"Will AI large - scale models replace traditional search engines?"

When I posed this question to ChatGPT, its response was measured: This is an ongoing debate that has yet to reach a definitive conclusion. In the short term, AI large - scale models won't entirely supplant traditional search engines; instead, they'll "consume half and transform half" of the search market.

Search engines are far from obsolete. However, user demands are undergoing a seismic shift, compelling the transformation of search boxes into intelligent hubs. This is also the transformation path that both Google and Baidu are pursuing.

Baidu has been particularly proactive. It's directly leveraging AI to revamp search result pages, transforming search into an AI - driven application dominated by rich media content like images and videos, rather than just incorporating AI - generated summaries.

While the new search paradigm offers an enhanced user experience, it also poses significant challenges to Baidu's monetization engine, Baidu Marketing. How can AI - native marketing be effectively monetized? How can customer return on investment (ROI) be increased while simultaneously elevating the user experience? How can we harness the power of "AI - native" to unlock new business values and inject long - term growth momentum?

Recently, during in - depth discussions with the Baidu AI Marketing team, NoNoise uncovered preliminary answers to these questions. After three years of exploration, Baidu AI Marketing has transitioned from the "broad testing" phase to a stage of large - scale effectiveness.

01

Bridging the Information Gap

How can marketing precisely align with user demands?

One of the most significant shifts in the marketing landscape over the past two years stems from the transformation of user demands.

Traditional search advertising focuses on maximizing the alignment between customer selling points and user demands centered around a single search query through an efficient monetization engine. Its primary goal is to optimize the monetization efficiency of top - ranking positions.

With the widespread adoption of large - scale model technology and the rise of chatbot products, the barriers to accessing information for users have significantly diminished, and information gaps are rapidly closing. Simultaneously, user demands are becoming increasingly niche and complex.

In essence, user demands have become more diverse and exacting.

This implies that the traditional search advertising model, which relies on keywords and static landing pages, is facing challenges. Precisely meeting user demands is becoming an increasingly daunting task.

How to achieve more seamless commercial matching and monetization based on "demands" has become the core challenge for Baidu AI Marketing.

To adapt to the evolving user demands, Baidu Marketing has developed two native monetization paths in the AI context:

The first path is e - commerce recommendations. AI detects potential purchase intentions of users during chats and recommends relevant e - commerce products along with links.

This approach is not exclusive to Baidu. In November of this year, Morgan Stanley analysts highlighted the prospects of "agent - based e - commerce" in their latest research report. They predicted that it would reach a gross merchandise volume (GMV) of $385 billion by 2030, accounting for approximately 20% of the total US e - commerce market. Moreover, current data indicates that general large - scale model platforms (such as ChatGPT) have a much higher shopping penetration rate compared to AI developed by retailers. Users are not just engaging in "chatting" but also "buying" - about 30 - 40% of AI users have made purchases through platform recommendations.

During the discussion, Zhang Lihong, the head of Baidu's commercial products, revealed that the GMV growth of e - commerce product recommendations within Baidu's AI ecosystem has been "extremely rapid." For instance, when a user expresses an interest in buying a specific laptop, the chatbot can generate a list of recommendations and even include promotional offers.

The second path revolves around agents. For example, when a user needs to compare postgraduate entrance exam preparation institutions near Shangdi, Beijing, the large - scale model summary may first present several criteria for selecting such institutions, such as faculty strength, enrollment rates, and fees. It then attaches relevant recommendations, such as four or five institutions with high comprehensive rankings in the vicinity.

When users click on these recommendations, they discover that the "institutions" are no longer cold, static web pages but merchant agents. These merchant agents can respond efficiently to the user's refined demands, providing more transparent information, quotes, and even emotional value.

Under this model, users can access information and services more conveniently. The starting point of marketing has been completely reconstructed - everything revolves around user demands, and opportunities for user conversion are sought based on meeting those demands. The business value is also amplified. Recommending four or five postgraduate exam preparation institutions means providing users with sufficient choices while achieving business value conversion through multi - point touchpoints. As a result, customer ROI has increased. The lead opening rate of human customer service is 5 - 6%, the number of merchant agents has doubled, and agents can provide 24/7 uninterrupted service.

Within Baidu, merchant agents, along with digital human live streaming, are considered the "new infrastructure" of AI marketing. Baidu's Q3 2025 financial report shows that AI - native marketing service revenue (primarily agents + digital humans) increased by 262% year - on - year, indicating significant business potential.

According to Zhang Lihong's recollection, when merchant agents were first introduced in 2023, few clients were willing to give them a try. AI customer service was even derogatorily referred to as "AI idiocy" by some clients due to its verbosity and tendency to generate hallucinations. After two years of evolution, the data trends for merchant agents, from the C - end user experience to the B - end conversion, have been remarkable. They now cover over 30 industries with an increasingly diverse client base.

For example, Feizhuo Technology, a B2B intelligent sensor company, previously faced limited customer service personnel and was unable to promptly respond to user inquiries during nighttime and non - working hours. After integrating a voice - based merchant agent, the average daily phone lead volume increased by 30%.

As of now, Baidu's merchant agents serve an average of 33,000 clients daily, indicating that this monetization product is entering a stage of large - scale application.

These positive cycles have also convinced Baidu that, after transitioning from broad testing to focused validation, this definitive path is worth continuing to invest in and expand for large - scale implementation.

02

Does Baidu AI Marketing Still Have a "Comfort Zone"?

The impact of AI large - scale models on traditional search engines is profound. From Baidu's series of strategic responses for its monetization engine, its approach can be summarized as leveraging strengths and compensating for weaknesses.

If reconstructing commercial products in the new search field and enhancing information flow ad presentation are considered as compensating for weaknesses, then deepening its presence in vertical industries is equivalent to solidifying Baidu AI Marketing's comfort zone.

The person in charge of Baidu AI Marketing believes that after large - scale models have brought about information equality, a large number of knowledge - based and general demands can be adequately met by almost every media platform. Traditional search engines' advantage in general knowledge has been "neutralized." However, from a commercialization perspective, this portion of traffic also has lower business value. "Now, it seems that among the traffic with higher commercialization value, our advantage still persists."

The basis for this lies in Baidu's ability to access more industry - specific know - how, namely the specialized aspects of vertical industries, which general large - scale models are not as proficient in. After years of accumulation, Baidu Marketing has developed vertical expertise in scenarios such as healthcare, education, local life, and traditional search - dominated B2B production, manufacturing, agriculture, forestry, animal husbandry, fisheries, energy, and chemical industries. This includes parameters of B2B production and manufacturing - related models, offline knowledge bases in the local life sector, and customer psychology. How can these be structured to precisely match user demands with customer demands? All of these have opened up more possibilities for Baidu to deepen its presence in vertical industries.

For example, the dialogue strategy of language agents can truly align with the business models of B2B clients. As mentioned earlier, for the intelligent sensor company, the agent created by Baidu Marketing can accurately identify professional terms such as "liquid level sensor model" and "radar liquid level gauge" and even determine whether maintenance demands fall within the business scope, achieving effective conversion while meeting user demands.

An AI that has a deeper understanding of industries can naturally bring more certain high - quality growth to enterprises.

The aforementioned person in charge revealed that education, real estate, local life, and legal services are industries that have significantly benefited from AI - native marketing this year.

While AI - native marketing is reconstructing the entire marketing chain, AI has learned to answer phone calls, conduct lead live streaming, and negotiate business. Currently, the large - scale implementation of these AI capabilities not only brings Baidu into the "value realization period" of AI but also further broadens the space for enterprise operational growth.