The Best Adblocker – Using Artificial Intelligence to Perfectly Block Ads

Princeton and Stanford’s new computer vision ad-blocker may soon change the relationship between advertisers, publishers, and readers.

The Best Adblocker - Using Artificial Intelligence to Perfectly Block Ads
Illustration by Mike Smith

Ever since Google figured out a way to monetize an internet business via ads, it’s become the go-to monetization strategy for most internet companies. But as ads began to clutter websites, developers fought back by distributing ad-blockers for free. Since then, it’s been a back and forth race as advertising firms and publishers try to improve ads and readers trying to find the best adblocker.

While ad-blocking is legal, its use has had significant ramifications for how websites pay their operational costs. On the one hand, heavy and invasive advertisement scripts slow down website performance and degrade user experience. On the other hand, it allows many companies to distribute content or products for free in exchange for its users watching a couple seconds of ads.

Regardless of where you stand on this issue, the growing adoption of ad-blocking software has already made companies think about new approaches to monetize their product. If you’re reading this on Medium, their recently introduced membership program is one example of an attempt to maintain the core product experience without resorting to ads. Either way, the team at Princeton and Stanford believe that their new ad-blocker will fundamentally change this relationship.

So How Does The Best Adblocker Work?

Popular ad blockers such as Adblock Plus scan the webpage for common urls, scripts, and markup codes used by ads and use these characteristics to detect and subsequently block the ads from rendering on the page.

However, Adblock Plus resorts to an open source list of standard scripts to pick out such ads. Thus, whenever advertisers and publishers change how they deliver their ads (e.g. Adblock Plus can’t block WebSockets so YouTube ads will not be blocked), ad-blockers will become less effective. Also, traditional ad blockers struggle with native ads that have similar code structure as normal content with the prime example being sponsored ads on Facebook.

The new ad-blocker, called perceptual ad-blocker, uses computer vision, instead of code matching, to detect and block ads. It uses OCR (optical character recognition), container searches, and other computer vision techniques to mimic how humans recognize ads.

This method is more effective since regulations exist to enforce that advertisements must be clearly labeled. Essentially, the FTC ruled that people must be able to recognize ads. Thus, if computer vision software can look for clues the same way a human would to recognize ads, it can always beat ad firms who must comply with the regulations, becoming the best adblocker possible.

Is This The End of All Ads?

Unfortunately for users — and fortunately for ad firms — this tool isn’t fully functional. It only detects ads, but doesn’t block them because the creators decided to “avoid taking sides on the ethics of ad-blocking.” Also, while the legal climate is currently friendly towards ad-blockers, a significant dent in how current publishers and advertising firms operate may pressure them to call for newer regulations and standards.

Perceptual ad-blockers also take advantage of the fact that browser extensions are given a higher priority and code privilege than ads or anti-ad blockers. One can imagine Google, Apple, and Microsoft changing their browser policies so that browser extensions get limited capabilities on their respective browsers. (Edits on April 22: it appears that Google already has plans to introduce its native Ad-Blocking features on Chrome — via WSJ).

Still, this research is the latest example of the shift in the online advertising space. More and more users are using ad blockers and as this war between blockers and anti-blockers intensifies, companies must adapt and shift towards different models.

Yitaek Hwang
Yitaek is the Director of R&D at Leverege who loves learning about IoT, machine learning, and artificial intelligence. He graduated from Duke University with a dual degree in electrical/computer and biomedical engineering and is a huge Cameron Crazie.