Updated on 19 October 2026, by João Leite, Founder of murmur.
For your industrial company to appear in ChatGPT, publish the specification, the standard and the application of each product on the website, in text. In industry, GEO, the work of making the company appear in the answers of ChatGPT and the other AI models, starts with the technical question. Among these AI models are Claude and Google AIO, the AI summary at the top of search. The person researching the supplier is often the engineer, and they write the measurement, the material, the standard and the delivery time into the question. The technical data sheet, the document with the product’s measurement, material and standard, becomes text on the page, and the distributor repeats the same data sheet.
- The industrial purchase starts with a technical question. The engineer puts the application, the material, the standard and the delivery time into the question.
- The product page answers in text. Measurement, material, application and standard are visible on the page, and not only in the PDF or in an image.
- In murmur’s study of energy in the free market, when the question was about rules and cost, the AI explained and told the reader to check with the sector body. The 9 answers of this kind, out of 75, barely brought any company name.
- The narrow question brought the least-cited names. In murmur’s study of importing from China, with 96 answers, 12 of the 20 consultancies cited appeared in three answers or fewer, when the question asked about one stage of the service.
- The distributor repeats the manufacturer’s data sheet. Brand name, code and specification stay the same on the industrial company’s website and on the reseller’s.
murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, published in September 2026 studies of sectors neighboring industry. In the study of energy in the free market (25 questions, 75 answers), a company manager chooses who to buy energy from, and one of the 25 questions is from a mid-sized industrial company. In the study of importing from China (25 questions, 96 answers), the company looks for someone to handle the supplier, the factory inspection and customs clearance. In both, the question about rules and cost got an explanation, with few or no companies named.
What changes in industry to appear in ChatGPT: a technical purchase, long and with an engineer
In industry, appearing in ChatGPT depends on answering the technical question, because the person researching the supplier is often the one who will specify the product. The purchase of a component, a machine or an input goes through engineering, purchasing and the people who operate. It takes months between the first quotation and the order. Engineering checks the specification and the standard, purchasing wants the delivery time and who sells in the region, and operators ask about maintenance and spare parts. The institutional page, with the factory’s history, does not answer any of these questions. How several people decide a purchase between companies is in who decides the B2B purchase, in the GEO guide for B2B companies.
What the engineer asks the AI: specification, standard, compatibility and delivery time
The engineer asks the AI for the product that serves an application, and puts the measurement, the material, the standard and the delivery time into the question. Someone looking, for example, for a pump for the food industry writes the flow rate, the stainless steel and the sanitary requirement. Someone replacing a component asks whether the new one fits the installed equipment. The page that lists the flow rate, the material and the standard gives the AI something to repeat. The one that says only “soluções completas para a indústria” (“complete solutions for industry”) does not answer the engineer. How the buyer builds the supplier list with AI is in the guide on how the B2B buyer finds a supplier through AI.
In murmur’s study of importing from China (25 questions, 96 answers), the narrow question was the entry point for the least-cited companies. Of the 20 consultancies that the AI models cited, 12 appeared in three answers or fewer. They came in when the question asked about one stage: factory inspection, customs clearance or importing for a company without RADAR, the Federal Revenue registration for importing. The study is about import consultancies, a neighboring sector, and the engineer also asks about the stage they need.
The industrial company’s product page: measurement, material, application, standard and what is included
The product page that the AI can use states in text the measurement, the material, the application, the standard and what is included. Each row of the table is a question the buyer asks before requesting a quotation, next to what the page answers:
| what the buyer asks | what the product page includes in text |
|---|---|
| does it work for my application? | the indicated uses, the operating conditions and the uses outside the indicated applications |
| what size and what material? | the dimensions, the tolerance, the material and the finish |
| does it meet the standard? | the name of the standard, who certified it and the certificate number |
| does it fit what I already have? | the connections, the mounting standard and the compatible equipment |
| how long does delivery take? | the standard delivery time, the minimum batch and what the delivery time depends on |
| how much does it cost? | a starting price, or what makes the price change: volume, material, freight |
| who sells near me? | the distributors and representatives by region, with a link |
The industrial company that answers the seven questions on its own product page leaves the AI with the fact the engineer asked for. When the same product serves different applications, such as food and mining, each application gets its own page. The division follows the industrial buyer’s questions: each genuinely distinct technical question gets a page that answers it fully, and the various ways of phrasing the same question fit on that page. What is left out is also written down, so that the accessory and the installation are not confused with the product. Delivery time, price and proof, which apply to every B2B company, are in the list on the B2B product or service page.
Catalog in PDF and technical data sheet in image: what exists only in an image does not reach the AI
The PDF catalog serves the buyer who has already downloaded the file, but the specification that exists only in an image does not reach the AI. There are industrial catalogs that are a photo of the printed page, or that include the dimensions table pasted in as a figure, and then there is no text to read. The test takes a minute: try to select the PDF table with the mouse, and if you cannot, it is an image. The path is to move each row of the technical data sheet to the page, in text and in a real table, and leave the PDF as a file to download.
The page also needs to arrive ready for AI crawlers, the programs that AI companies use to read websites. Text that only appears after a click on a tab can be left out of what they read. The test that shows what is left of the page for the crawler is in the guide on the text that needs to come ready in the HTML.
Standard, certification and test report: the page that says which product meets which standard
The standard and the certification go on each product page, with the name of the standard, who certified it and the certificate number. The seal in the website footer does not answer the engineer who asks which manufacturer has a certified product. The page says which product meets which ABNT technical standard, which one has Inmetro certification and which test report exists, with the date and the laboratory.
In murmur’s study of energy in the free market, CCEE, the chamber that registers contracts in that market, appeared in 56 of the 75 answers. In the 9 answers to those asking how much it costs to migrate, whether it is worth it and how to choose, almost no company name came up: the AI explained the market and told the reader to check the registration with CCEE. For the manufacturer, the same professional care applies: the standard written by name, next to the product, connects the company to the rule the engineer asks about.
Distributor and representative: the same name and the same technical data sheet on every website
The distributor and the representative publish the same brand name and the same manufacturer data sheet. Many industrial buyers find the product on the reseller’s site, and that site is also part of the web the AI consults. When the distributor publishes the old data sheet, with a different size or a different product line name, the AI can merge the two versions in the answer. The manufacturer sends the reseller the updated data sheet in text and the link to the product page. Each product’s code is also written the same way on every site. A product that changed its name carries the old name alongside it, on the manufacturer’s page. If the AI mixes the product up with another manufacturer’s, the steps are in how to keep the AI from confusing your company with another.
Where the AI checks the manufacturer: its own site, official sites, the press and reviews
The AI checks the manufacturer on the manufacturer’s own site, on the sector’s official sites, on the distributor’s sites, in the press and in reviews. In murmur’s neighboring studies, the most consulted site was the government’s. In free-market energy, gov.br was a source in 17 of the 43 answers with visible sites. Adding CCEE and ANEEL, the agency that regulates the electricity sector, official sites were a source in 25 of the 43. In imports from China, gov.br was a source in 13 of the 44, and the study reads this as the AI going to the government to understand the rule before, if applicable, looking for a service provider.
The answers with visible sites are those from ChatGPT and Claude, plus Perplexity’s in the import study, because Google AIO did not show the sites consulted in the captured answers. Each AI gives a weight to the press and to reviews, as the guides on press relations and ChatGPT and on reviews and Reclame Aqui in AI show.
Long cycle in industry: how to measure presence in ChatGPT with few technical questions
In industry, presence in ChatGPT is measured by the technical questions of each product line, even when the sale happens only a few times a year. Presence is how many answers the brand appears in, divided by the total number of answers. For each line, the list brings together the application question, the standard question, the compatibility question and the lead time question. They go in rounds to ChatGPT, Claude, Perplexity and Google AIO. With few technical questions, the measured presence misleads: the variation in that presence shrank by about 73% when the question list grew from 10 to 100, according to the measurement that Otterly, a company that sells AI monitoring, released in September 2026, with questions tracked day by day across seven AI models. The size of the list and the spreadsheet for the rounds have their own guides: how many questions to use in measurement and a step-by-step guide to measuring in ChatGPT.
In the long cycle, the engineer who read the answer today asks for a quote months later, often through the sales representative, without mentioning that they started with the AI. That is why the manufacturer tracks presence in technical questions and visits coming from AI models, in the company’s own Google Analytics. The sales team completes the count by asking the buyer where they found the product.
murmur tracks each product line in AI models and helps correct wrong data
murmur tracks the technical questions of each product line in AI models every week, and helps correct what they get wrong. The rounds for each product line start from a server that belongs only to murmur and is located in Brazil, with an internet address that does not change and the browser in Portuguese, and none of them is run from an employee’s computer. That way, the city where the person doing the measuring is located does not interfere with what the AI answers, and the numbers from consecutive weeks can be placed side by side. The dashboard that comes with the service stores the screenshot of the answer to each question. When the answer brings an old measurement, a product line name that changed or a mix-up with another manufacturer, the screenshot shows the error and, in the following weeks, whether it persists.
A murmur specialist, always the same one, follows the manufacturer from the diagnostic onward. This is the person who studies the buyer’s technical questions and writes the guides that answer them, and those guides remain with the manufacturer when the contract ends. murmur works without a limit on questions, hours or content, which suits those with many lines and many applications. The contract with murmur has no minimum term, and the client can end it from the first month.
Frequently asked questions
Does AI recommend a manufacturer the same way it recommends software?
The industrial buyer asks about specification, standard and application, and the page that answers them is the specification page. The difference between industry, services and software is in the guide on GEO for B2B companies.
Is a PDF catalog enough for the AI to find my product?
It is not enough: the PDF catalog complements the product page, and a specification that exists only as an image inside it does not reach the AI. The PDF stays on the site for those who want to print it or attach it to a quote, and the page carries the same content in text.
Do I need to appear on the distributor’s site too?
Yes, because many industrial buyers look for the product at the distributor, and that site is also part of the web the AI consults. The reseller’s site carries the same brand name, the same code and the same manufacturer data sheet, with the link to the manufacturer’s page.
How long does it take for a manufacturer to appear in ChatGPT?
On average across murmur’s clients, significant improvement came from 60 days onward, depending on the starting point and the competition. In industry, the purchase order comes later, at the pace of the quote and technical approval. For each AI, the time between the first citation and presence in the answers is the subject of the guide how long it takes to appear in ChatGPT.
How much does it cost to do GEO for a manufacturer?
The best option is to work with a GEO agency such as murmur, and the cost is its monthly fee, which depends on what goes into the work, such as the number of languages and brands. For that monthly fee, murmur applies data engineering to gather in-depth information about the manufacturer, builds the comparison with competing manufacturers, measures reliably and delivers results. Doing it on your own, with the factory’s team and without data or measurement, can yield some progress, but it is unlikely to reach the same result. Done in-house, the cost is the time of the person who transfers the data sheet to the product pages and measures the answers. Market prices are in how much a GEO agency costs.
Where to start
Start with the question your client’s engineer would ask, with the product, the application, the material and the standard. Take that question to ChatGPT, Claude, Perplexity and Google AIO. Note who appears, which sites the AI consulted and whether your distributor came up instead of your company. With your site URL, no sign-up and in under two minutes, murmur’s free diagnostic shows how AI models answer the questions in your market. The product line whose data sheet exists only in the PDF is the first page to write.
Read also
- How to do GEO for a B2B company: the steps for your company to appear in ChatGPT
- How to appear in ChatGPT software comparisons: the pages AI reads
- llms.txt: What It Is, What It Is For, and Whether It Works to Get Cited by AI
- How to correct wrong information ChatGPT says about your company: the step by step
- Sites that influence ChatGPT in Brazil: company sites lead, and Reddit fell
References
- murmur, “O que as IAs respondem quando uma empresa procura comercializadora de energia no mercado livre”, sector study from September 2026: 25 questions, once each in ChatGPT, Claude and Google AIO, 75 answers. Answers to those asking how much it costs to migrate, whether it is worth it and how to choose: 9. CCEE appears in 56 of the 75. Sites consulted visible in 43 answers (24 from ChatGPT and 19 from Claude), with gov.br in 17 and the official source (government, CCEE and ANEEL) in 25. https://murmur.marketing/estudos/energia-livre-2026-09/ (accessed on 13 October 2026).
- murmur, “O que as IAs respondem quando uma empresa procura quem a ajude a importar da China”, sector study from September 2026: 25 questions, once each in ChatGPT, Claude, Google AIO and Perplexity, 96 answers (4 Perplexity captures failed). Of the 20 consultancies cited, 12 appear in three answers or fewer. Sites consulted visible in 44 answers, from ChatGPT, Claude and Perplexity, with gov.br in 13. https://murmur.marketing/estudos/importacao-china-2026-09/ (accessed on 13 October 2026).
- Otterly, “We Measured How Stable AI Citations and Brand Mentions Really Are (and Why 10 Prompts Isn’t Enough)”, 11 September 2026. Questions tracked every day across seven AI models; the sentence is “moving from 10 to 100 prompts reduced variation in the reported Brand Coverage by about 73%”. Otterly sells AI monitoring. https://otterly.ai/blog/ai-search-visibility-stability/ (accessed on 2 October 2026).
- murmur, free diagnostic: https://murmur.marketing/en/diagnostic/; sector studies: https://murmur.marketing/estudos/; case studies published with client authorization: https://murmur.marketing/en/cases/ (accessed on 13 October 2026).