Generative artificial intelligence tools face intense financial pressures as OpenAI and Google test advertisements beneath chatbot responses to offset massive server and energy expenses, balancing user retention against commercial monetization.
Building and operating large language models consumes enormous amounts of electricity and capital equipment, creating a stark economic reality for developers. Software and internet services historically scaled with near-zero marginal costs for each additional user, but generative AI incurs distinct compute costs for practically every single query processed by the underlying hardware infrastructure.
Most consumers want to use AI daily, but only a smaller subset is willing to pay enough continuously to cover those operational expenses, Karoline Ströhlein, a behavioral economist at the University of Regensburg, told faz.net.
OpenAI started testing advertisements in the United States in early February, generating an annualized revenue run-rate of one billion US dollars in less than 200 days, according to David Dugan, who leads the advertising business at OpenAI, speaking to faz.net.
OpenAI Expands Ad Tests to Germany
OpenAI has since expanded advertisement tests to numerous international markets including Germany, placing sponsored links directly beneath ChatGPT responses with clear labels. Dugan told faz.net that no personal chat logs, memories, or user data are shared with advertisers, and queries concerning political topics or personal and psychological health remain entirely ad-free.
Personalized advertising is not initially deployed within European Union markets. In the United States, however, enabled personalized advertising allows ChatGPT to utilize specific signals from general user behavior to display targeted ads, though OpenAI has declined to disclose the exact nature of those behavioral signals when asked.
Alphabet subsidiary Google maintains a strict separation between paid search options and organic web results, a principle Dan Taylor, vice president of ads at Google, told faz.net also applies to AI-driven searches. Google is currently testing advertisements inside AI modes and AI overviews in the United States, while German users view sponsored links exclusively alongside the AI overview, and Gemini mobile chatbot applications remain free of advertisements for now.
Google relies on advertising for roughly three quarters of its 403 billion US dollar annual revenue, maintaining an approximately 90 percent search engine market share. Taylor told faz.net that AI integration has triggered a growth surge by prompting users to submit longer inquiries in conversational tones and utilize smartphone cameras.
Controlled Tests Reveal Sponsoring Bias
Controlled experiments conducted by researchers evaluated how language models respond to sponsored flight options, testing whether financial incentives sway algorithmic recommendations. In a simulated environment, models chose between a cheaper unsponsored flight and a more expensive sponsored alternative.
| AI Model | Developer | Sponsored Flight Recommendation Rate |
|---|---|---|
| Grok-4.1 Fast | SpaceX | 83 percent |
| Qwen-3 Next | Alibaba | 70 percent |
| GPT-5.1 | OpenAI | 50 percent |
| Gemini 3 Pro | 37 percent | |
| Claude 4.5 Opus | Anthropic | 28 percent |
Eighteen out of twenty-three tested models recommended the sponsored option in more than half of all evaluated test cases, according to findings reported by faz.net. Elon Musk’s SpaceX model Grok-4.1 Fast and the Chinese model Qwen-3 Next recorded the highest sponsorship preference rates at 83 percent and 70 percent respectively.
Lower recommendation rates for sponsored options appeared in Gemini 3 Pro at 37 percent and Claude 4.5 Opus at 28 percent, prompting researchers to note a decision tendency favoring users. Separate research observed that lower-income users encountered advertisements from ChatGPT more frequently, while other models altered recommendations based on estimated socio-economic status.
Chatbots frequently displayed sponsored alternatives even when users explicitly requested bookings through non-sponsored providers. While models maintained accurate details regarding flight durations and pricing, researchers identified transparency concerns because systems often omitted the sponsorship status more frequently than the underlying price.
OpenAI stated on inquiry that because they do not know in detail how these independent studies were designed or operated, they could not comment on the findings. Google offered a similar response, stating that the study relied on a hypothetical experiment that fails to reflect how their products actually function.
Ströhlein told faz.net that structural product incentives present underlying risks because advertising creates an inherent conflict between optimizing for user benefit and maximizing financial returns for platforms and advertisers. Traditional search engines present clear visual boundaries between sponsored links and organic results, whereas conversational chatbots simulate personal consultations that can obscure commercial intent.
Consumer adoption metrics show no immediate negative impact from monetization efforts so far. Dugan told faz.net that user satisfaction and behavioral patterns have remained stable across markets, pointing to a 50 percent year-over-year increase in weekly active users in Germany, where 90 percent of individuals aged 18 to 34 now utilize the application at least once per week.
When asked whether commercial partners could ever receive preferential treatment inside AI responses through advertising agreements, Dugan told faz.net that he could not make predictions for the future.
- Health Experts Say Holiday Travelers Should Consider Wearing High-Filtration Masks
- WhatsApp Patches Actively Exploited Android IP Leak Flaw in ‘Click to WhatsApp’ Ads
- Inflation Since October 2024 Adds About $4,720 a Year to Household Costs (daybreakwire.com)
- Foo Fighters to Close Take Cover Tour at Croke Park September 2027 (archyworldys.com)