Fewer Export Inquiries? Your Buyer Asked AI First

Fewer export inquiries do not necessarily mean the market has worsened. It may be that the buyer asked AI before ever contacting a supplier — and you were not in that round of screening.

But the statement does not work in reverse. Fewer inquiries can also mean a category cycle is out of step, platform traffic has become more expensive, or your own quoting and response have slipped. All of those causes are real. What this article is about is the one that is easiest to overlook and hardest to notice yourself: the buyer's screening now happens earlier, somewhere you cannot see it.

Four possible reasons inquiries fall — and the one most people miss

Break "fewer inquiries" down and there are only four possible reasons.

The first: demand really has fallen. A category has hit a cyclical trough, or your main export market is contracting. This one is the most visible and the easiest to spot.

The second: platform traffic has become more expensive. The same budget no longer buys the exposure it did a year ago.

The third: your own follow-through has a problem. Slow quotes, incomplete materials, samples sent with no follow-up, a salesperson who left and took the customer records with them. Problems like these are internal — you can find them if you look.

The fourth, and the least discussed: the buyer already ran one round of screening through AI before contacting you.

Why is the fourth one hard? Because it leaves no signal. Your back office shows no visitor, no comparison, no one crossing you off a list. So you assume it is one of the first three — and you buy more ads, switch platforms, push prices down. But the problem was never in that direction.

Asking AI is no longer a niche behaviour

The China Internet Network Information Center (CNNIC) published its Generative Artificial Intelligence Application Development Report (2025) in October 2025. Several figures in it deserve a careful read from anyone in export trade.

As of June 2025, China's generative AI user base reached 515 million people, up 266 million from December 2024 — growth of 106.6% in half a year, with a penetration rate of 36.5%.

The use cases matter more. According to the report, the most widespread use of generative AI products is "answering questions", at 80.9% — first among all scenarios. Separately, more than 90% of users reach for a Chinese-developed model first.

Read the two together and the point is clear: most people are using AI for the same thing — asking questions. And what they use most are Chinese engines such as Doubao, DeepSeek, ERNIE Bot, Kimi, Tencent Yuanbao and Qwen.

Sourcing research is, at bottom, a string of questions. Who can make this product? What certifications does this factory hold? What is the minimum order quantity? How long to produce a sample?

Buyers used to open a dozen pages and compare them by hand. Now they can ask all of it in a single conversation.

One distinction has to be made here: 515 million is the size of the generative AI user base, not the number of export buyers who use AI to find suppliers — no authoritative statistic exists for the latter. But the 80.9% figure for "using AI to answer questions" shows that AI is, first of all, being used as a place to ask questions.

What buyers ask AI is exactly what your website does not say

Sort the questions buyers ask AI and you get roughly seven types:

(This breakdown comes from publicly available summaries of B2B buying-decision stages — it is not official statistics, so adjust it to your own industry.)

All seven share one thing: every one of them is asking for objective facts — MOQ, certification numbers, sampling lead time, capacity ranges, what cases you have delivered.

And those facts are almost certainly not on your website — not one of them. Your product pages say "supports all kinds of customisation", "reliable quality", "on-time delivery": all adjectives. The numbers your sales team knows by heart have never made it onto paper.

When AI cannot find the numbers, it names the competitors who wrote theirs down.

Why AI names your competitors, not you

This brings us to GEO (Generative Engine Optimization) — what people now call AI search optimization.

It is not the same thing as traditional SEO (Search Engine Optimization). SEO competes for rank on a search results page; GEO competes for being cited inside an AI answer. The first works on a ranking algorithm, the second on whether a model thinks of you at all.

Whether AI mentions you comes down to roughly three filters: are the sources credible, can the content be parsed, and do the facts line up.

The first filter: your information has to sit inside the pool of sources AI draws on. That pool is built from public web pages and content platforms. Whether a B2B platform's store detail pages are open to AI differs from platform to platform and is mostly not disclosed — so you cannot put all your chips on your platform store page.

The second: the content has to be machine-readable. Subheadings, lists, tables, Q&A — AI can extract those directly. Long blocks of prose are hard work for it.

The third: the facts have to line up. Before citing you, AI cross-checks. Your website says the company was founded in 2021; an article says "more than ten years in the industry". When the two dates do not match, it simply drops you.

Corporate spending on AI search optimization (GEO) has already started to move. A report by People's Daily Online in March this year cited figures from the China Academy of Information and Communications Technology (CAICT) in its 2025 Generative Engine Optimization Industry White Paper: by the end of 2025, China's GEO market had reached RMB 4.2 billion, with a compound annual growth rate of 38%, and more than 68% of large and medium-sized enterprises had already put GEO into their annual marketing budget.

The same article also cited Gartner: the global GEO services market passed USD 1.2 billion at the end of 2024 and is expected to reach USD 3.5 billion by 2026.

The same distinction applies here: RMB 4.2 billion is the domestic Chinese market; USD 1.2 billion is the global market. These are not the same statistical scope, and they cannot be converted into each other or compared.

What GEO actually fixes: where your information sits

To pull the previous sections together: the customer has not disappeared. They have simply moved somewhere else to do their homework. Your information has not changed either — it is just not sitting where they are reading. That is the core problem AI search optimization addresses.

So the first step in GEO is not writing articles. It is getting the facts that are already true on your workshop floor and in your sales team's heads down on paper: your minimum order quantity, how many days for a sample, which materials you work in, capacity for small batches and for bulk, what you have made, and what you do not take on.

None of it needs to be invented. You already operate this way — you have just never written it down.

One more caution: what a company says about itself is not necessarily accepted at face value by AI. Company evidence falls into roughly three kinds — self-attested (your own specification pages and FAQ), third-party attested (customer cases, real discussion in industry communities) and authoritatively attested (certification body records, industry association material, third-party test reports). Only when all three are present is AI more likely to mention you consistently in its answers. Self-attestation on its own is just self-promotion.

On "how many buyers use AI" — why I will not give you a number

Before writing this I went through the literature. I had planned to give you a "XX% of export buyers already use AI to find suppliers".

After checking, I decided not to.

Because the numbers contradict one another. For the same year, different sources give six versions: 29%, 45%, 51%, 73%, 84%, 94%. Even the single figure of 94% is attributed to Forrester in some articles and to 6sense in others. China's GEO market size is the same story: for one and the same year, reports put it at RMB 250 million, at RMB 4.2 billion, and at RMB 28.6 billion.

Trace those articles back and most are vendors' own promotional copy. A set of contradictory numbers does not help you decide anything. It only helps someone sell something.

What can be confirmed is the direction and the order of magnitude: a generative AI user base of 515 million, doubling in half a year, published officially by CNNIC, and "answering questions" as the number one use case. As for how deeply export buyers actually use AI, there is no credible authoritative statistic — that gap is something you should remember too.

Three questions about inquiries and AI that need straight answers

Fewer inquiries — does that mean customers no longer want me?

No — the two are not the same thing. There are four reasons inquiries fall, and this article covers only one of them. Spending money on GEO without knowing which one applies to you is just a different way of wasting it.

The test is simple. Open Doubao or DeepSeek, type in the product specification you are strongest on, and see whether you appear in the answer. If you do, the problem is not where your information sits. If you do not, it is worth dealing with.

How soon will I see a change after doing GEO?

No promises on timeline, and no promises on results. Nobody can give a definite schedule for this — any claim that you will "be in AI recommendations within 90 days" lacks verifiable support.

The openly observed pattern: on time-sensitive platforms, content weight concentrates in the first week or two; on some platforms it takes about three months; evergreen content keeps getting cited. So this is ongoing work — not one batch of articles and done.

My customers are mainly in China — do I still need this?

You do, but the emphasis is different. Buyers in China and buyers overseas use two different sets of AI. Chinese buyers ask mainly in Doubao, DeepSeek, ERNIE Bot, Tencent Yuanbao and Qwen; overseas buyers mostly use English-language tools such as ChatGPT and Perplexity. The two sets of engines draw on different content pools, so the content has to be planned separately.

Also, more than 90% of Chinese generative AI users reach for a Chinese-developed model first (CNNIC data). If domestic traders make up a meaningful share of your customer base, you cannot leave the Chinese line empty.

A closing note

With fewer inquiries, the thing to fear is not the falling number. It is diagnosing the wrong cause.

If customers really are fewer, the move is a different market or a different category. If they have only changed where they do their homework, the move is to change position — put the facts that are already true about you somewhere AI can read them.

Those are two entirely different actions. Work out which one you are in before you decide where the money goes.

When did you last feel inquiries were thinning out? Has a customer ever told you they found you through AI? Tell us in the comments.

Shenzhen Yihang Network Technology Co., Ltd. expanded into GEO in 2024 and has served a good number of export businesses across the Pearl River Delta and the Yangtze River Delta. We have written this approach up as an open method at www.chyihang.com/en/geo/. To understand where your own company's information sits in AI, start there.

References

1. China Internet Network Information Center (CNNIC), Generative Artificial Intelligence Application Development Report (2025), published 18 October 2025. cnnic.net.cn
2. People's Daily Online Finance, "How to handle the 'double-edged sword' of AI search," 20 March 2026 (cites the China Academy of Information and Communications Technology's 2025 Generative Engine Optimization Industry White Paper and related Gartner data). finance.people.com.cn
3. Xinhua News Agency, republished by People's Daily Online, "China's generative AI user base reaches 515 million," 19 October 2025. finance.people.com.cn

Data and scope note

The user-base and industry figures cited here are macro-level statistics, used to illustrate changes in how information is obtained. They do not mean that any specific industry or individual company will necessarily receive inquiries or orders. Descriptions of how each AI platform builds its sources and cites content come from public material and industry observation; no platform publishes its recommendation mechanism officially, and behaviour may change as platforms iterate. Before acting on this, a company should still assess its own products, target markets and content base independently.

#ExportTrade #CrossBorderEcommerce #AISearch #GEO