Dustin Lloyd, a Democratic primary candidate for the Missouri state legislature, was a deeply rooted member of his community long before his political aspirations began.
However, he encountered a significant gap in his digital footprint: A.I. chatbots seemed entirely unaware of his candidacy.
When voters queried platforms like OpenAI’s ChatGPT or Google’s Gemini about Mr. Lloyd, the responses yielded only generic data that failed to highlight his specific commitment to supporting small businesses.
To resolve this, Mr. Lloyd optimized his online presence by publishing a detailed Q&A on his official campaign website. This strategic update worked; subsequent chatbot inquiries successfully linked his personal history to his primary policy objectives.
The process of refining A.I. outputs was surprisingly straightforward, yet Mr. Lloyd noted that he must continue to monitor and update this information daily.
While politicians have traditionally managed their reputations among human voters, they must now contend with the perceptions of artificial intelligence. This shift has given rise to an entirely new sector of political consulting.
Mr. Lloyd, 33, utilized an A.I. presence report from CampSight, a tool launched by the progressive group Run For Something Action Fund. The tool identified that Mr. Lloyd’s digital footprint across Wikipedia, Ballotpedia, and his personal website was insufficient for AI recognition. It even recommended increasing engagement on platforms like Reddit to ensure chatbots would ingest more of his campaign’s messaging.
“Everyone is searching for tools to at least analyze these AI results and, ideally, influence them,” explained Jordan Haines, Chief Technology Officer at Run for Something.
This burgeoning industry, known as Answer Engine Optimization (A.E.O.), is a direct response to the evolving way chatbots generate information. Early models relied on static, massive datasets that often became outdated quickly. Modern chatbots, however, can browse the live internet to provide real-time information, including details on current political candidates and elections.
While current A.E.O. tools often focus on updating Wikipedia or refining a candidate’s official website, experts suggest that well-funded campaigns may soon deploy much broader, aggressive content strategies specifically designed to be “parroted” by AI models.
“There is a palpable sense of urgency because we are dealing with a technology that is both poorly understood and rapidly evolving,” said Beth Simone Noveck, a professor at Northeastern University specializing in democracy and large language models (LLMs). “Even for those who claim to master these techniques, the unpredictable nature of LLMs makes full control nearly impossible.”
Politicians also face the danger of “AI smear campaigns,” where search results provide factually incorrect or defamatory information. An experiment preceding Scotland’s May parliamentary elections revealed that over a third of AI-generated responses regarding the vote were inaccurate. Researchers at the British think tank Demos found that chatbots frequently hallucinated details, including fabricating scandals, misstating candidate positions, or inventing political figures entirely.
Despite these risks, voters are increasingly relying on AI for political information. Caucus AI, a political A.E.O. research firm, estimates that at least 16 million voters now utilize chatbots or AI-driven search results to gain election insights.
“As AI reliability increases, we expect a significant surge in usage by 2028 and beyond,” said Meg Schwenzfeier, a founder at Caucus AI and former chief analytics officer for Kamala Harris’s presidential campaign.
Caucus AI’s research into how quickly new Wikipedia content is indexed by AI found a window of only about 12 minutes. This suggests that candidates can rapidly influence AI responses by making quick, unvetted updates to public information.
Caucus AI is currently monitoring AI sentiment regarding all Senate, House, and gubernatorial candidates. They found that certain messaging styles successfully penetrate AI responses. For instance, when asked about Democratic Senate challenger Mary Peltola in Alaska, multiple chatbots explicitly cited her “Fish. Family. Freedom.” slogan.
“We are investigating what causes specific messaging to be prioritized,” said Ms. Schwenzfeier. “The models themselves remain largely ‘black boxes.’”
This ability to nudge AI responses raises concerns among digital strategists regarding foreign interference and disinformation. If legitimate candidates can manipulate AI, so can bad actors.
“Disinformation actors are early adopters by nature—they will utilize every available technology to achieve their goals,” said Tim Chambers of Dewey Square Group, a public affairs firm serving various government and labor clients.
Dewey Square Group’s research has found that far-right and low-integrity factual sites often optimize their websites to be more “scrappable” by AI tools compared to highly factual, center-left outlets.
Mr. Chambers noted that his clients are now considering “rebuilding websites for machines as much as for humans,” ensuring their presence is visible in multilingual AI queries.
For Missouri’s Dustin Lloyd, the tactical changes yielded immediate results. Initially, when asked for a recommendation in the Democratic primary, a chatbot suggested his opponent, Tanya Lakins, due to her small-business platform. After Lloyd optimized his digital content, the AI changed its response.
“If you’re voting in the Democratic primary,” the chatbot stated, “Dustin Charles Lloyd appears to have the strongest and most explicit small-business platform.”