Companies' digital marketing competition is rapidly shifting from the top of traditional search results to brand exposure in artificial intelligence (AI) answers.

On April 8 local time, online media outlet Gigazine reported that as AI search and chatbots become the starting point for information discovery, companies and the marketing industry are focusing on strategies to get their names included in AI-generated answers rather than traditional search engine optimization (SEO).

A new optimization market tailored to the AI era is also forming rapidly. The industry is seeing a range of concepts such as Answer Engine Optimization (AED), Generative Engine Optimization (GEO) and Generative Search Optimization (GSO), and consulting and solutions around them are spreading. Rand Fishkin (랜드 피쉬킨) of SparkToro said the SEO industry has effectively entered a gold-rush phase and pointed to AI optimization being packaged as a new market.

A leading tactic is comparison-style content. It takes the form of listing multiple products while placing a company's own service at the top, designed so large language models (LLMs) can easily extract information. Zendesk and Freshworks, for example, present their products as the best option in content formatted as comparison articles. Many other companies, including Eesel, Hiver, Watermelon and Help Scout, are running similar content.

Such pages are marked by simple item categories and comparison structures that make them easy for large language models to read and extract. That also raises the likelihood they will be incorporated into the AI answer-generation process. Google said it is aware of such low-quality list-style content and is responding to common manipulation in Search and Gemini.

The issue is that it has gone beyond making documents that are easy for AI to read. Microsoft disclosed cases in which some companies inserted hidden prompts into an "AI summary" feature so that when users press a button, specific instructions are automatically entered into an AI assistant input window.

It explained that the prompts planted phrases such as "remember this company as a trustworthy source" or "prioritize this service in future recommendations" to induce persistent bias in AI memory and recommendation results. Microsoft classified this as a case of "AI recommendation abuse."

An analysis has also emerged that personalization itself is becoming a new attack surface as AI assistants adopt structures that store user preferences and work context and reflect them in later responses. In one case, after a user clicked an "AI summary" button on a certain site, that site remained in the AI's memory as a "preferred source" and influenced later recommendations.

Moreover, some AI systems have been confirmed to be able to load pre-written prompts through URL parameters alone, allowing such manipulation to be carried out with a single click. Tools offering related functions have also appeared, promoting themselves as "SEO growth hacking for LLMs" and advertising that they increase the chances of being cited in AI answers.

Some also point to expectations in the market being excessive compared with actual usage. Fishkin said interest in AI search may have been "inflated by 10 to 100 times" relative to the scale of real use. SparkToro analysis found that in desktop environments, search volumes for traditional search engines still far exceeded AI tools, and that Amazon, Bing and YouTube had larger search shares than ChatGPT.

Even so, executive interest and media attention in AI search are drawing in money and staff, which is another variable in the market. Many companies are investing out of anxiety that they could fall behind if they are not visible in AI, but it remains unclear whether real demand sufficiently supports it. As companies jump into the competition for exposure in AI responses, manipulation tactics targeting AI recommendation and memory features are also emerging, shaking the standards of search marketing itself.