GenAI & MFAs: The bullies that want your advertising lunch money

GenAI is actively being used to produce filler content for ad farms by bad actors seeking to make a quick buck. And, throughout the years, MFAs have been straightforward enough to identify and cut from your site lists with a combination of brand safety pre-bid tools and human oversight. This is still true, but the growth and accessibility of genAI is complicating matters.

How is GenAI making the MFA problem worse?

Until recently, if a programmatic trader noticed their campaign had served thousands of impressions on an unfamiliar URL with little to no performance, a quick visit to the site would confirm if it was an ad farm. These sites would take eons to load due to the high volume of ad calls; ad formats or sizes would appear in spaces not designed for them; and most importantly they had no content. There may have been a bland, general appeal homepage surrounded by banners, but the rest was an empty shell stuffed with ads.

The made-for-advertising sites of today are a bit sneakier, in part due to genAI. Ad fraud has always been a moving target, but one that was easy to spot. Fraudulent sites rely on content scraped from real publishers to appear legitimate, designed to fool an advertiser’s spot check. Typically, they won’t be shut down due to fraudulent ad revenue, but because of copyright violations.

GenAI enables the builders of fraudulent sites to minimize this risk, quickly creating their own ‘new’ content. This approach also allows fraudsters to go beyond scraping general interest and news; they can use genAI to craft more customized or niche-interest content with minimal effort, all while capturing more traffic and appearing more legitimate.

If they have content and my audience is there, why should I care?

MFAs are far from the only sites compiling material scraped from across the web – or the only ones experimenting with genAI to pad out their content. So, what makes an MFA different than audience-beloved staples like Buzzfeed, Reddit, or Upworthy? The answer is twofold: terrible user experiences, and terrible advertiser experiences.

MFAs have no real reason to care about users’ satisfaction with their website experience, except perhaps to minimize bounce rates. This is where poor advertiser and audience experiences are one and the same; unlike a standard web page that sends an ad call only when a page is loaded, MFAs often only display ads for seconds at a time before sending out fresh ad calls and rotating the ads on screen.

The rationale behind this is likely the expectation that someone landing on the page will have minimal time-on-site – so if the MFA has captured the attention of a user willing to stay a while, it can maximize profits by serving as many impressions as possible. This leads to a poor user experience for the audience, who will be turned off by the flashing-banner feel if they hadn’t left the site already, and for the advertisers accidentally purchasing the inventory equivalent of shouting their brand name out of a moving vehicle.

How to filter out the worst of the worst

Both MFAs and genAI are likely to proliferate in the coming years — without strong quality monitoring, your risk for wasted impressions will grow. Due to dynamic ad fraud, the problem will have to be tackled both pre-bid and post-bid. The need for blacklists and consistent site/app monitoring will never go away, but thankfully there are preventative steps to take to ensure your strategy is as clean as it can be.

First, embrace an active role with the tech and inventory partners attached to your campaigns. Take advantage of your demand-side platform (DSP)’s partnerships with quality and fraud monitoring settings pre-bid, but also consider investing in a secondary post-bid partner of your own to ensure no one company is grading their own homework.

Take a hard look at the exchange partners included in your default settings – not every campaign can scale using only programmatic guaranteed or private marketplace inventory, but you have more choices than executing everything directly or fully throwing bids into the void. Use your post-bid analyses to inform which supply-side partners are contributing the most questionable inventory, and filter them out where necessary.

Finally, use the past year’s AI fervor for your own benefit. AI tools’ understanding of speech has never been better, allowing for much more sophisticated content blocking than the keyword matches of yesteryear. Experiment with in-platform blocking at category- and keyword-level, and engage with partners who offer additional filters like sentiment analysis to find the ideal content for your creatives.