Search Ad Revenue Grew 107 Percent. Search Engines Stood Still.

🔢 A number that reads one way and means another

A Chinese industry report puts growth in "in-platform and AI search" advertising revenue at 107.4 percent for 2025. Read alone, it says search advertising is booming.

The same report puts traditional search engine advertising revenue at 47.95 billion yuan for the year, described as broadly flat against 2024. It also notes that search advertising's share of the total market recovered to 9 percent — the level it held in 2022.

Put those three sentences together and the picture changes. Search advertising did not grow as a category. It gave up share for two years, and in 2025 it clawed that share back by moving to a different kind of search. The part that stayed still is the part most people mean when they say search advertising.

The source is the China Internet Marketing Development Report 2026, published on 23 September by the Zhongguancun Interactive Marketing Lab with the Miaozhen Marketing Science Institute, Beijing Normal University's School of Journalism and Communication, and the Media Economics and Management Committee of the Chinese Journalism History Society. It is a blue book from the People's Daily Press, edited by Yu Guoming and Chen Yong.

This is not breaking news. It is a five-week-old report that has had less attention than it deserves, and the numbers in it bear directly on how a foreign brand should be budgeting for search in China.

📏 What the report actually measured

Four figures set the frame, and all four are consistent across the report's coverage.

Total internet advertising in 2025 was about 725.7 billion yuan, up 11.50 percent. Internet marketing, a wider measure that includes e-commerce promotion and social marketing, was about 881.8 billion yuan, up 13.56 percent. Together the two came to about 1.6 trillion yuan, up 12.62 percent. The report projects 2026 at close to or above 800 billion yuan for advertising and marketing, growing at 10 to 12 percent.

Inside that, three structural facts matter more than the totals.

E-commerce platform advertising reached 279.75 billion yuan, 38.55 percent of the market. Short-video platform advertising reached 143.21 billion yuan, up 14.46 percent, and overtook comprehensive e-commerce for the first time to become the largest single media platform type by revenue. Pure display advertising, the banner, fell to 8 percent of the market.

And then the search split described above: traditional search engines flat at 47.95 billion yuan, in-platform and AI search up 107.4 percent, the category's share recovering to 9 percent.

↔️ Search grew. Search engines did not.

The distinction is not semantic. It separates two different purchases.

A search engine ad buys a position against a query a person types into a general-purpose search box. An in-platform search ad buys a position against a query typed inside a commerce or content platform. An AI search ad buys an appearance inside an answer rather than a list of links.

Those are three different inventories, sold by different companies, optimised by different logic. In 2025 in China, the first was flat and the second and third grew fast enough to lift the whole category's share off a two-year low.

The report's own framing is that the structure of the search advertising market is being rebuilt. That is a stronger claim than a growth statistic, and the flat 47.95 billion is what makes it credible: you cannot rebuild a market that is expanding everywhere.

🧭 Where the growth went: inside the platforms, and into the assistant

Two mechanisms are documented, and they are not the same trend.

The first is in-platform search. The report attributes e-commerce advertising growth mainly to interest-based commerce, which grew 18.9 percent, while comprehensive e-commerce was described as little changed over three years. The distinction matters: the growth is in commerce that generates demand through content and then captures it in search, not in commerce that waits for a shopper to arrive with a query.

The second is AI search, and here the report's language shifts from measurement to behaviour. It describes a consumer decision path moving from passive searching to actively trusting an AI recommendation, and says competition stops being about keyword ranking and becomes about structuring information so that an AI can identify it, understand it and judge it valuable. That is the definition of generative engine optimisation, or GEO.

Two supporting facts from the same reporting round. A CTR survey cited in the report's coverage found 31 percent of advertisers planning to increase spend on AI conversational search advertising in 2025. And Tan Beiping of the Miaozhen Marketing Science Institute said at the launch that a group standard for GEO has already been published under the China Advertising Association framework.

A group standard existing is a different signal from a vendor pitching a service. It means the practice has a written definition that buyers and sellers can point at.

💬 The new condition is being quotable, not being ranked

The most useful contribution in the report is not a number. It is a description of what has changed about being visible.

Li Biao of Renmin University's School of Journalism argued at the launch that as AI agents mediate consumer decisions, the B2C path is becoming B2A2C — the agent becomes a new intermediary between brand and buyer. Under that structure, the threshold for competing stops being only whether you can be seen. It also includes whether a large model can understand you consistently, identify you accurately and cite you.

Yu Guoming, the report's editor, framed the same shift as three eras of brand equity: the product brand of the industrial era, the traffic brand of the platform-algorithm era, and now the mind brand of the human-machine era. He described AI as a second author of brand narrative, and argued that brand management has expanded from managing human cognition to managing human cognition and machine comprehension at the same time.

For a practitioner, the operational reading is plain. A keyword map optimises for a ranking that a person sees. A structured claim about what your product is, who it is for and what it does not do optimises for a summary that a model writes. The second is a different artefact with different failure modes, and most accounts do not have one.

🎯 What this changes for an overseas brand's search budget

Three adjustments, in order of how soon they bite.

Separate the two searches in your plan and in your reporting. A budget line called "search in China" now mixes a flat inventory and a growing one. If the plan does not distinguish a general search engine from in-platform and AI surfaces, the performance of the growing part will be read as the performance of the whole, or the reverse.

Assume the assistant is a distribution channel with its own requirements. If 31 percent of advertisers are increasing spend on AI conversational search, the constraint is not budget, it is whether there is anything to cite. Descriptions, specifications, comparison tables and explicit scope statements are the raw material. Marketing copy written for a human reader often contains none of it.

Do not read a 107 percent growth rate as a market size. The report does not publish the absolute base for the in-platform and AI search category in the coverage available. A growth rate on a small base can be large and still small in absolute terms, and the 9 percent category share is a reminder of that: the whole search category, both halves, is under a tenth of the market.

🏢 Where Baidu PPC Pro comes in

The China side of this has four prerequisites, and none of them changes with a market report.

An account without a Chinese entity. Baidu requires a Chinese business licence to open an advertising account. Most overseas brands do not hold one, and obtaining one takes months. Baidu PPC Pro opens and manages the account without it, which decides whether a brand can be in a Chinese auction at all.

A keyword map built from local terms. Chinese consumers search the local term for the category, the local term for the feature and the local phrase for the problem they want solved. A translated keyword list reaches people and then fails to match them.

Creative that clears review the first time. Advertising in China is filed and reviewed per industry, with claims checked against permitted phrasing. A creative that performs in another market can be rejected here on submission, usually with no advance warning.

Landing pages that answer the search. Chinese-language pages that resolve the comparison the user just typed, and that state plainly what the product is and is not — the same material an assistant would need to cite.

📅 What to watch, and notes on the source

Four things would settle what is still open.

1. The absolute size of the in-platform and AI search category. Without it, 107.4 percent is a direction, not a quantity.
2. Whether the 9 percent category share holds in 2026. A recovery that reverses would say the 2025 figure was a one-off reallocation rather than a new baseline.
3. Whether advertisers report the two searches separately in 2026 budget surveys. That is the point at which the split becomes visible in planning rather than only in retrospective data.
4. What the GEO group standard actually specifies. A standard that defines citation-worthiness would give the practice something testable.

Notes on the source:

• The report is the China Internet Marketing Development Report 2026, subtitled Decoding a New Map of Chinese Internet Marketing. It was released on 23 September 2026 by the Zhongguancun Interactive Marketing Lab, the Miaozhen Marketing Science Institute, Beijing Normal University's School of Journalism and Communication, and the Media Economics and Management Committee of the Chinese Journalism History Society. It is published as a blue book by People's Daily Press.
• Totals, the e-commerce and short-video figures, the 8 percent display share, the 107.4 percent figure and the 9 percent category share are as reported by China Press and Publishing Journal on 29 September 2026, and cross-checked against coverage by CTR and China Financial Information Network.
• The 47.95 billion yuan figure for traditional search engine advertising, and the statement that it was broadly flat, appear in secondary summaries of the report rather than in the coverage cited above. They are used here because several independent summaries agree on the figure, but they rest on one line of transmission and should be read as such.
• The report coverage available does not publish the absolute base for the in-platform and AI search category. No absolute value is asserted here.
• Per-platform revenue growth figures for individual companies that circulate in third-party summaries were not used, because those summaries do not state the methodology or the originating dataset.
• The 31 percent advertiser figure for AI conversational search comes from a CTR survey cited in report coverage; the GEO group standard statement is from Tan Beiping's remarks at the launch. Neither is an audited figure.
• The 50 percent of China 500 companies using AI agents for marketing decisions by 2026, and 15 percent of daily work decisions taken autonomously by 2028, are the report's projections and are not used as current facts here.
• No effect on cost, approval rates or campaign performance is quantified, because none has been published.