Artificial intelligence is becoming increasingly capable of uncovering personal information in ways many internet users would never expect. New Australian research has found that AI systems can infer sensitive details about people simply by analysing the online advertisements they are shown — even without access to browsing history, search data or personal accounts.
Australian Researchers Warn of a New Privacy Risk
A study led by researchers from the University of New South Wales (UNSW) and Queensland University of Technology (QUT) has revealed that targeted advertising itself may expose more about individuals than previously understood.
The research examined how large language models (LLMs) — the technology behind modern AI chatbots and analytical systems — can analyse advertising patterns to predict personal traits such as age, gender, education level, political preferences, employment status and financial circumstances.
Lead author Baiyu Chen from UNSW said the ads people see online are far from random.
“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 said.
“Our study shows that LLMs can analyse those patterns and infer private attributes from ad exposure alone.”
The study, titled When Ads Become Profiles: Uncovering the Invisible Risk of Web Advertising at Scale with LLMs, adds to growing concerns about the rapid expansion of AI-driven surveillance and data analysis.
How AI Builds Profiles From Advertising Data
The research formed part of the Australian Ad Observatory project run by the ARC Centre of Excellence for Automated Decision-Making and Society (ARC ADM+S).
Researchers analysed approximately 435,000 Facebook advertisements viewed by 891 Australians. According to the study, AI systems were able to process and interpret the advertising data around 50 times faster and roughly 200 times cheaper than human analysts.
The findings suggest that even short periods of internet browsing can generate enough advertising data for AI systems to construct highly detailed personal profiles.
While the AI-generated profiles may not perfectly represent an individual, researchers said they are often accurate enough to reveal significant insights into a person’s lifestyle, financial position and stage of life.
The study also found that publicly available AI models performed as well as — and in some cases better than — humans at predicting private characteristics from ad exposure patterns.
Major Platforms Still Expose Sensitive Signals
Large digital platforms including Google, Meta, TikTok and X officially restrict advertisers from directly targeting users based on highly sensitive information such as health conditions, sexuality, political beliefs or financial hardship.
However, the researchers argue that these characteristics remain indirectly embedded within the algorithms that determine which advertisements users receive.
In practice, this means AI systems can still extract meaningful information from ad patterns without needing direct access to protected data.
The report warned that ordinary digital tools could potentially be repurposed for large-scale data harvesting.
“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 stated.
Browser Extensions Could Become a Weak Point
The study identified browser add-ons such as ad blockers, coupon tools and translation extensions as potential privacy vulnerabilities.
Because these tools often request broad permissions to access web content, researchers warned they could theoretically gather advertising data without users fully understanding the implications.
For cybercriminals using AI-powered profiling systems, this creates new opportunities for sophisticated scams and personalised fraud attempts.
Instead of stealing passwords or hacking devices directly, attackers could potentially build behavioural profiles based purely on advertising exposure and use those insights to tailor convincing phishing campaigns.
This concern arrives as Australians continue to face rising levels of cybercrime, online scams and identity fraud. The Australian Competition and Consumer Commission’s Scamwatch program has repeatedly warned that scammers are increasingly using personalisation techniques to make fraudulent communications appear legitimate.
Calls for Stronger Privacy Regulation
The researchers argue that Australia’s privacy laws and digital platform regulations may need to evolve to address not only the collection of personal data, but also what can be inferred from the content users are exposed to online.
Chen said responsibility cannot rest solely with individual users.
“This is not something users can fully solve on their own, because the broader issue is systemic,” Chen said.
“People cannot easily opt out of the ad ecosystem altogether, so stronger platform safeguards are also needed.”
How Users Can Reduce Risk
Researchers recommend several steps that may help reduce exposure:
- Limit unnecessary browser extensions
- Review extension permissions carefully
- Use browser privacy settings
- Reduce ad personalisation where possible
- Regularly audit app and browser access permissions
Even so, the study suggests the broader challenge lies in the structure of the modern digital advertising economy itself.
AI Is Expanding the Meaning of Personal Data
For years, concerns around online privacy focused mainly on information users knowingly shared — such as browsing histories, search queries and social media activity.
This latest research suggests that simply viewing online advertisements may now reveal just as much.
As AI systems become more sophisticated, digital footprints are no longer limited to the data people intentionally provide. Instead, the content users are exposed to online may itself become a powerful source of personal intelligence.
The findings raise difficult questions for regulators, technology companies and consumers alike about how privacy can realistically be protected in an increasingly AI-driven internet.
“Student. Subtly charming organizer. Certified music advocate. Writer. Lifelong troublemaker. Twitter lover.”
