AI in Politics, Communications at Scale, the 2024 Election, & More
How AI is changing our approach towards message distribution and amplification
Key insights for AI leaders
AI's biggest impact wasn't in creating content but in amplifying it through user engagement that triggers further user shares and platform network effects.
The technology democratized sophisticated campaign tactics, allowing smaller campaigns to compete with well-funded operations in areas like microtargeting and data analysis.
Campaign messaging speed and precision increased dramatically through AI-powered testing and iteration, particularly effective in fundraising where success metrics are clearly defined.
New AI-powered campaign tools transformed traditional outreach methods, from multilingual robocalls to automated text messaging and personalized phone banking with AI voice agents.
By now, it looks like the dust has largely settled on the 2024 elections. Regardless of which candidate you were rooting for, this was a monumental year for voter turnout, approaching the historic levels of the 2020 election, and we saw incredible numbers for campaign donations and fundraising as a whole.
I think one of the biggest differences in this election versus past years, was the increasing role of AI in political campaigns and how integration of AI tools across the board has changed the way strategists approach campaigning as a whole.
Compared to previous election cycles, this was the first time we’ve had AI tools and capabilities so widely available for anyone to use.
This type of technology used to be only available to the most technologically advanced and well-funded campaigns, but with the introduction of open-source LLMs and cheaper compute resources, there are numerous new companies in the political tech space who are offering AI-based products, effectively leveling the playing field for campaigns of all sizes.
While public discourse has focused heavily on the threat of AI-generated deep fakes, based on what I’ve seen, AI has had the largest impact on messaging distribution and campaign operations. From a communication perspective, while these are closely related verticals, the way AI has been integrated into new and existing tools has a number of important differences.
Let’s unpack what all this means and how this will fundamentally change the way we view and approach political campaigns in the future.
The new megaphone: AI’s amplification effects
The photo above describes a modeled social media experiment, where the red dots are fake social media accounts, light blue dots are human users exposed to higher-quality content, and black dots are human users exposed to lower-quality content.
Users are exposed to more low-quality content when fake accounts infiltrate users’ networks and when the fake accounts generate more deceptive content. The right column shows greater infiltration, and the bottom row shows greater amounts of deceptive content.
Broadly speaking, AI’s amplification effects have irreversibly changed social media dynamics and content distribution strategies.
Engagement has always been the “north star” in terms of metrics, but the efficacy and manner in which we pursue engagement has changed in the era of algorithmic feeds and the new “attention economy”.
Social media dynamics
The algorithmic feeds used by social media platforms are designed to serve users content that they’ll find most engaging, but this also means that they’re vulnerable to manipulation.
With bot capabilities becoming increasingly refined with the use of AI, it’s become increasingly difficult to differentiate between real users and fake users.
“The problem is not that AI is too powerful; rather it’s that AI is not powerful enough.” -Zelly Martin, Senior Research Fellow, UT Austin
Using bots, adversarial actors can automate engagement (likes, retweets, shares) to game social media companies’ recommendation algorithms.
While isn’t a particularly new phenomenon, in the context of an election, outrage is a powerful motivator for human action. This fake engagement will amplify content that is poignant on a personal level to real users who in turn, promote it to their own followers.
This happened with a baseless voting clam being amplified by a network of social media accounts and video being circulated of a Haitian man who claimed to have voted multiple times in Georgia.
It’s becoming increasingly clear that the sophistication of these fake engagement networks has grown dramatically with the integration of AI and LLMs. Earlier this year, researchers identified a network of 1,140 bots that used ChatGPT to generate content and boost engagement of the generated content through replies and retweets. For social media platforms, fighting this battle will be increasingly difficult. Whereas before, the main concern was detecting AI generated content, these social media companies will now need to contend with AI content dissemination – which is a much trickier task.
Content distribution strategies
“It’s easy to write text that proposes a certain view, but hard to get many people to read it.” - Darrell West, Brookings Institute
Anyone can write a tweet or shoot a video, but the real challenge is getting your message seen. With the increasing adoption of AI tools, the approach to content distribution has changed significantly. Now, it’s easier than ever to increase the volume and frequency of messaging.
What would’ve taken a team of writers and content specialists can now be automated and done with AI.
One person can write a single core message and then use AI to generate a wide variety of similar messages in different tones or styles.
They can then take these “AI-rephrased” messages and automate their distribution – making it possible to continuously create and send out “fresh” content.
It’s not hard to imagine the scale and network effects on engagement a team of say 4-5 people could have with this method.
Another avenue in which AI has changed how we think about content distribution is SEO manipulation and the creation of data or information voids.
Using AI to scale up content generation and frequency, you can now have a deluge of misleading content that makes it significantly more difficult to find truthful and reliable information.
This creates “data voids”, topics around which there is a lack of quality, reliable information, that can be further exploited by the promulgation of misleading content. For non-english language content, this problem is even more salient as there is less of an existing knowledge-base to contend with.
For example, by using specific and curated keywords to game the ranking algorithms of search engines, politically motivated actors can ensure the proliferation of content when someone searches for certain topics.
Even if the content isn’t particularly engaging, it can still influence the narrative through repetitive reinforcement. The more times something misleading is seen, the higher the likelihood is that people believe it to be true. This cycle continues and makes it increasingly challenging to find and recognize reliable content amid a sea of disingenuous sources.
Campaign operations transformation
What’s particularly notable about AI and the way it’s transformed campaign operations is that it’s broken down the cost and expertise barrier for smaller, less-funded campaigns. AI technology has democratized many of the sophisticated methods employed in voter outreach and message targeting that were once the exclusive domain of well-funded operations.
Data analysis and microtargeting
The concept of microtargeting really came into the spotlight with Cambridge Analytica and the 2016 election, but things have changed quite a lot since then.
The landscape of voter profiles and targeted advertising looks nothing like it did back in 2016. The use of data in campaigns and electioneering is now the norm, and has even taken precedence over typical forms of voter “pulse finding” so to speak. With a wealth of tools at their disposal, campaign strategists can utilize large swaths of data to determine, with some degree of numerical certainty, the best path forward to secure a win.
These large data sets offer detailed insights into citizens’ behaviors, interests, and locations in order to advance their key goals of voter mobilization and fundraising. However, in years past, analyzing this data and actually putting it to work required a substantial amount of labor and domain expertise, as well as significant tooling and infrastructure. Now, with AI, data-driven insights are far more accessible and even more powerful than before.
Companies like FiscalNote and Chorus AI offer a full suite of AI-powered data management solutions and tools that make it easy to synthesize information and pull meaningful insights from a wide range of sources.
What would’ve taken multiple teams of experts can now be done by a small group of people.
Rapid testing and iteration
The speed with which campaigns can test and iterate on messaging has accelerated dramatically with the integration of AI across the communications tech stack.
Campaigns can now:
Create and deploy multiple message variants simultaneously
Analyze engagement in real-time
Refine content based on performance data
Scale successful messages quickly across platforms
Leveraging AI content generation tools, such as Quiller, campaigners can create hyper-personalized advertisements, customized for specific voter profiles based on geographic and demographic information.
For a political campaign wanting to address the unique concerns of different voters, such as the rising cost of groceries or reproductive health care costs, AI can very feasibly assist with developing and targeting messages that address all of these concerns and more.
On the platform side, companies like Meta and Google have already implemented AI-powered tools for advertisers that make it easier to personalize ads and serve them to the right audience at the right time.
This AI-driven methodology has proved to be particularly effective in fundraising efforts, based on interviews with campaign strategists.
Why is this the case you might ask?
Well, compared to voter persuasion, where it’s not entirely clear what factors influence a voter’s decision between when they receive a message and the time they cast their ballot, fundraising is tied to a very clear outcome: did they donate or not?
Keeping this example in mind, it’s evident that AI-powered messaging and targeting isn’t a silver bullet for campaigns, but it definitely makes a strong impact in moving the needle.
Tools and technologies
The 2024 election cycle has seen the emergence of a large ecosystem of AI-powered campaign tools. These range from comprehensive campaign management platforms to specialized tools for specific tasks.
These tools can also be combined to accomplish an even wider range of campaign objectives. For example, in the NYC mayoral election, Eric Adam’s team used AI tools to make multilingual robocalls to potential voters. Numerous other campaigns across the US utilized a combination of demographic data and AI voice agents to automate and tailor phone banking efforts based on geographic location.
I want to give you a few examples of current products just so you can understand what’s possible now and begin to imagine what’s possible in the future as AI capabilities continue to develop.
Data Analysis and Voter Profiling
CIVA for civic intelligence gathering
BallotReady for voter behavior analysis
Communication and Outreach
DaisyChain for automated personalized text messaging
Synthflow for AI voice agents used for personalized phone banking
Content Creation and Distribution
Quiller for targeted content generation
BattlegroundAI for advertising optimization
Considerations for product leaders using AI
The 2024 election revealed that AI’s been incredibly impactful on political messaging with regards to distribution and amplification. Campaigners and communications experts have been incredibly effective in utilizing AI tools to iterate rapidly, test, and distribute at scale.
AI content generation, while being a useful cost-saving tool, has been less of a needle mover than previously expected. However, generative AI will continue to be a useful tool in brainstorming and ideation, allowing users to focus their creative efforts where it’s most needed.
The future of product campaigning will likely be shaped not by who can create the most convincing content, but by who can most effectively harness the tools at their disposal to ensure their message reaches and resonates with the right audience.
Disclaimer: Amit may be an investor in the companies discussed.