Quảng cáo sắp xuất hiện trên các chatbot AI — và nó có thể chi phối các câu trả lời bạn nhận được

When Google founders Sergey Brin and Larry Page were still university researchers, they warned that advertising could undermine the trustworthiness of search engines. In their landmark paper presenting Google, they also wrote:

“The goals of the advertising business model do not always correspond to providing quality search to users[…]We expect that advertising–funded search engines will be inherently biased toward the advertisers and away from the needs of the consumers.”

More than two decades later, a similar question is emerging around artificial intelligence. As companies race to commercialize AI systems to offset the amount of cash they are burning through, advertising is becoming an increasingly attractive source of revenue.

But advertising within AI systems presents a new challenge. Unlike online ads, which generally appear alongside information as banners or sponsored links, AI advertising is being integrated into the conversation itself, potentially changing the answers you receive.

AI’s advertising model

Today, around half of consumers report using AI conversations for search. Indeed, conversations with AI are becoming so common that some even develop dependencies with AI companions. As advertising merges with AI, it could reshape these conversations.

OpenAI has introduced ChatGPT ads designed around the way people interact with AI assistants, while Google is testing sponsored answers within Gemini conversations. Both approaches appear to follow closely how advertising technologies currently work, including real–time bidding and micro-targeting users, which allow advertisers to reach individual users based on behavioural data.

Organizational theorist Henry Chesbrough has argued that “technology by itself has no inherent value; that value only arises when it is commercialized through a business model.” His insight suggests that emerging technologies can only be examined as commercial systems shaped by financial incentives.

AI companies say advertising will not change answers. Yet advertising technology is difficult to audit because outsiders often lack access to the data, and AdTech companies weaponize complexity to prevent oversight.

Preventing paid influence from shaping AI-generated answers remains largely voluntary and depends on AI companies honouring their promises.

Some AI providers have already raised concerns about the impact of advertising on trust. Perplexity, an AI-powered answer engine, tested sponsored conversations but discontinued the program.

Even with clear labels, users may struggle to distinguish advertising from organically generated information when both appear within an AI conversation.

New risks in the zero–click internet

What happens when advertising no longer directs traffic to a website, but instead becomes part of a monetized AI conversation?

Commercial AI systems are heading towards a “zero–click internet” in which users make purchases within a corporate–controlled AI environment, not independent websites.

This represents a new form of e-commerce where advertising and transactions are integrated directly into AI chats. As a result, valuable user attention and commercial opportunities remain under the control of AI providers.

Our research explores the implications of merging advertising into AI environments. It identifies risks for free speech and free markets. AI-generated answers compress competing perspectives and may make alternative viewpoints less visible. This could create challenges for minority voices and dissenting perspectives.

It also raises questions about whether large corporations could purchase preferential visibility within AI systems. Smaller firms may face pay–to–play barriers, potentially increasing market concentration.

AI and disinformation

Another concern is the amplification of disinformation through AI. The United Nations warns that embedding micro-targeted adverts into AI-generated content could amplify misinformation and polarizing material.

We know that AI answers can be steered by their controlling corporations. One example is SpaceXAI’s Grok chatbot, which generated false “white genocide” claims in response to unrelated prompts and produced other extremist content reportedly due to an unauthorized code change an employee had made.

Merging advertising with AI adds another layer of risk. Existing digital advertising technologies allow messages to be targeted at specific users based on their personal data.

If similar techniques are integrated into AI conversations, misleading or manipulative messages could be delivered privately, without the public scrutiny that accompanies traditional political advertising.

This is one concern associated with “dark money” in AdTech: when third–party entities use digital advertising technologies for illegitimate purposes, such as election interference.

Why current defences fail

Current defences against disinformation were not designed for AI conversations.

Some of the most advanced legislation to counter disinformation, such as the European Union’s Digital Services Act, was designed for social media and search engines, not chatbots. ChatGPT falls under the act’s scope if it works as a search engine. However, the act’s rules do not guarantee the accuracy of each chat.

The principle responses to disinformation, namely fact–checking and media literacy, were developed in an information environment where misleading claims were public and persistent.

Fact-checkers assess claims with public relevance, such as statements from politicians or hoaxes circulating widely on social media. AI conversations, by contrast, are personalized and ephemeral, meaning misleading claims made to individual users may never enter the public information environment where they can be examined.

Imagine a chatbot repeating a fabricated allegation about a candidate in a private conversation with a voter. The practical challenge is that a personalized falsehood may remain invisible to fact-checkers, leaving the voter misled.

Media literacy faces similar limitations. Users are encouraged to examine sources and compare evidence. However, while search engines presented users with ranked sources, AI systems increasingly provide a single generated response: not simply “an” answer, but “the” answer.

For example, a company may flood the web with fabricated reports. The reports can poison or taint the sources an AI system retrieves. When a consumer asks about the product, the AI turns those planted claims into an authoritative answer, complete with citations.

Setting boundaries

Advertising–funded social media is addictive and often rewards conflict. Because controversy attracts attention, platforms can turn it into engagement metrics that advertisers value. Advertisers are also not limited to commercial brands; anyone willing to pay can attempt to increase their reach.

While policymakers are increasingly focused on the governance of digital platforms, there are still few guidelines for sponsored AI conversations.

Policymakers should establish clear boundaries preventing sponsors from shaping the evidence and substance of AI-generated conversations. A label alone is unlikely to be enough.

AI companies promise that advertising revenue will not influence the substance of their answers. However, such promises may come with a “for now” qualification.

The authors do not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.