Generative AI is transforming industries everywhere. Politics is no different.
For the last year, Higher Ground Labs has been exploring and advancing the applications of AI* in politics. One of the most powerful applications of AI we identified is in content generation: using AI to draft, produce, and edit content. From HGL’s landscape analysis, we knew this was one of the top emerging usage areas for the political ecosystem. So we sought to study it.
What role might AI play in political content creation? What resource savings might AI drive? How might the use of AI affect persuasiveness?
As we enter one of the most consequential elections of a generation where over $10 billion is estimated to be spent on political ads, we set out to research the appropriate role AI might play in political content through this series of randomized controlled trials (RCTs).
*In this report, we refer to “generative AI” as “AI”.
Executive Summary
Our research question: Can AI save time and money without compromising persuasiveness?
The answer: Yes, but not always.
Our findings show that AI-assisted content can perform as well as human-created content, but not consistently. While AI can reduce costs and speed up production, it is not guaranteed.
Why? While AI can save resources, its effectiveness depends on how it is integrated into the content creation process. In our experiment, greater AI usage led to progressively better outcomes. Teams with more AI expertise gained a competitive edge, saving resources while achieving results on par with the human-created ads from our industry’s best.
It is clear that AI does not negatively impact performance. Our tests showed no backlash or negative reaction to its use, confirming that experimenting with AI is safe.
These tests capture a snapshot of a particular moment in time — and an important one, with Election Day less than 50 days away. AI is a rapidly changing space. New tools that weren’t on the market when our test started became available after it concluded. Staff are upskilling and adapting their workflows to incorporate new AI features. We hypothesize that with increased testing and time, AI will become an even more potent resource for political content experts.
This is the experimentation stage of AI, and we should take this opportunity to do so.
The Experiment
In August 2024, we partnered with several industry-leading agencies to create ad content across varying ”lanes” of AI usage. The task was to create short video and image ads aimed at influencing presidential vote choice, focusing on the issue of abortion. A total of 15 pieces of content were submitted and tested, with a placebo used as the control baseline for comparison.
We designed four “lanes” of AI to test, with each lane progressively increasing in AI usage. Across each lane, we worked with agency partners to produce content according to specific instructions.
We ran survey experiments sampling adults in 11 key battleground states, with the final analysis consisting of 37,867 respondents.
Learn more about the tests and results by reading the full research report. We go more into depth on the experiment design, detailed results, our interpretation, and recommendations.
We are grateful to our partners who helped make this possible: our excellent research partners Grow Progress and our industry-leading agency content partners AL Media, Bully Pulpit International, Jumpstart AI, Rising Tide Interactive, Tavern Research, and Trilogy Interactive.
If you have any questions or feedback, or want to partner with us on AI experiments, reach out to ailab@highergroundlabs.com.
Higher Ground Labs is the leading funder, accelerator, and incubator of political technology. HGL’s Progressive AI Lab is the central hub for AI innovation within the Democratic and progressive landscape.