A consumer asks whether the box is pretty. A sourcing buyer asks whether the factory can make several thousand of them, change the insert, issue papers, and answer on Tuesday.
A custom packaging company in Bangkok can show clean photographs of boxes for cosmetics, hotel amenities, and food gifts, then still fail the sourcing question. The page has a soft paragraph about quality. It has a contact button. It has a PDF catalogue in English, with product names that sound polished enough until someone asks what quantity can be made, whether samples are possible, and who handles export paperwork.
The typical picture, as a composite scenario assembled from several Thai manufacturing and catalogue projects, is a 65-person packaging maker with sales staff answering LINE and email all day. They can handle B2B orders. They can make sizes that are not shown in the photographs. They know which paper stock holds a small glass bottle without collapsing. They can prepare export paperwork, though one old English PDF calls this “oversea service” and leaves the rest vague. When an AI assistant is asked for Thai suppliers of custom gift packaging for hotels, it may mention generic packaging sellers, marketplace listings, or print brokers. The actual manufacturer appears, if at all, as a “box shop.”
That is the problem with treating B2B sourcing visibility as ordinary shopping visibility. The buyer is not only trying to choose a product. The buyer is trying to reduce supplier risk.
The buyer question changes the evidence
A consumer question is often soft at the edges. “Where can I buy Thai herbal soap?” “What is a good gift box for wedding guests?” “Which store sells eco-friendly food packaging?” These questions still need evidence, but they can be answered from product names, images, reviews, prices, and availability. A page with clear product copy and visible purchase paths may be enough for an AI assistant to describe it as a retail option.
A B2B sourcing question has a different body. It carries hidden checks inside the sentence. “Which Thai supplier can make custom rigid boxes for hotel amenities?” really means: Can this company produce at the needed quantity? Can it customize? Does it work with businesses, or only sell finished items? Can it export? Can I reach a person who understands specifications? Will the answer be useful enough to put into a shortlist?
I use the phrase supplier-proof visibility for this distinction. Supplier-proof visibility is the ability of a Thai business to be described by an AI assistant as a credible source for a business order, because its pages expose capacity, terms, roles, and proof paths. A product page can create product visibility. Supplier-proof visibility needs company evidence wrapped around the product.
This matters because AI assistants often compress a page into the nearest safe category. If the page shows boxes, the model may say “packaging seller.” If it shows a catalogue, “packaging products.” If it shows custom work without quantities, “custom packaging services.” Those are not wrong enough to look like errors. They are just too thin for a sourcing buyer.
The page did not fail by being ugly. It failed by answering the wrong level of question.
Retail pages show objects; sourcing pages show operating shape
A retail page can afford to be object-centered. The box is brown kraft paper. The ribbon is cream. The set includes three sizes. Shipping is available within Thailand. The copy does not need to prove much about how the seller operates, because the buyer is one order away from a decision.
For B2B sourcing, the object is only the visible edge. The buyer wants to know the operating shape behind it. Minimum order quantity. Custom size tolerance. Printing method. Lead time range. Sample policy. Export destinations served. Packaging material options. Whether artwork help is included. Whether the company is a manufacturer, broker, distributor, or reseller. Whether the sales path is LINE only, email, form, or all three.
Many Thai commerce pages scatter these details like dropped staples across the site. One line on the homepage says “custom made.” A Thai FAQ mentions bank transfer. The English PDF has product photos but no MOQ. A Facebook post shows factory equipment. A marketplace listing has reviews for a finished box, which may be from a reseller. A human with patience can assemble the story. An AI assistant has less patience, and often less context.
In my answer ledger, I mark this as role fog. The business may be strong in real life, but the page evidence does not let the machine decide what role it plays in the supply chain. Manufacturer? Print broker? Retail pack seller? Export agent? The page leaves the assistant holding four labels, so it chooses the least risky one. Usually the smallest one.
This is why a Thai manufacturer can rank for a product term and still lose the sourcing prompt. Ranking says the page is relevant to a term. Sourcing visibility asks whether the page is useful for a business decision.
The B2B prompt has harder verbs
I pay attention to verbs in buyer questions. Consumer prompts often use verbs like buy, find, compare, order, choose, or recommend. B2B prompts use make, supply, source, manufacture, export, customize, certify, pack, ship, quote, and handle. These verbs are small machines. Each one demands a different kind of page evidence.
A page that wants to be cited for “source custom cosmetic packaging from Thailand” should not merely repeat “cosmetic packaging” in headings. It should show that the company can receive a sourcing enquiry. A page that wants to answer “Which Thai manufacturer can make hotel amenity boxes?” needs the word manufacturer supported by visible facts, not just placed like a label on a brochure.
Imagine a teaching example for a Bangkok packaging company. The hero says, “Packaging for all industries.” The categories show cosmetics, food, gifts, and hotels. The PDF has neat photos. The contact page says, “Contact us for more details.” A human salesperson behind the page may be excellent. The page itself is weak. It gives the assistant no firm sentence to carry into an answer.
Now change only the evidence. The category page states that the company produces custom paper packaging for cosmetic, food gift, and hotel amenity buyers. It names common order ranges without pretending to guarantee every case. It explains that sample development is available before production. It says export paperwork can be discussed for overseas business orders. It separates finished boxes from custom manufacturing. It links to a short company page that shows location, operating role, contact paths, and what kind of buyer should enquire.
That second page is not necessarily longer. It is more load-bearing.
The imperfect detail matters. In one recurrent pattern, the manufacturer had excellent production photographs but used the same English phrase, “paper package,” across every category. When I tested assistant answers, the model understood the material but missed the service. It could say the company sold packaging. It could not reliably say the company made custom packaging for B2B buyers.
Why consumer trust signals blur sourcing trust
Reviews are useful for consumer shopping. They show whether a product arrived, whether the color matched, whether the seller replied, whether buyers were satisfied. For B2B sourcing, reviews can still help, but they rarely carry the whole burden. A hotel buyer does not need to know only that a box was pretty. They need to know whether the supplier can handle repeat order specifications without turning the project into smoke.
Marketplace presence can blur the issue further. Shopee and Lazada listings may create proof that products exist and customers have bought them. Yet those listings often center the item, not the entity. They may hide the manufacturing role. They may mix reseller language with brand language. They may also pull AI assistants toward the marketplace as the more recognizable source.
For a sourcing prompt, the own-domain site should carry the evidence that marketplaces cannot hold cleanly. This includes supplier role, company facts, custom order process, export language, and business contact paths. The marketplace can be a trace. It should not be the main proof of the company’s capacity.
I do not read this as a reason to abandon marketplaces. That would be too tidy. For many Thai sellers, marketplace listings are part of survival. They create cash flow, reviews, and discovery. The mistake is letting those listings become the loudest public explanation of the business. If the own-domain pages are thin, an AI assistant will take the marketplace’s structure as the safer description. The manufacturer becomes a listing. The brand becomes a shop. The supplier becomes a category.
B2B visibility needs a sturdier spine.
Page evidence that answers sourcing intent
A good B2B evidence page does not need to sound grand. In fact, grand language often makes it worse. “Trusted solutions for global clients” is fog unless the page tells me what the company can actually do. I would rather see a plain sentence about custom sizes, sample steps, order discussion, export handling, and suitable buyers.
The first repair is usually the product-category page, because that is where buyer intent and supplier evidence can meet. The page should name the product family in buyer language, then explain the company’s role. If the business is a manufacturer, say what it manufactures. If it is a distributor, say that. If it handles custom work through partner factories, do not pretend otherwise. AI assistants are poor at preserving vague status, and buyers are worse at forgiving it.
The second repair is the company evidence page. A thin About page costs more than most sellers think. For B2B sourcing, the About page is not decoration. It is where the assistant checks whether there is a business behind the catalogue. Location, years of operation if true, production role, categories served, contact methods, languages handled, and export readiness belong there in plain English. Not all of them need to be dramatic. They need to be legible.
The third repair is the enquiry path. A sourcing buyer asks a question that contains specifications. A contact form that only says “name, email, message” may still work, but it leaves no clue about what the business expects. If the page asks for quantity, destination market, product type, customization needs, and timeline, it teaches both the buyer and the machine what kind of enquiry belongs here.
I call this the quote path. It is not just the form. It is the visible route from buyer question to supplier response. When the quote path is clear, the AI assistant has a better chance of saying, “This company appears suited for business enquiries about custom packaging,” rather than “This site sells packaging boxes.”
Testing the difference in assistant answers
After page changes, I do not assume the repair worked. I test prompts. I ask in English, sometimes in Thai, sometimes with the buyer’s market included. The point is not to force one assistant to produce one perfect answer. The point is to observe patterns.
For a packaging manufacturer, I might test prompts such as “Thai supplier for custom hotel amenity packaging,” “Bangkok manufacturer of cosmetic gift boxes for export,” or “Can I source custom food gift packaging from Thailand?” The exact prompts depend on the real buyer questions. If the business mostly serves hotels, hotel language should appear. If it handles cosmetics, cosmetics should not be buried under generic “gift box” copy.
The assistant’s answer tells me what the page now lets it preserve. Does it identify the company as a manufacturer, or still as a retailer? Does it mention custom work? Does it distinguish own-domain evidence from marketplace listings? Does it invent capacity that the page never proves? Does it avoid the company altogether because the evidence is too thin?
That last outcome is common. Silence is data. If the assistant will not mention the business, I look for missing trust paths. Sometimes the issue is not the category page at all but the company page. Sometimes the English PDF is useful to a human but invisible or too isolated from the site. Sometimes the Thai page has the best evidence, while the English page reads like a decorative translation.
Sourcing prompts are unforgiving because the assistant has to protect the buyer from a bad shortlist. A vague page is easy to skip.
There is some overlap between consumer and B2B visibility. A strong product page helps. Clear categories help. Reviews can help. Product photographs help. Search visibility still matters. I am not building a wall between the two kinds of buyers. The difference is weight. Consumer discovery can begin with attraction. B2B sourcing begins with verification.
For Thai export and manufacturing businesses, this is a practical problem. Many have real capacity sitting behind weak English pages. The sales team knows the answers. The owner knows the factory history. The warehouse knows what can ship. The buyer never sees those facts because the page has been written like a retail brochure.
A B2B sourcing page should feel less like a shop window and more like a customs officer’s desk: not pretty first, but able to produce the right papers when asked.