.
Two-fifths of Gen Z investors are female, showing they are far more gender-balanced than common assumptions suggest, and they are 33% more likely than the average person to describe themselves as creative.
Just over half are arts and culture enthusiasts, 60% more likely than the average Gen Z consumer. A quarter watched a live gaming stream last month, making them 91% more likely than the average Gen Z consumer to have done so.
The richer picture you get from authentic human data can change what you actually do with the audience. It points you to different creatives, different partnerships, different channels and, ultimately, different places to put your budget.
Blurry audiences are expensive
Advertising is only one part of the marketer's toolkit, but it's one of the most visible examples of how getting audience understanding wrong can cost a lot of money.
The Association of National Advertisers tracked $123 million in real ad spend across 35.5 billion impressions and found that only 36 cents of every dollar reached consumers. Its 2024 follow-up showed some improvement, but still found that for every $1,000 spent programmatically, only $439 actually reaches consumers.
Furthermore, only 16% of internet users across all age groups feel represented in the advertising they see. 18% actively avoid all advertising. More people avoid ads entirely than feel represented by them.
This is the environment in which marketers are increasingly using AI to make decisions about their audiences. If the audience definition is wrong, everything downstream can be wrong too: who gets targeted, which channels get funded, what creative gets made and which consumers never come into consideration.
Used with good inputs, AI can genuinely sharpen audience understanding. But used to turn incomplete, inherited information into audience assumptions, it can make an existing problem worse, scaling those assumptions with more confidence, consistency and polish than they have earned.
The answer isn't to stop using AI for understanding audiences. It's to give it authentic and verifiable human data to work with.
.webp?width=495&height=317&name=pink_thumb_graphs%20(1).webp)
.webp?width=495&height=317&name=pink_thumb_letter%20(2).webp)