Partner
To rank in ChatGPT as a manufacturer or distributor, put every part number on its own readable web page, let the AI search bots in, and get your products listed in the places AI checks: product feeds, distributor sites, directories and reviews. ChatGPT can only recommend a part it can find. It names whoever put the answer on the internet.
This post is the written version of our video walkthrough. It follows Dave, a maintenance manager whose production line goes down at 11 PM, as he asks ChatGPT for a part you make. If you'd rather watch, it's 13 minutes:
Planning a bigger catalog project? See how we approach ecommerce for manufacturers and B2B ecommerce for distributors.
How ChatGPT picks which supplier to name
At 11:02 PM, Dave's line is down and his regular supplier is out of stock. He doesn't call anyone. He opens ChatGPT and types: half inch brass elbow, 90 degree, in stock, ships overnight.
ChatGPT searches the web and gives him a shortlist of three or four suppliers. He doesn't fully trust it, so he asks Gemini, then Claude, then Perplexity. Different logos, same shortlist. Your competitor is on it. You're not, even though you've made that exact part for 20 years.
Here's why. AI search tools reach your pages, read them, choose sources and then name a company. Most B2B catalogs fail early in that chain. The products sit in a 400-page PDF from 2009, behind a "Request a catalog" form, or in the head of Gary, who has worked at the company since 1987 and retires in March. AI can't read any of those.
Buyers have already made this switch. Gartner found that 45% of B2B buyers used generative AI during a recent purchase. Your office opens at eight. AI doesn't sleep.
The 7 fixes
1. Put every part number online
This is the least exciting fix and the most important one. Every part number gets its own page, with the specs written out in plain words. Not page 212 of a PDF. Not Gary's memory. A real page that AI can read at 11 o'clock at night.
Check that the part number, the price (if you show one) and the key specs are in the page's HTML, not loaded later by scripts. In our audit of 99 B2B websites, 1 in 4 product pages that showed a price only showed it after scripts ran, so many AI crawlers never saw it. If you sell equipment parts, see how parts catalog software compares with a parts finder in your store.
2. Give every size and finish its own page
Say you sell that elbow in three sizes and three finishes. That's nine products. On many B2B sites it's one page with two dropdown menus. A person can work that out. To AI, it's one vague product with a mystery inside, like a vending machine with no labels on the buttons.
Give every variation its own part number, name and specs. Dave isn't searching for "elbow." He's searching for the half inch, 90 degree, lead-free version in his hand.
While you're at it, list your old part numbers too. The number Dave reads off a broken fitting is probably the one you replaced years ago.
3. List it even if you don't sell online
"We don't sell online. Pricing is custom. It goes through reps. Our dealers would riot." That's fine. List it anyway. The goal isn't a checkout button. The goal is being found.
Put the product online with a button that says Request a quote, Find a dealer or Log in for pricing, and send that request straight to the rep or dealer. Dave gets an answer at 11 PM, and your rep gets a hot lead at 8 AM. Your website stops competing with your sales team and starts feeding it.
4. Publish cross reference charts and answer buyer questions
Dave has a part number from a competitor's catalog, so he types that in. If you're a distributor, this is your moment.
A cross reference chart is a translation dictionary: the competitor's part number on the left, your matching part on the right. Publish it on the product page, and when Dave types the competitor's number, AI finds your page. Same part, in stock, ships tonight.
Then answer the questions Dave would ask a salesperson, right on the product page and in plain English:
- Does this fit my older model?
- What's the lead time?
- Is it a drop-in replacement?
AI favors a page that answers the question before it's asked.
5. Check who's on your robots.txt guest list
Every website has a small file called robots.txt. Think of it as the guest list for the bouncer at your front door. ChatGPT, Claude, Perplexity, Google and Bing each send their own bots, and the bouncer checks every one against the list.
Some bots only collect training data, and letting those in is a business choice. Others are the ones that put you in the answers: OAI-SearchBot and ChatGPT-User (ChatGPT), Claude-SearchBot, PerplexityBot, Googlebot and Bingbot. Turn those away and you can drop out of AI answers entirely. Blocking GPTBot, OpenAI's training bot, does not remove you from ChatGPT search. Blocking OAI-SearchBot does.
Many companies never made this choice. Some web security services, Cloudflare among them, offer settings that block AI bots, and they're easy to switch on without realizing it, and in our audit 12 of 99 B2B sites turned AI bots away by accident, usually at the firewall or CDN. Ask your IT team one question: who's on the guest list?
6. Build trust on other people's websites
AI doesn't just trust what you say about yourself. That's like trusting a dating profile. It checks what other people say about you:
- Are you on the "where to buy" pages of the brands you carry?
- Are you in the industry directories your buyers use?
- Do you have real reviews?
- Are your engineers answering questions on Reddit and industry forums under their own names? Real answers, not marketing.
- Do you have install videos, sizing videos and "how to pick the right part" videos?
Each one is another place your part numbers show up. Industry research from Ahrefs found that brand mentions on other websites track AI visibility about three times as closely as backlinks do.
7. Send one clean product feed everywhere
This is the shortcut that puts you everywhere at once. A product feed is one clean list of all your products, sent straight to the platforms. It's like handing the librarian your catalog instead of making them read every page.
- Google Merchant Center is free to list in, and it feeds Google Shopping and the product answers in Google's AI Mode.
- Microsoft Merchant Center feeds Copilot.
- ChatGPT has its own merchant program for product listings.
Build the feed once from the same data as your product pages, keep it updated daily, and you're on the shelf in every one of them. One caveat: Google Merchant Center only accepts products with a price any customer can buy at, so quote-only and login-priced parts stay out of the feed. They still need their own pages (fix 3) so AI search can find them.
Story: the catalog that arrived by fax
"This sounds great, but our product data is a disaster." We hear that a lot, so here's the most extreme case we've worked on.
One manufacturer's product and inventory updates arrived one way: by fax. Paper, out of a machine, in this century. Most people would say those products can't go online. We asked where the fax machine was.
We scanned the pages, used software that reads printed text and turns it into data, and built a daily product feed that pushed everything live every day. That put hundreds of thousands of products online, and it lifted the manufacturer's revenue significantly because buyers could finally find those products.
The easy part of your catalog, your competitors already have online. The messy part, the fax and the PDF from 2009, nobody touches. Whoever does gets those searches almost to themselves.
What doesn't work
Some popular advice doesn't hold up. No major AI platform says it reads llms.txt, and Google says Search ignores it. Schema markup is worth having for accuracy and shopping eligibility, but a study of 1,885 pages found that adding it produced no meaningful lift in AI citations on its own. A nice homepage and a forklift video set to inspirational music don't help either. AI recommends the supplier whose part data it can actually read.
How to check if ChatGPT recommends you
Run Dave's test this week:
- Write down five questions your buyers really ask, including at least one exact part number and one competitor's part number.
- Ask each question in ChatGPT, Gemini, Claude and Perplexity a few times. Answers vary, so one try isn't enough.
- Note who gets named: you, a distributor, or a competitor.
- For every miss, check the matching fix above. The answer is usually fix 1, 2 or 5.
For the full method, read our guide to how B2B brands get recommended by AI. Its 7 steps follow the order AI works in and go deeper than the fixes above, with a 14-point scorecard and a 90-day plan. Running a consumer store instead? See AI search optimization for ecommerce.
Frequently asked questions
How do I rank in ChatGPT as a manufacturer or distributor?
Put every part number and variation on its own page with the specs in plain text, allow AI search bots like OAI-SearchBot in your robots.txt and firewall, publish cross references and buyer FAQs on product pages, get listed on distributor sites, directories and review sites, and send a clean product feed to Google, Microsoft and ChatGPT for every product with a public price.
Does blocking GPTBot remove my products from ChatGPT?
No. GPTBot collects training data. ChatGPT search uses OAI-SearchBot to find pages for its answers. Blocking OAI-SearchBot is what removes you from ChatGPT search results, so check both robots.txt and your firewall or CDN settings.
Do I need to sell online to show up in ChatGPT?
No. You need your products to be findable online. A product page with "Request a quote," "Find a dealer" or "Log in for pricing" gives AI something to recommend and sends the lead to your rep or dealer.
Why does ChatGPT recommend my competitor instead of me?
Usually because your competitor's part data is online and readable and yours isn't. Common causes are products locked in PDFs, one page for many variations, prices loaded by scripts, AI bots blocked by accident, and few mentions on other websites.
Is ranking in ChatGPT different from ranking in Google?
The basics overlap: readable pages, clear product data and trusted mentions help in both. The difference is that ChatGPT gives a short list of named suppliers instead of ten links, so if you aren't one of the three or four it names, the buyer may never see you.
Get your whole catalog in front of AI
Getting every part number, variation, old number, cross reference and feed online is a real project. You don't have to do it yourself. IWD Agency takes catalogs with thousands or millions of part numbers from wherever they live today, whether that's PDFs, spreadsheets, an ERP or a fax machine, and gets them online, organized and in front of ChatGPT, Gemini, Claude, Grok, Perplexity and Copilot. See our work with clients like DMC Tools, School Fix, QC Supply and Penn State.
Book a 30-minute assessment and tell us what you sell and where your data lives right now. And please get what's in Gary's head online before March.