A new academic study has revealed that AI models can figure out a lot about you from the ads you’re shown
Our personal traits are encoded in the algorithms that drive ads on major platforms such as Google, Meta, TikTok, and X
The researchers argue that policy “must evolve to address not just data collection, but what can be inferred from the content people are exposed to”
Not so long ago, your online search history and other personal data were needed for the digital masters to assemble a profile of you for advertising purposes.
Now a team of researchers from the University of New South Wales (UNSW) and Queensland University of Technology (QUT) have discovered that artificial intelligence (AI) doesn’t need such old-fashioned data points. It can suss you out just from the online ads you’re shown.
“The key point is that the ads a person sees are not random,” says lead author of the study, Baiyu Chen of UNSW. Its title serves as yet another warning about how AI can be misapplied: When Ads Become Profiles: Uncovering the Invisible Risk of Web Advertising at Scale with LLMs.
(LLMs are Large Language Models, an AI term for computer systems that can analyse and recreate massive amounts of human language and approximate thought.)
The key point is that the ads a person sees are not random
Lead author Baiyu Chen of UNSW
“Advertising systems optimise delivery based on inferred profiles and behaviours, so the overall pattern of ads shown to a user can carry signals about traits such as gender, age, education, employment status, political preference, and broader socioeconomic position,” Chen says. “Our study shows that LLMs can analyse those patterns and infer private attributes from ad exposure alone.”
The research team – which also included professor Flora Salim, professor Daniel Angus, Dr Benjamin Tag and Dr Hao Xue – are part of the ARC Centre of Excellence for Automated Decision-Making and Society (ARC ADM+S).
AI systems can build convincing profiles of us based on targeted ad data after we browse the internet for just a short while
The study is the result of the organisation’s Australian Ad Observatory project, which involved analysing around 435,000 Facebook ads seen by 891 Australian users. The findings mean AI can sift through markers of our selves we didn’t realise were there, and do it 200-times cheaper and 50-times faster than humans.
But what’s really concerning is that AI systems can build convincing profiles of us based on this targeted ad data after we browse the internet for just a short while, and they can infer personal characteristics as well as – or better than – humans.
Popular LLMs available to us all “can accurately reconstruct complex user private attributes”, the researchers write. And while the AI versions of who we are, based on the ads we see, may not capture everything, “they are often close enough to reveal meaningful insights about a person’s life stage or financial situation”.
All of this increases our risk of being scammed, since AI tools analysing ads can bypass current privacy protections, create profiles of us, and allow scammers to personalise their attacks.
Major digital platforms such as Google, Meta, TikTok, and X restrict certain sensitive categories of information from being used in targeted advertising, including an ad target’s health status, gender, sexual orientation, financial situation and political affiliation.
But these personal traits are still encoded in the algorithms that drive the ads on these platforms, and widely available AI models can extract this information.
“Everyday tools such as browser extensions could be repurposed to quietly collect ads and build detailed user profiles — bypassing platform safeguards and leaving little trace,” the researchers report.
Everyday tools such as browser extensions could be repurposed to quietly collect ads and build detailed user profiles — bypassing platform safeguards and leaving little trace
The vulnerable extensions – including ad blockers, coupon finders, and page translators – provide “perfect cover for data harvesting”, the researchers write. It means a cybercriminal skilled at AI tools wouldn’t have to hack your device or obtain personal information from you in order to set up a targeted scam.
The researchers argue that policy and regulation “must evolve to address not just data collection, but what can be inferred from the content people are exposed to”.
Chen says platform users can reduce the risks by limiting the use of browser extensions, limiting unnecessary permissions, and using any available privacy and ad-personalisation settings in their web browser settings.
But the platforms are going to have to tackle the new privacy risks brought on by the age of AI as well.
“This is not something users can fully solve on their own, because the broader issue is systemic,” Chen says. “People cannot easily opt out of the ad ecosystem altogether, so stronger platform safeguards are also needed.”
In other words, we’re leaving digital footprints for AI to follow every time we go online, and what will be done with this record of our personal attributes is anybody’s guess.
Get the inside story on our investigations into consumer rip-offs and bad business practices.
Andy Kollmorgen is the Investigations Editor at CHOICE. He reports on a wide range of issues in the consumer marketplace, with a focus on financial harm to vulnerable people at the hands of corporations and businesses. Prior to CHOICE, Andy worked at the Australian Securities and Investments Commission (ASIC) and at the Australian Financial Review along with a number of other news organisations. Andy is a former member of the NSW Fair Trading Advisory Council. He has a Bachelor of Arts in English from New York University. LinkedIn
Andy Kollmorgen is the Investigations Editor at CHOICE. He reports on a wide range of issues in the consumer marketplace, with a focus on financial harm to vulnerable people at the hands of corporations and businesses. Prior to CHOICE, Andy worked at the Australian Securities and Investments Commission (ASIC) and at the Australian Financial Review along with a number of other news organisations. Andy is a former member of the NSW Fair Trading Advisory Council. He has a Bachelor of Arts in English from New York University. LinkedIn
For more than 60 years, CHOICE has been fighting the good fight for Australian consumers.
In the past year alone we've uncovered systemic issues with sunscreens, investigated shonky supermarket pricing, fought for stronger scam protections and helped make complex energy pricing fairer and clearer.
CHOICE is here to provide unbiased advice and independent testing in our world-class labs. We buy the products we test, just like you do, and our expert reviews are influence free. We’re here to help you choose smarter. Hopefully you’ll also save some money along the way.
Thanks to CHOICE, you’ll never be alone when a business treats you unfairly. You can support our work by joining or donating to our cause.
