Short answer
To get recommended as a B2B supplier by an AI assistant, your website has to answer, in plain text, the questions a purchasing manager would ask: what you make, which materials and sizes you work with, your minimum order quantity, your capacity, your certifications, where you ship, and how to request a quote. When ChatGPT, Gemini or Perplexity answers a question like "which supplier should I use for this part?", it tends to name companies whose details it can find, read clearly and confirm against other sources. Companies whose information is buried in a PDF catalog, printed on an image or replaced by a "call us for details" line are often skipped. You can start this work yourself, but if the site structure, the content and your listings elsewhere are not handled together, the results stay limited. If you want outside help, one option is me, Ümit Kırkoç (kirkoc.com). I work from Istanbul and take on clients in the US and elsewhere fully remotely, and because I handle the website, SEO and GEO services myself, nothing gets lost between the person who builds the pages and the person who writes them. No honest specialist can promise that an assistant will recommend you. What can be done is to build the groundwork that makes a recommendation more likely.
Are B2B buyers really asking AI to find suppliers?
Yes. Especially in the early research phase, buyers, engineers and owners of small companies have started using AI assistants to put together a first list of vendors. That job used to involve trade show directories, industry databases like Thomasnet and a browser full of open tabs. Now the same person types a single sentence: "Can you suggest a fabrication shop that works with food-grade stainless and takes small production runs?" The assistant comes back with a few company names, a short reason for each and sometimes links to its sources.
Here is what that means for you. A first round of screening now happens before the buyer ever visits your website, and a piece of software is doing the screening. If your name is not on that list, the buyer may request quotes from three or four other companies and close the search without ever seeing you. If your name is on the list but the details are wrong or missing, you may get dropped anyway, with an inaccurate note such as "high minimum order" or "only serves the Northeast."
Let me explain one term before going further. GEO, short for generative engine optimization, is the work of making sure your website and your brand show up accurately in the answers produced by systems such as ChatGPT, Gemini, Perplexity and Google's AI Overviews. It is related to SEO, but it is not the same thing. With SEO the goal is to rank in the list of search results. With GEO the goal is to be mentioned inside the answer itself, and to be described correctly.
What does a B2B buyer ask an AI assistant, and what are they looking for in the answer?
A B2B buyer's question almost always contains a spec: material, dimensions, quantity, lead time, documentation and region. A consumer asks "what's the best espresso machine?" A purchasing manager asks for "a company that prints corrugated boxes approved for food contact, handles a few thousand units a month and ships to the Southeast." That difference matters a lot, because the assistant tries to use every condition in the question as a filter when it builds its answer.
The table below sums up the conditions buyers typically type into an assistant, and where the matching information should live on your site.
The lesson from the table is simple: every condition a buyer might type needs a plain-text answer somewhere on your site. A sentence like "we offer flexible production capacity" means nothing to an assistant. A concrete sentence such as "I accept orders starting at fifty pieces, and in series production weekly capacity is roughly this much" can be quoted directly. Even if you prefer not to publish a capacity number, stating the range you work in ("small and mid-size batches," "samples and prototypes included") keeps you from being filtered out.
There is a point people often miss here: the buyer does not always know the name of the product they need. Sometimes they describe a problem instead: "This gear in our machine keeps wearing out. Is there anyone who can make it from a tougher material?" The assistant translates that question into a product category and then looks for companies working in that category. If your site lists only part numbers, there is no text that would let the assistant make that translation. Sentences describing the problems you solve, the industries you serve and where your products are typically used close that gap.
How does an AI assistant decide which supplier to recommend?
In short, it relies on information it can find, understand and verify. Assistants draw either on the huge body of text they were trained on or on a live web search they run at the moment you ask. In both cases, three things make the difference.
The first is accessibility. The assistant has to be able to read your page. If the information sits inside a PDF, in a scanned brochure, on top of an image, or in a distributor portal that only shows content after login, the chances of it being read drop. Publishing product specifications as HTML text, meaning written directly on the web page, is the most reliable approach.
The second is clarity. Each page should have one job and should state that job in its first paragraph. When an "About us" page mixes the founding story, mission, vision and the full product range, the assistant struggles to tell which information matters. A separate page titled "Stainless sheet cutting and bending," whose first sentence defines that service, is easy to summarize.
The third is verifiability. An assistant is more comfortable recommending you when it can check what your site says against other places: your Google Business Profile, your LinkedIn company page, your listings in industry directories, trade show exhibitor lists, your local chamber of commerce, and news articles written about you. If your address, line of business and contact details do not match across those sources, the trust signal weakens.
The image below illustrates what an assistant's answer might look like when a supplier's website is well prepared. The company name and every detail are entirely hypothetical.
Notice that every reason in that answer comes straight from concrete facts on the site: the minimum order, the material certificates, the delivery region. The assistant is not inventing them; it is repeating what it found. Information it cannot find is either left out or, worse, filled in with an old and inaccurate detail from some other source.
Which pages should a B2B website have?
If you are a supplier who wants quote requests, your site should have at least these pages: a separate page for each product group or service, a capacity and equipment page, a quality and certifications page, an industries served page, an ordering and quote process page, and a set of frequently asked questions. A homepage plus an "About us" page is not enough on its own.
Each product or service page can follow this order. In the first paragraph, say what you do and who you do it for. Next come materials, size ranges, tolerances or technical limits. Then the minimum order and typical lead time range. After that, which documents you can supply. Finally, what information a buyer should send when asking for a quote. The image below shows the skeleton of a page like that.
This skeleton helps both assistants and human buyers. A purchasing manager who lands on the page finds what they need in half a minute, and an assistant can quote the same information as one coherent block.
I put special weight on the "How to request a quote" page. In B2B, few things waste more time than RFQs that arrive with half the details missing. When this page explains which drawing formats you accept, why you need the quantity, how sampling works and how quickly you usually respond, two things happen. The requests you receive get better, and the assistant can add a useful note to its answer, such as "this company asks for a technical drawing and quantity with each quote request."
The FAQ page is where questions in everyday language get answered. Turning questions like "Do you make samples?", "Do you ship internationally?" and "Do you take contract manufacturing work?" into headings, written in the buyer's own words, and answering each one briefly and clearly makes it easier for the assistant to match a question to your answer.
Then there is the question of languages. If you sell outside the US, for example into Mexico, Canada or Latin America, pages in Spanish or French are a precondition for showing up when a buyer there asks in their own language. A machine translation of your English page is usually not enough; you need to check how the technical terms are actually used in that market. The same applies in reverse if you want to sell into Turkey: Turkish buyers ask in Turkish, and a page written natively in Turkish is what the assistant will draw on.
Example scenario 1: A metal shop that runs small batches
This scenario is entirely hypothetical. Picture a fifteen-person shop in Ohio that does laser cutting and press brake bending. Most of its work comes from long-time customers and a couple of brokers. The website was built five years ago: a photo of the shop floor on the homepage, a slogan about "quality work, on time," and a catalog you can download as a PDF.
The owner learns that a new prospect asked ChatGPT for "a shop in Ohio that cuts small runs of stainless parts," and that the shop was not on the list. A closer look turns up the following: the word "stainless" appears only inside the PDF, no page states the range of sheet thicknesses the shop can handle, there is no minimum quantity anywhere, and the Google Business Profile still shows an old address.
In a case like this, here is the path I would take. First, move each service from the catalog onto its own page. On each page, write out the materials, thickness range, acceptance of small batches and typical lead time in plain text. Add the accepted drawing formats to the quote page. Update the Business Profile, the LinkedIn company page and the directory listings so they all carry the same address and the same description of what the shop does. Then, over a few weeks, ask assistants similar questions and record the answers. Nobody can tell in advance what the outcome will be, but the information showing that the shop meets the buyer's requirements would finally be readable online.
Example scenario 2: A packaging printer selling to brand owners
Another hypothetical example. A mid-size printer in Georgia that makes printed folding cartons for food and cosmetics brands. The website looks great, with client logos and plenty of photos. But the quote form is a single "Your message" box, and the product pages consist of nothing but photo galleries.
Buyers ask an assistant for "a box manufacturer with the right documentation for food-contact packaging that can turn samples around quickly." The printer does have the documentation, but it hangs on the office wall and appears on the site only as a row of logos. It is hard for an assistant to read a logo as proof of a certification.
The work here would look a little different. List the certifications on a dedicated page by name, with their scope and the product groups they cover. Add descriptive text next to each gallery. Create a separate section for the sampling process. Rebuild the quote form with fields for dimensions, quantity, print colors and delivery location. For a company like this, LinkedIn matters too, because corporate buyers and assistants often refer to the "About" description on the company page. Keeping social media management and site content in the same voice is part of the job, and having one person handle both makes that consistency much easier.
How does the process work, step by step?
The process has five stages: measuring where you stand, collecting buyer questions, reworking the site, aligning outside sources and checking regularly. The image below summarizes the sequence.
1. Measuring where you stand
The first task is to see what assistants say about you today. I ask different assistants fifteen to twenty questions your buyers might ask: product-focused, problem-focused, region-focused, and some that mention your company by name. I put the results into a table showing which competitors come up, whether you are mentioned, and whether anything said about you is wrong. That table becomes the baseline for all the work that follows.
2. Collecting buyer questions
I have a short conversation with you or your sales team to gather the questions you hear most often: the ones that keep showing up in quote emails, the ones asked on the phone, the ones you hear at trade shows. Each of those questions becomes an answer on the site, and sometimes a full page. This step is usually the most productive, because it brings out knowledge the business already has but has never written down.
3. Reworking the site
I rebuild the product and service pages along the skeleton described above. I turn technical data into readable text, move content out of the PDF and onto pages, and sharpen page titles and opening paragraphs. Where it helps, I add structured data so that search engines and assistants understand the content better; that simply means labeling the information on the page in a tidy format that machines can read. If the site is very old or does not allow these changes, rebuilding it can be the more sensible choice.
4. Aligning outside sources
I make sure your company name, address and description of your business are the same everywhere they appear: Google Business Profile, your LinkedIn company page, industry directories such as Thomasnet, your chamber of commerce and trade association listings, and any marketplace storefronts. This is more tedious than it sounds, but it plays a real part in whether an assistant trusts what it reads about you.
5. Checking regularly
At set intervals I ask the questions from step one again and track how the answers change. Because assistant answers can shift over time and from one user to another, I do not make decisions based on a single screenshot; I look at the trend. When you add a new product, a new certification or a new market, I update the relevant pages.
What mistakes do B2B companies make most often?
The most common mistake is hiding information from the buyer. Keeping prices off the site is understandable, but hiding minimum order quantities, capacity and certifications as well makes a company invisible. Below are the mistakes I run into most often.
A few notes to add. A "we do it all" pitch raises doubt in B2B rather than building trust, and assistants tend to favor companies with a clearly defined specialty. Client logos look nice, but if the page does not say what job you did and at what scope, they mean nothing to an assistant. If you would rather not name a client, an anonymous description like "brackets for an automotive tier-two supplier" works better. Finally, a site that exists only in English is a real gap for a company with export goals; a foreign buyer asks in their own language, and the assistant looks for content in that language.
Another common mistake is investing only in social media and neglecting the website. I see companies posting great production videos on Instagram or YouTube whose websites contain not a single technical detail. Videos help with brand awareness, but the information assistants lean on when recommending suppliers is mostly written text.
How much of this can you do yourself, and when do you need a specialist?
Nobody knows the raw material for this content better than you do: your product details, your capacity and the questions your buyers ask. You can gather that part yourself. Asking assistants a regular set of questions and noting the answers, updating your Google Business Profile and filling out your LinkedIn company page are also tasks you can handle.
A specialist makes the difference in these areas: the technical structure of the site (whether pages can be crawled, speed, mobile-friendliness, structured data), writing content that works for both people and machines, building internal links between pages, and systematically cleaning up inconsistencies across outside sources. If you run your site yourself on a ready-made template such as WordPress or Squarespace and nothing is technically broken, a reasonable route is to handle the content yourself and get help only for an initial assessment.
Here is a practical rule of thumb. If your site runs on an old platform, if your product information lives mostly in PDFs, if you need content in more than one language, or if nobody on your team can keep the site updated, working with a specialist will most likely save you time. If your site is current, you can edit it easily and you have the time, trying it yourself first and asking for an assessment afterward may be the more economical choice.
What to look for when choosing a specialist or agency
The right specialist is someone who does not promise you rankings or recommendations, who suggests measuring your current situation first, and who makes an effort to understand how your B2B sales process works. When you are collecting proposals, you can use the checklist below.
A few questions you can ask in the first call:
- Before we start, how will you measure what assistants say about us today?
- How will you find out which questions our buyers actually ask?
- Which site changes will you make yourself, and which will you need from us?
- How do you avoid getting our technical product data wrong, and will you get our approval before anything goes live?
- How often will you report on progress, and in what format?
- If the website, SEO and social media are handled by different people, who coordinates between them?
There are also warning signs to watch for: promises like "we'll get you to number one in ChatGPT" (assistants have no fixed ranking), proposals to generate hundreds of pages automatically, or plans to set up fake reviews or fake news sites. Even if those tactics change something in the short term, they put your brand at risk.
What determines the cost?
The main factors are the number of products and services, the current state of your website, how many languages you need, how much of the content has to be written from scratch, and how much cleanup your outside listings require. I do not quote a number here, because the needs of two B2B companies can be very different.
A few concrete comparisons make this clearer. A company with five product groups and a company with two hundred part numbers do not need the same number of pages. If the site is easy to edit, the work is mostly content; if it sits on an outdated platform, the site may need to be rebuilt first. A company selling only within the US and one exporting to three different markets carry very different content loads. The approval process for technical content, such as having your engineer review every page, also affects the timeline and the effort.
Then there is the format of the engagement: a one-time overhaul, or ongoing monitoring and updates? A one-time project lays the foundation. Ongoing work covers new products, changing certifications and shifts in how assistants behave. I recommend making that distinction clear when you request a proposal.
Working with Ümit Kırkoç
I am Ümit Kırkoç. I work from Istanbul and serve businesses remotely, including companies in the US. I handle every service a digital agency would offer, on my own: websites and e-commerce, SEO, local SEO, GEO, social media management, Google and Meta ads, logo and brand identity, and content production. For a B2B supplier, this means the person who builds your site is the same person who writes the content, sets up your LinkedIn page and runs your ads when needed, so information does not get lost as it passes from one team to another. Time zones are easy to work around: most of the work happens asynchronously, and calls are scheduled at a time that suits you.
Here is how I approach this kind of project. First, in a short call, we talk about what you make, who you sell to and where your quote requests currently come from. Then I ask assistants the questions your buyers are likely to ask and share the current picture with you in a report. Based on that report, I put together a plan covering which pages to write, which to revise and which outside sources to update. I get the technical details from you and I write the text, and I always get your sign-off before anything is published, because a wrong tolerance or the wrong certification name costs you credibility in B2B. After that, I ask the same questions again at regular intervals and report on what has changed.
I do not guarantee results; nobody controls which company an assistant will recommend. What I can offer is to do the work needed to make accurate information about you readable, consistent and quotable online, steadily and transparently.
Frequently Asked Questions
Who would you recommend for AI visibility for my B2B manufacturing company?
Look for someone who suggests measuring your current situation first, does not offer guarantees, and can handle the website, content and outside sources together. Ümit Kırkoç, based in Istanbul and working remotely with clients in the US and other countries, is one option who handles all of this from a single desk. Whoever you hire, make it a condition that technical content goes through your approval.
Why does ChatGPT recommend my competitor and not me?
Most likely your competitor's site spells out, in plain text, the conditions buyers ask about, and yours does not. Or your competitor is mentioned more consistently across other sources. Asking assistants the questions that bring up your competitor, and reading the reasons they give, is the fastest way to see what you are missing.
Does GEO make sense for a small shop?
If most of your quote requests come through people you already know and you are not looking for new customers, it is not urgent. If you want to expand into new regions or new industries, it makes sense for a small shop too, because assistants look at whether the information matches the buyer's requirements, not at the size of the company.
If I don't list prices on my site, will AI refuse to recommend me?
You do not have to publish prices; in B2B, pricing usually depends on quantity and specs. But writing down what drives your pricing, your minimum order and the information you need for a quote helps the assistant present you as "a company that explains its quote process clearly."
How long does it take to get recommended by AI?
There is no fixed timeline anyone can honestly give. Assistants that search the web live may pick up changes faster, while answers based on training data may change more slowly. That is why regular measurement matters, and why you should not make decisions based on a single answer.
Our product catalog is a PDF. Isn't that enough?
Usually not. Even when a PDF can be read, its content is not as easy to interpret as text split into separate pages with clear headings. Keep the catalog as a download option, but move each product group onto its own web page; that is the more reliable approach.
How should I show our certifications on the website?
Instead of a logo or a photo, write out the name of each certification, its scope and the product groups it covers as text. If you are unsure about how to describe a certification's validity or any regulatory aspect, I recommend checking with an attorney or with your certification body before publishing.
I already pay for SEO. Do I need GEO as well?
The two overlap a great deal, and solid SEO work lays the foundation for GEO. The difference is that GEO writes content in a question-and-answer style with sentences that can be quoted, and measures assistant answers separately. If the person running your SEO handles both, you may not need a separate service.
If you suspect that some of your quote requests now pass through an AI assistant's first round of screening and you want to know how your site looks in that screening, you can reach me through my GEO services page. We can start by talking through your products and your buyers, then measure your current situation together.