Marketers are Wasting Advertising Spend on AI-generated Content

Photo by Tima Miroshnichenko.

TL;DR

As LLMs become more sophisticated, fraudsters and spammers will realize the opportunity to use generative AI to scale inauthentic content across the web with the intention of stealing advertising dollars.

Google has already taken action to downrank AI generated publications in their search algorithm, but brands and advertising agencies need to develop solutions to identify and filter AI generated content before their ads are misplaced on made-for-advertising content.

Background

Back in March, Google announced it would be taking significant action against the rise of AI generated content appearing in users’ search results. After the mainstream adoption of LLMs in 2023, bad actors realized they could produce large volumes of low quality content, stuffed with keywords, to game search ranking algorithms and boost their website’s ranking. Google classified this type of SEO hacking as “scaled content abuse” and changed their policies to reduce the visibility of these sites in results - or at least by 40% as Google claims.

Brands typically work through digital agencies to manage their online advertising campaigns. Marketing copy is sold and distributed most effectively through programmatic advertising, allowing advertisers to target users by placing those ads adjacent to relevant content (referred to as buying inventory on a publisher’s site). This process allows a supplements company, for example, to automatically place their marketing copy next to a fitness blog post or on a yoga forum.

Why should advertisers care?

Programmatic spend, which exceeded $150B in 2023 for spend in the USA alone, is a system vulnerable to manipulation. Already, advertisers have had to deal with advertising fraud related to Made-for-Advertising sites (MFAs) - sites created, usually with low-quality or plagiarized content, with the intent of selling as much ad space as possible.

Example of MFA (source - nonstopnostalgia.com)

Premium brands will find that their ads are sometimes on these sites. Just last year, 21% of ad impressions in a sample were attributed to MFAs and 15% of ad spend, or around $13B, was spent on these sites.

MFAs aren’t a new problem, but AI is becoming an easy, accessible tool for spammers to spin up content farming operations with low effort. While LLMs won’t significantly reduce the cost of artificial content generation, they will certainly compound the volume of these sites as inferred by Google’s recent reactions.

Fraudsters now have the tools to quickly scale low-quality sites within hours and capitalize on misallocated ad spend. In a study conducted last year, over 140 major brands ended up paying for ad space on these AI generated content farm sites. As models and tooling have improved over the past year, it will only keep getting easier for AI-generated MFAs to be created and published.

How should advertisers respond?

The most important takeaway for advertisers is recognizing the difference in value between human generated content and artificial generated content. Not all AI generated content is inherently bad or should be considered a brand safety risk, but its value certainly ranks lower in the eyes of the consumer.

• Reporting from the Edelman Trust Institute found that US consumer trust in artificial intelligence has fallen by 15% in the last five years
• A 2024 survey conducted by Bynder found that 52% of consumers were likely to disengage from AI generated content and that millennials were more likely to prefer human generated content.

All this means is that advertisers need more visibility in where their ads are being placed and more transparency into the authenticity of the content they are sponsoring. Agencies and brands may also eventually need to develop or adopt teams and solutions specifically to deal with inauthentic content in order to protect their spend. Gartner predicts that by 2027, 80% of marketers will have developed “content authenticity teams” to serve this purpose.

Fortunately, Pangram Labs is building the most accurate detection solutions to identify AI generated text in near-real time. We see a future where most digital marketers will need to react in real-time to block sites with AI generated spam. To that end, Pangram Labs is working to develop a frequently-updated MFA block list, and look forward to helping brands and advertisers recapture spend lost to AI generated content.

To learn more, get in touch with us at info@pangram.com!