“i googled my own company's main key words, and the result was steady in the first two rows. Turning around, asking the `specified supplier of chinese tractor fittings', looking for three rounds of answers, i can't see the shadow of my company. I'm starting to wonder if ai slipped me into the blacklist."
The person who said that was zhang, the owner of an export company for agricultural machine parts in hebei. They have been doing this for almost 20 years, and the main tractor clutches, brakes and falsing are the old lines, mainly to the middle east and eastern europe。
Zhang is not a technical person — the website is his own eyesight, and seo has spent money on perfecting. What makes him wonder is why ai refused to “recognize” his company
In the early spring of this year, liu hebei found a team of pandas heading out of the sea to perform a thorough “ai search medical examination”。
This is not a complete geo service optimization, but a deep-seated cause-of-health mapping -- positioning "ai why not you" at the very heart of the problem, and seeing if there's any need for a systematic overhaul after that。
I sorted out the entire diagnosis. If you're also a trader in mechanical parts, and you feel like you've been invisible for a long time in the ai search, maybe this article will help you find a way out。
I. Pre-diagnostic: ai is almost invisible
Based on standard processes, the study simulates 35 typical overseas buyers ' issues (covering product selection, quality validation, vendor comparison, procurement decision-making, etc.) from the chatgpt, perplexity and gemini platforms, and records brand reference rates, references and answers。
The following is the baseline data before optimization:
After seeing the result, liu fell into a long period of deep contemplation that “the foreign trade industry, for many years, seems to be in a `unrecorded' state in the ais”
Ii. Diagnosis: four key symptoms gradually surfaced
The team of pandas out of the sea, geo, took two weeks to complete three in-depth screenings: a comprehensive audit at the technical level of the network, a reverse evolution of the ai search performance, and a mapping of the customer procurement habits. Ultimately, they have identified four core barriers that allow ai to “blind”。
Question 1: business entity information “fights” on multiple platforms
At the time of searching, ai could not confirm that the three sources of information were directed to the same entity. Therefore, it does not combine information from three sources to verify credibility, but rather to score independently — each source has a low “trust score”。
Question 2: the product description is vague. People and ai can't read it
The common writing on the official web product page is made of high-quality products, weather-resistant and durable, capable for various trackor models
It works on traditional websites, but for ai it is basically nothing. What ai needs is something specific that can “capture”, not a bunch of words:
1 “high-quality materials” — what materials, asbestos-free ceramics, or organic compounds
2 “technology” - for how long? How many hours did the lab run
3 “application of various types” — which type is the type of brand
So, for example, it says clearly, "clutchdisc for new holland t6 series, uses kevlar-reinforceed competition, spoke to 4,200 hours, meets iso 14113 standing."
See? This is the first time that ai can bring out the real entities of “newholland” “kevlar” “iso 14113”. When the customer asks for “high-temperature centrifuge tablets for t6 in new netherlands”, ai will naturally give priority to such specific companies。
Question 3: there is a lack of "helping people make decisions" on the website
On liu's website, it's a product page and company profile, and there's hardly any of the following types of ai's favorite content:
• procurement advice category, such as “how to select tractor clutches according to horsepower and operating environment”
• comparative analysis, for example, “whether ceramics in deep-farming are more suitable for organic pieces”
• after-sales, for example, "why do you slide with a new clutch?"
When the customer asked ai “what are the differences between these suppliers” or “what is to be taken into account in purchasing such accessories”, ai did not have anything on hand to quote liu and naturally did not push his company out。
Question 4: the website has been frozen for over two years
The most recent case on liu's website was in 2022; the product parameter page was still the old model that had been discontinued; and the company news column, the latest one, was issued in 2021。
The ai algorithm is actually “pick-and-take” and prefers fresh and continuously updated content. The competition against liu is almost weeklyInkedin posted technical posts, and new cases are posted on a quarterly basis. And on liu's side, the content is like a pause button. Ai will be conscious that the company may not be running much, and it naturally goes back when it's recommended。
Iii. Summary of diagnostic results: a table showing the problem
After looking at the form, liu xiao said, “although i used to believe that artificial intelligence had limitations, it seems that i did not provide the relevant information to artificial intelligence for `identify me'.”
Solutions: a phased approach
After a clear explanation of the problem, the panda team went out to sea to prepare a phased overhaul for liu. It's not the kind of thing that goes up and down, but it's a priority, a phase forward。
Liu decided to start with the focus of the first and second phases, without making any immediate technical changes to the website - because of the age of his website system, the addition of schema would need to be rebuilt, while the current budget did not allow it。
V. How about an "ai search visibility check"
This farm spare parts factory, as mentioned earlier, is not an example. Over the past year, a team of pandas out of the sea from geo has conducted a similar diagnosis for dozens of foreign-trade enterprises, and has found that more than 80 per cent of the “ai can't find” problems cannot escape these three types of problems: inconsistencies in entity information, lack of key entities in product descriptions, and lack of content on the website to help clients make decisions。
If you also feel like a "transparent person" in ai, do three things first:
A self-search: enter your brand name in chatgpt, perplexity, gemini, some of the main product words, and three or four of the business questions that clients often ask. See if ai mentioned you or not。
Nuclear information on different platforms: put your network, lInkedin, company name, address, telephone listing on the main b2b platform, to see if it is fully consistent。
“dry goods” in the description of a number of products: select a description of a single product page and how many specific factual points there are — such as materials, standards, appropriate models, test data. If it's less than three, it's really too “fake”。
After these three steps, if you find a bunch of problems, or if you want to get a professional i. I. D. Like liu, you can come to the geo. In the same way, we'll help you figure out the real reason why "ai can't see you" and give you a dry road map。
The diagnostic methodology and data for this paper refer to the “geo diagnostics” section of the white paper on panda out of the sea, b2bgeo growth engine for foreign trade, where customer names and some of the information have been desensitized. Welcome to the comment section to talk about your "neglected" by ai, and we can personally get a free a. I. V. Examination。
