Dialogue with Wancheng Cloud Commerce: Publishing articles does not equal GEO optimization. Large models don't just push whatever they're fed.

"Help me put together a set of clothes suitable for a seaside vacation."

"Recommend a Western restaurant suitable for a date for me."

"With a budget of 3,000 yuan, which floor cleaning robot is the most cost - effective to buy?"

This kind of Q&A with AI is happening to hundreds of millions of users around the world. In the past, people would input these questions into search engines such as Baidu and Google, which gave rise to SEO (Search Engine Optimization) marketing. When the traffic entrance shifts from search engines to AI tools, GEO (Generative Engine Optimization) has also emerged.

Different from the "search result ranking" pursued by traditional SEO, the core goal of GEO is to make brand information be preferentially cited and recommended in the answers generated by AI. In an era where AI fully penetrates into users' information acquisition and consumption decision - making scenarios, for overseas - going enterprises, doing well in GEO means seizing a new traffic entrance.

What are the fundamental differences between the logic of GEO and SEO? Why do some people say that GEO is a false proposition? How should overseas - going enterprises layout GEO? How can small and medium - sized enterprises start with low costs... With these questions that overseas - going enterprises are concerned about, we had a special conversation with Tan Li, the general manager of the brand overseas - going division of Wancheng Cloud Commerce, to analyze how to make enterprises be seen, trusted, and recommended to users by AI in the era of AI search.

36Kr: GEO is called the evolution of SEO in the AI era. Compared with the core of SEO, which lies in the ranking of search results, what is the key to GEO optimization? What are the core differences between it and traditional SEO?

Tan Li: The core difference is that traditional SEO is more about "rule - based competition", while GEO is more like "trust - based competition". In the past, when doing SEO, our goal was to rank the page on the first page of Google so that users would click in; but now many users directly ask questions in AI and don't even click on the links. So what you need to solve is - how to make AI "cite you" or even "recommend you" in the answer.

So the key to GEO is, first of all, whether you are a "trusted source" (whether anyone in the whole network recognizes you); secondly, whether your content can be understood, disassembled, and reused by AI; finally, whether your content has a complete evidence chain, such as data, cases, or third - party endorsements.

36Kr: After GEO, some people also put forward the concept of AEO. What are the differences between the two?

Tan Li: Many people mix up these two concepts, but I personally distinguish them like this: AEO is more about "optimizing Q&A results", and its core is to make your content directly become the standard answer to a certain question, such as FAQs and selected summaries. GEO is broader. It faces generative AI, not just answering a question, but participating in the "output after multi - source information integration". So AEO is more about "structural skills", while GEO is more about "the overall brand and content ecosystem".

36Kr: The concept of GEO has sparked extensive discussions in the past two years. Why do enterprises need to pay attention to GEO at this time? In the context of the continuous increase in the penetration rate of AI search, what strategic position should GEO occupy in the enterprise marketing system?

Tan Li: Let me share some very real changes: Now users are getting more and more used to asking AI directly instead of screening search results by themselves; and AI has begun to replace the "information collection stage" at the front - end of decision - making in many industries; also, some customers may form a judgment about you in AI before even opening your official website. So enterprises can't ignore GEO. The reality is right here. I think GEO occupies a very crucial position in the enterprise marketing system. It won't replace SEO, but is a "second traffic entrance" parallel to SEO, and may even become one of the main entrances in the future.

36Kr: How to understand that some enterprises think GEO is a false proposition after some attempts, or that they can continuously feed language materials to the large model by themselves?

Tan Li: I think this is a bit of a misunderstanding of the working mechanism of AI. The large model doesn't "recommend whatever you feed it". It has several screening logics: First, whether the information is consistent (for example, whether it can be confirmed from different sources); second, whether the source is trustworthy (whether it is a long - existing content asset); third, whether it is mentioned on multiple platforms (rather than from a single source). So when you "feed language materials" by yourself, in essence, it is just a single - point information input, and it is difficult to form a "trust network".

However, I also understand why some people say that GEO is a false proposition, because there are indeed many "false practices", such as only doing in - station content, only piling up AI articles, and not doing external verification. These basically hardly work.

36Kr: When enterprises layout overseas GEO marketing, what key steps do they need to go through? What differences will there be in the paths for enterprises in different industries and of different scales?

Tan Li: From a practical perspective, there are four key steps: First, disassemble how your customers screen suppliers and what this path is like. Second, reconstruct your content structure, including the official website, blog, FAQs, cases, etc., all in a way that can be understood by AI. Third, build an external trust system. For example, evaluation platforms, media exposure, social media discussions, and industry citations are all very important. Fourth, do multi - channel distribution and continuously make your content be seen and cited on different platforms.

Different enterprises may have obvious differences in two aspects: industry complexity and brand foundation. For example, the medical and industrial industries need more professional endorsements than other industries. Enterprises with a brand can more easily amplify their advantages, while those without a brand need to "make up for trust" first.

36Kr: During the implementation process, how should enterprises make changes or even reshaping at the content strategy and technical infrastructure levels to meet both AI understanding and user resonance?

Tan Li: In terms of content strategy, it is necessary to have a clear structure, information with evidence and data sources, and the expression should be close to real - life communication. At the technical level, the basics are more likely to be ignored: such as Schema structured data (Product, FAQ, Article, etc.), robots, index, hreflang and other basic configurations, and also the readability of the page. In summary, it is to create "content that can be understood".

36Kr: When laying out the GEO overseas - going plan, what is the most easily overlooked but crucial link for enterprises? For small and medium - sized enterprises lacking a professional team, what feasible solutions or cooperation models are there?

Tan Li: After serving so many cross - border enterprises, we found that the most easily overlooked point is actually: "Let AI access you". It sounds very basic, but many enterprises will fall into misunderstandings in practice: for example, the robots block AI crawlers, the page noindex setting is wrong, or there is a mess in multiple languages... All these will directly lead to AI not being able to see your content no matter how much you create.

The second easily overlooked point is off - site trust building. Many people still think that "publishing articles = doing optimization", but GEO values more: comments, reviews, social media discussions, and media citations.

For small and medium - sized enterprises, I suggest two paths: ① Lightweight: Focus on a niche area to create in - depth content + a small amount of high - quality external endorsements; ② Cooperative: Find a service provider with resources for integration (content + channels + data), and don't blindly make mistakes by yourself, as the cost will be very high.

36Kr: When enterprises choose a GEO service partner, what core capabilities of the partner should they focus on (such as whether they have a vertical industry knowledge base, cross - platform channel resources, an effect monitoring and analysis system, etc.) to avoid the trap of "ineffective investment"?

Tan Li: This question is very crucial, because there are indeed quite a few "concept - based service providers" now. I suggest focusing on four capabilities:

First, the ability to understand the industry. Do they understand your industry, not just know how to write content?

Second, the quality of content production. Do they have real cases and the ability for localization, rather than just pure AI batch generation?

Third, the ability of channels and distribution. Do they have overseas platform resources, such as media, communities, and review channels?

Fourth, the data monitoring system. For example, after doing it for a while, see if it has been cited by AI? Which content is effective? What is the ROI?

If the other party can only talk about "how many articles to publish", it can basically be directly excluded.

36Kr: The popularity of Seedance 2.0 at the beginning of 2026 made the industry see the possibility of AI video generation leaping from "trial - and - error like drawing cards" to "industrial production". What does the "cost collapse" of only 8 yuan for a 10 - second product promotion video mean for the content production logic of cross - border e - commerce?

Tan Li: This change is actually quite subversive. In the past, making videos had high costs and long cycles, so what people competed for was "whether there was content"; but now the cost is close to zero, which means that content is no longer a threshold, and "whether the content is effective" has become the threshold.

36Kr: When the video production cost approaches zero and everyone can generate high - quality content in batches, how will the focus of enterprise competition change? From the perspective of GEO layout, how should enterprises build marketing thinking in the AI era?

Tan Li: When the supply is infinite, what is scarce changes. I think future competition will focus on three points: First, cognitive occupation. Are you in the users' minds? Second, trust structure. Has it been verified by multiple parties? Third, data ability. Can you quickly iterate the content?

From the perspective of GEO, I think enterprises should establish a new thinking: They should regard brand content as an asset for long - term accumulation, and at the same time build up trust such as word - of - mouth and reviews bit by bit. More importantly, before publishing content, ask yourself: Will AI use it? Will users trust it? Avoid creating "self - indulgent" content.

36Kr: The popularity of Seedance 2.0 has also sparked discussions about deepfakes and copyright. How should cross - border e - commerce sellers avoid potential risks of portrait rights and copyright when using such tools to generate marketing materials?

Tan Li: This issue really cannot be ignored. I suggest trying to avoid using recognizable people (especially public figures) as generation materials. Try to use commercially - licensed libraries for material sources. Confirm the scope of authorization for brands, Logos, music, etc. Be particularly cautious when it comes to medical and efficacy - related content, as the compliance risks are higher. After all, AI can help you reduce costs, but it can't help you be "exempted from liability".