AI chatbot-tailored promotional strategies enter U.S. politics as voters change how they seek information
Competition for exposure in AI answers heats up in the commercial sector, intensifying GEO and AEO marketing
Concerns over the spread of false and biased information make verifying information reliability a growing challenge

Candidates in U.S. elections are adopting new campaign strategies targeting artificial intelligence (AI) chatbots. As more voters use generative AI to research candidates and policies, a practice known as generative engine optimization (GEO) is gaining ground. Some observers, however, warn that because generative AI can present inaccurate or biased information as fact, turning political information into a battleground for AI optimization could distort voters’ judgments about the information they receive.

AI Chatbots Become Part of Election Campaigns

On the 26th (all dates hereafter are local time), Reuters reported that it had obtained a campaign strategy document prepared for Mike Minogue, a Republican candidate for governor of Massachusetts. The document outlines a plan to get major AI chatbots, including ChatGPT, Claude and Gemini, to portray Minogue favorably when answering questions about key policy areas such as taxes, housing and public safety. It was prepared by Visigent, a digital strategy firm based in Austin, Texas.

Visigent proposed creating and publishing hundreds of pages and related materials explaining Minogue’s policies on websites frequently cited by AI chatbots, while paying highly trusted external sites to publish related content. The strategy also included distributing posts promoting his policies on YouTube and Reddit and continuously monitoring how major AI chatbots answer policy-related questions. Online information would then be revised and supplemented if the chatbot responses were deemed not to adequately reflect Minogue’s policies. It has not been confirmed, however, whether the strategy was actually adopted or implemented.

Voters Use AI to Research Candidates

Visigent said it charges political campaigns a minimum of USD 5,000 (approximately USD 5,154) per month to carry out such strategies, with costs potentially rising to USD 50,000 (approximately USD 51,385) per month when paid content placement and influencer recruitment are included. Presentation materials reviewed by Reuters also set out performance fees of up to USD 200,000 (approximately USD 205,500) for a so-called “category win”—securing AI chatbot recommendations in a specific policy area. Visigent CEO Thomas Howell told Reuters that the company had 6 political campaigns as clients but did not disclose their names.

The emergence of these AI-tailored campaign strategies reflects a shift in how voters seek information. At a media briefing on the 23rd, Cameron Hickey, CEO of the U.S. civic organization National Conference on Citizenship (NCoC), said the proportion of voters using AI chatbots to research election candidates was growing rapidly. According to an NCoC poll of eligible U.S. voters, the share of respondents who said they had used chatbots to look up candidate information rose by 11 percentage points, from 14% two weeks earlier to 25% in the latest tally. These figures, however, were preliminary results released ahead of the formal publication of the survey findings; details including the sample size, methodology and margin of error have not yet been confirmed.

GEO and AEO Gain Ground

The trend toward seeking information through AI chatbots extends beyond politics into online marketing. As more users rely on AI-generated summaries and answers when making purchasing decisions rather than visiting multiple websites to compare information themselves, AI chatbots have emerged as a new arena of marketing competition for businesses. GEO and answer engine optimization (AEO) are concepts that have emerged in response to this shift.

GEO is a strategy for optimizing how online information is structured and distributed so that generative AI references and cites content from a particular company or organization when answering users’ questions. While traditional search engine optimization (SEO) focuses on raising a website’s ranking in search results and attracting visitors, GEO emphasizes increasing the likelihood that relevant content will be incorporated into AI-generated answers themselves. AEO is also a strategy for adapting to AI-based search, but the term generally places greater emphasis on providing direct answers to users’ questions. It involves structuring content so that search engines’ featured answers, voice assistants and AI summary services can easily extract information addressing specific questions. However, the industry has not established a clear distinction between the two concepts, and their definitions may vary somewhat depending on who uses them.

Examples of Content Optimization in Business

Companies are increasingly entering search marketing aimed at generative AI. A leading example is U.S. software company Adobe. Last October, Adobe officially launched Adobe Large Language Model (LLM) Optimizer, which analyzes how brands are mentioned and cited in AI services such as ChatGPT and Gemini and helps optimize online content. It also applied the solution to its own products, Firefly and Acrobat. Alongside restructuring its website content, Adobe expanded the visibility of related information on external platforms, including specialist media outlets and social media. As a result, citations of Firefly in AI answers increased fivefold within a week, while Acrobat’s AI visibility relative to competitors rose by 200% and AI-driven website traffic increased by 41%.

Other companies are pursuing similar strategies. U.S. automaker General Motors (GM) optimized its webpages to help AI chatbots accurately recognize its brand and product information when answering vehicle-related questions, increasing brand visibility in AI search by 23% and citations by 35%. Swiss pharmaceutical company Roche also adopted a strategy using search marketing platform Semrush’s AI optimization solution to analyze how frequently and in what ways its brand appeared in AI chatbot answers. Subsequently, according to Semrush’s figures, Roche’s share of brand mentions in AI answers across the markets tracked rose to 67.3%.

Table 1. Examples of Generative AI Search Optimization by Global Companies

AI Chatbots Still Face Technical Limitations

Some observers, however, are voicing concerns about the search market’s current shift toward generative AI. Given the nature of LLMs, the risk of information whose factual accuracy has not been verified spreading to a broad audience cannot be ruled out. LLMs essentially learn statistical patterns between words and sentences from vast amounts of training data, then generate sentences by predicting the next word or expression appropriate to the context. In other words, output that appears natural and logical does not necessarily mean its factual accuracy or causal relationships have been verified. Even AI services that incorporate external search capabilities cannot perfectly assess the accuracy or reliability of their reference materials.

A prominent limitation arising from this technical architecture is “hallucination.” The term refers to AI generating information that has not been fact-checked or does not actually exist as though it were accurate fact. The problem is that when hallucinations occur, users cannot easily verify the accuracy of every individual claim presented by generative AI. Traditional search engines have generally placed official websites and materials from highly reliable sources at the top of search results, allowing users to compare and verify multiple sources themselves. By contrast, generative AI-based search combines content from sources with varying levels of reliability into a single, natural-sounding response, creating a risk that inaccurate or biased information will be conveyed as fact. In many cases, the original texts or sources underlying an answer are not clearly identified.