This week, New York is hosting Advertising Week, and unsurprisingly, one of the main topics of discussion is AI and its influence on everything: advertisers, agencies, adtech, and publishers. With that in mind, we can expect everyone to showcase their AI capabilities and products.
For instance, TikTok today announced several new AI-powered ad tools, which we’ll discuss in more detail later.
To connect with that AI focus, we conducted our own (random) experiment. We took The New York Times’ ads.txt file and tasked the top gen AI platforms—ChatGPT, Gemini (Google), Perplexity, and Anthropic—with analyzing it. Each platform received the same two prompts: “What is this?” (accompanied by a screenshot of the file) and “Are there any interesting or notable patterns in the file?” Our goal: to just see what happens.
Here’s a summary of each platform’s response.
ChatGPT
Prompt 1: “What is this?”
Prompt 2: “Are there any interesting or notable patterns in the file?”
Gemini (Google)
Prompt 1: “What is this?”
Prompt 2: “Are there any interesting or notable patterns in the file?”
Perplexity
Prompt 1: “What is this?”
Prompt 2: “Are there any interesting or notable patterns in the file?”
Anthropic
Prompt 1: “What is this?”
Prompt 2: “Are there any interesting or notable patterns in the file?”
Why This Matters:
Does this matter? We’re not so sure. Still, while this experiment was mostly for fun, it got us thinking about whether AI could be used to check ads.txt files for issues. The New York Times’ file is fairly simple to understand, but larger files might require more automated analysis.
For example, it took ChatGPT just seconds to analyze Newsweek’s ads.txt file, which has a greater number of entries (neither good nor bad, just a fact), and noted: “Newsweek’s file shows a larger proportion of RESELLER entries, indicating they may rely more heavily on indirect sales partners for their programmatic ads.”
Experts React:
We asked Alexandre Nderagakura, adtech, programmatic, data, and privacy expert, and former Tech Director of IAB Europe, for his take on our experiment.
Regarding ChatGPT, he noted that “the AI incorrectly identified the file as sellers.json; it’s actually ads.txt. It seems the AI attempted to make a connection with sellers.json, which makes sense.” He also added that it “was also wrong about ‘6a698e2ec38604c6’—this is not a campaign ID but a certification authority ID, such as a TAG ID.”
Ultimately, according to Nderagakura, “Overall, the conclusions made about the ads.txt file, the relationships, and the volume of entries per ad tech are fairly simplistic. However, special mention goes to the last AI (Anthropic), which showed some understanding of the ecosystem by discussing Adswizz and audio ads for The New York Times.”
Our Take:
This was fun. If you have any other experiments we should consider running using an AI, sound off in the comments or DM us at @AdTechRadar.
