ChatGPT Ads Hit $1B Annualized Run Rate: What It Means for China's LLM Monetization and GEO Landscape

As a landmark event in conversational AI monetization, this latest development from overseas can help us better understand the changes happening in the domestic market.

When looking at this data, it's important to first clarify the statistical basis: the $1 billion figure is an annualized run rate, an extrapolated full-year estimate based on current business performance, not actual revenue that has been settled.

In terms of product form, ChatGPT ads are only shown to free and low-tier paid users; paid subscribers do not see ads. Ads appear as standalone cards at the bottom of answers, and official rules clearly state that the ad module does not interfere with the model's generated output.

This overseas pay-per-click advertising model has proven viable, but it's difficult to replicate directly in China. Compared to the overseas focus on conversational ads, leading domestic LLMs are more focused on in-platform transaction loops within conversational scenarios.

Currently, the market has formed two clear development paths:

Doubao, Tongyi Qianwen, and Tencent Yuanbao leverage their group ecosystems to build complete transaction chains; Kimi and DeepSeek have not yet deployed self-operated transaction shelves, relying more on web-wide search data sources for their answers.

Domestic In-Platform Loops: Two Paths in E-commerce and Local Life Scenarios

E-commerce Scenario

Currently, Doubao, Tongyi Qianwen, and Tencent Yuanbao have all established e-commerce transaction chains, while DeepSeek and others have not yet integrated with e-commerce.

Doubao connects to Douyin Mall, Qianwen links with Taobao and Tmall, and Yuanbao integrates with JD.com. When users make shopping requests in conversations, the model prioritizes structured product data from within the ecosystem and outputs product cards that can be directly clicked through.

The triggering of product cards is not simple keyword matching; it depends on the user's purchase intent strength and the completeness of product inventory, price, and other information.

Therefore, for brands to get their products recommended, the key is to fully complete their store and product information, rather than trying to buy placement.

Based on current public information, there is no channel for paid intervention in AI recommendation ranking in e-commerce scenarios. At this stage, the chain mainly focuses on ecosystem redirects and displays. Some platforms have piloted transaction service fees for AI channels in certain categories, but the pace of implementation varies.

Local Life Scenario

Local life is the scenario where domestic AI transaction loops are progressing fastest.

Doubao integrates with Douyin Laike, Tongyi Qianwen connects with Fliggy, and Tencent Yuanbao integrates with OTA platforms like Tongcheng Travel and Ctrip. When users inquire about hotels or vacations, the AI outputs corresponding store cards that can be clicked through to complete bookings.

Among these, Doubao is currently the only platform in China that has publicly set independent transaction commission rules for AI conversation channels: starting August 10, for hotel orders completed through Doubao's entry point to Douyin Laike, the platform charges approximately 12% in comprehensive channel service fees (11.4% software service fee + 0.6% payment processing fee). Impressions are not charged; fees are only deducted after a transaction.

Qianwen's integration with Fliggy and Yuanbao's integration with OTA platforms like Tongcheng Travel have not yet publicly announced separate commission rules for AI conversation channels; transactions are still settled under the original platform commission systems.

In contrast, models like DeepSeek, which primarily rely on web-wide search, have not reached official transaction partnerships with OTA platforms. After users ask questions, they are redirected to third-party platforms like Ctrip to complete bookings, and the platform does not participate in fulfillment or commission.

Whether closed-loop or search-based, the common boundary is consistent: Payment cannot buy AI recommendation priority; whether a store enters the recommendation candidate pool depends on the completeness of its information onboarding.

In-Platform Loops Do Not Equal Paying for AI Recommendations

There is a common misconception in the market: since large models now support in-conversation product cards and transaction conversions, brands can pay to get AI priority recommendations.

Based on existing platform rules, whether in e-commerce or local life scenarios, payment cannot change the model's recommendation weight.

For brands to qualify as candidates, they must fully complete information onboarding in the corresponding ecosystem. Under the CPS commission model represented by Doubao, commercial costs are only incurred at the transaction stage.

Here, it's helpful to clarify the three models, which are often confused in industry content:

• ChatGPT Ads Overseas Ad Cards: Pre-paid purchase of exposure slots, an independent add-on module that does not alter the model's main text. China has not yet launched a similar self-service bidding ad backend.

• Domestic In-Platform Transaction Loops (represented by CPS commission): Candidates qualify by completing information; commissions are generated after transactions, and payment does not interfere with recommendation results.

• GEO (Generative Engine Optimization): Does not purchase ads or participate in transaction commissions; by polishing web-wide sources, it encourages large models to retrieve and cite the brand in native Q&A responses.

This is easy to understand in the context of cultural tourism attractions:

When users search for regional attraction recommendations, if the brand has completed POI and package information onboarding in the closed-loop ecosystem, it has the opportunity to display ticket transaction cards, enabling conversion paths.

When users care about play experience, project suitability, or pros and cons comparisons, the model retrieves public web-wide sources—this is the mental scenario where GEO delivers value.

Transaction cards handle direct conversions; GEO influences brand trust during the user decision-making stage. They are complementary, not substitutes.

Practical Implications for Brand Practitioners

Brands across consumer electronics, automotive, FMCG, beauty, services, hospitality, and cultural tourism are affected differently by closed loops, but the underlying signals from industry changes are common.

First, the barriers to AI presence have shifted; you can't just focus on web-wide sources.

Traditional GEO focuses more on polishing web-wide sources, including authoritative media reports, authoritative testing and certifications, encyclopedia content, Q&A platform content, professional reviews, and genuine user experiences and recommendations across public internet information.

As major tech ecosystems mature, completing in-platform sources has become a prerequisite for visibility in high-purchase-intent scenarios. If product parameters, brand profiles, and POI data are not fully completed in the corresponding ecosystems, even the most polished web-wide sources will find it increasingly difficult to get card-type exposure in consumption-oriented queries.

But this does not mean the value of web-wide sources is diminished.

Once users enter the comparison, review, or brand research stage, models still heavily cite public web-wide sources. High-credibility content such as authoritative media, testing certifications, and professional reviews still carry significant weight in large models' citation ranking. For products like Kimi and DeepSeek, which primarily rely on web-wide search, traditional GEO logic remains fully effective.

Second, rationally assess the possibility of domestic conversational ads in the future.

Due to computing cost pressures, there is a possibility that China may launch similar ChatGPT-style conversational ad cards in the future.

But given the existing ecosystem landscape, ads are more likely to be a supplementary monetization method: ads would be strictly separated as standalone cards from native answers, for example, initially available to free users, with members likely retaining ad-free benefits; transaction loop monetization remains the core direction for major tech companies.

Final Thoughts

ChatGPT ads reaching a $1 billion annualized run rate is a valuable overseas industry signal, but it cannot be directly used to extrapolate the domestic market's trajectory.

The main line of domestic large model monetization is in-platform transaction loops, not general conversational pay-per-click ads. For all brands, AI presence has entered a stage where both types of sources are equally important: on one hand, complete in-platform sources across relevant ecosystems; on the other, continuously polish web-wide sources.

Regardless of how future monetization models evolve, these two types of sources are the foundational foundation for brands in the AI environment.