Publishers Shift Strategy in AI Era

Note held a media study session in Tokyo on July 21, where CEO Sadaki Kato and Takayuki Fukatsu, who became chief strategy officer in July, discussed how the spread of generative AI is changing the distribution of information and creating difficult choices for companies seeking both to protect and promote their content.

"Content that cannot be read by AI is, from the AI's perspective, the same as content that does not exist," the executives said, warning that businesses and media companies could lose traffic if their information cannot be accessed or referenced by AI services.

Kato and Fukatsu said companies need to discuss more seriously whether data should be kept closed or made accessible in the AI era, including questions surrounding paywalls. They argued that industries need to develop best practices rather than leaving individual companies to address the issue separately.

One of the biggest changes is the shift from traditional search engine optimization toward what is sometimes called AIO, GEO or LLMO, in which companies seek to ensure that their products, services and information are understood and recommended by generative AI.

As more people ask AI systems questions instead of conducting conventional internet searches, companies may increasingly compete not for the highest position in search rankings but for favorable inclusion in AI-generated answers.

The executives offered a simple example: if a consumer asks an AI system which laundry detergent to buy and the system consistently recommends only Company A's product, rival companies B, C and D face a serious competitive problem.

The same change is already appearing in employment decisions. Kato said he had recently spoken with someone in the recruitment industry who told him that job seekers are increasingly asking AI whether they should move to a particular company.

If the company does not appear in AI results, or if the information available about it produces an unfavorable assessment, that could influence a person's decision, the executives said.

For companies, this means ensuring that AI systems can properly understand their products and services. For media companies, however, the situation is more complicated because allowing AI unrestricted access to content could undermine businesses that depend on selling that material.

Note has about 80 million pieces of content on its platform, with slightly more than 20% behind paywalls. Paid material can generally be discovered above the paywall but cannot be searched or accessed in full beneath it.

The company believes the ideal system would allow AI to recognize and understand the existence of paid content without using it for unrestricted model training, while showing enough information to interest users without revealing the substance of the paid material.

AI could then potentially guide a user from a recommendation to the point of purchase, the executives said.

Note has been discussing whether such arrangements could be developed in cooperation with AI companies, potentially through broader negotiations involving platform and publishing companies.

The alternative carries risks. Some publishers and newspapers have responded to generative AI by blocking AI crawlers entirely to protect their material.

But if an AI system cannot access or reference a company's website, it may have no knowledge of the content and therefore no reason to recommend it. That could eventually reduce traffic and make paid content harder to sell.

Note itself is moving aggressively in the opposite direction internally, seeking to apply AI throughout its operations.

The company has AI transcribe all management meetings and review the discussions, effectively allowing the technology to evaluate the company's senior executives.

Note is also working to feed a wide range of corporate data into AI systems so they can identify new developments, detect unusual spikes in figures, flag potential risks and suggest areas that managers should examine.

The executives described the system as being similar to having an exceptionally intelligent outside director constantly reviewing the company and prompting management to take necessary action.

Note is also attempting to log virtually all internal activity.

Meetings are conducted through Google Meet and transcribed, with records shared across the company. AI systems are being developed to review the progress of individual departments against their goals and other performance indicators.

Information that does not exist digitally is captured through daily reports submitted by employees. Those reports are intended to bring offline activity and other missing context into the company's information systems.

AI can then produce summaries and reviews for individual teams, which can in turn be reviewed at higher levels of management, creating a structure in which information flows upward through the organization.

Fukatsu also argued that efforts to manipulate AI recommendations through superficial technical tricks are likely to fail.

He cited a form of LLMO manipulation in which websites hide instructions in invisible text, such as placing white text on a white background telling an AI system that the page is excellent and should be rated more highly.

Such techniques may initially appear clever, but Fukatsu described them as "extremely foolish."

Unlike earlier search engines that processed webpages largely through mechanical signals, large language models can evaluate and reason about content to some degree. A website that openly tries to manipulate an AI system risks effectively announcing that it is attempting to deceive the technology.

Fukatsu said techniques that once formed part of the tactical side of SEO could therefore become counterproductive in an AI-driven information environment.

Instead, he said companies should abandon superficial tricks and concentrate on improving the fundamental value of their webpages, books, products and other content.

The key will be whether AI systems recognize that a site is providing, in a straightforward manner, the information that AI itself needs and that users genuinely want.

SEO gradually moved in that direction over roughly two decades, but Fukatsu expects the transition to happen far more rapidly with AI.

The executives also expect the way people use smartphones and the internet to change substantially.

Rather than opening individual applications or typing traditional search queries, users may increasingly enter questions directly into an AI interface. Kato said that even when he wants to know something as simple as the opening hours of a Tokyo passport office, he now asks AI rather than visiting the Tokyo Metropolitan Government's website.

In the short term, smartphone search boxes may increasingly become input fields for questions to AI.

Over the longer term, users may not even need to ask questions themselves.

AI systems connected to calendars, documents and other personal services could detect a user's situation automatically and provide information before it is requested, such as warning that a meeting begins in 30 minutes, recommending a document to read and preparing profiles of the participants.

That would make AI the primary interface through which people interact with information, increasing the importance of ensuring that content circulates properly within AI systems.

The central challenge for the next several years, the executives said, will be finding ways to connect content safely to AI while protecting its value and ensuring it can still be discovered and recommended.

Source: 産経ニュース