Customers Ask AI About Suppliers... 'GEO' Becomes a New Challenge for Corporate PR

Hoffman Agency Korea Launches GEO Diagnostic Tool "GEDI"
Asks 12 AI Systems the Same Questions... Compares Competitor Mentions and Cited Sources
Checks Whether AI Accurately Describes Product Specifications and Delivery Track Records

"AI, tell me which suppliers of data center cooling equipment are  available."

If an employee on the purchasing team at a data center construction company asks an artificial intelligence (AI) system this question, the first company to appear could be among the first considered. For companies, it is becoming important to know whether AI presents them as a candidate even when customers do not yet know their name, and whether it accurately describes their products and delivery track record.

Generative engine optimization (GEO) is a technique for reviewing these answers and improving a company’s publicly available materials. While search engine optimization (SEO) focuses on making webpages more visible in search results, GEO also examines which information is cited in answers AI generates by synthesizing multiple sources.

Researchers from Princeton University in the United States and the Indian Institute of Technology Delhi (IIT Delhi), among others, presented a paper systematizing GEO at the 2024 ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD). Adding sources, quotations, and statistics to webpages improved by 30-40% a metric that reflected how much space the materials occupied in AI answers and where they appeared. By contrast, repeatedly inserting keywords, a common SEO practice, had almost no effect. The most effective methods varied depending on the topic of the question.

The first step in applying GEO is to identify questions customers are likely to ask during the purchasing process. When evaluating data center cooling equipment, they might ask about cooling capacity and installation requirements; when narrowing down potential suppliers, they might ask about delivery track records and after-sales support. Since the companies AI mentions and the materials it cites can vary depending on the question, companies need to examine what information is conveyed at each stage of the purchasing process.

Kwon Kijeong, CEO of Hoffman Agency Korea

Hoffman Agency Korea, which specializes in technology company communications, announced on November 11 that it would launch GEDI, a GEO diagnostic tool that automates this review, in Korea. The tool poses the same questions to 12 generative AI systems, including ChatGPT, Gemini, and Claude, and compares the contexts in which the company and its competitors are mentioned, as well as the materials cited by AI.

The company said the tool analyzes company websites and media reports that AI cites as the basis for its answers, down to the sentence and paragraph level. For example, if AI says a discontinued product is still on sale and cites an outdated specification sheet on the company’s website, the company can identify which materials to update first.

Companies can also compare results by changing how questions are phrased for different markets, or track changes in responses by asking the same questions again after a set period. However, AI cites sources when it answers after conducting a web search. Answers based on knowledge learned by a model may not change until its next training, even if the source materials are updated.

Kwon Kijeong, CEO of Hoffman Agency Korea, said, "Understanding the gap between the messages companies want to convey and what generative AI actually says is a new starting point for brand communications in the age of AI."

This content was produced with the assistance of AI translation services.