Industry Pains

Why ChatGPT Does Not Recommend My Company

Because there is no text out there tying your company name to the problem the buyer is describing. The model does not recommend the best factory — it recommends the one it can quote from. If your website only says you have tradition and quality, and no third-party page ever wrote your name next to the application you serve, there is nothing for the model to pull. Below: how the answer is actually assembled, the five reasons a Brazilian factory gets left out, and a twenty-minute test that tells you where you stand today.

What the Model Needs From You

2 routes
The answer comes from training memory or from a live search — you have to be present in both, and they are won differently
3rd party
A mention on someone else's site weighs more than the same claim on your own — self-praise is not evidence for a model
10 questions
What the twenty-minute test costs: ten buyer questions across three AIs, logged in a spreadsheet
3-6 months
Realistic window between publishing extractable answers and seeing the first consistent citations

How the answer is actually assembled

The Short Version

The model does not rank companies. It assembles a paragraph out of text it can quote — and only names whoever appears inside that text.

There are two routes into the answer, and they behave differently. The first is what the model absorbed during training: text published on the open web long enough ago to be part of what it read. The second is the live search the assistant runs while answering, which pulls pages right now and quotes them with a link. Winning the second route is faster and measurable; the first takes longer and depends on your name showing up across many independent sources. A company absent from both simply cannot be named — not because it is worse, but because it is not written anywhere the model can reach.

One consequence surprises most owners: the model prefers text that already reads like an answer. A page that opens with two direct sentences stating the criterion, the number or the trade-off gets quoted far more often than a page that spends four paragraphs on company history before reaching the point. It is not a style preference — it is what fits inside the paragraph the model has to produce.

The model is not choosing the best supplier. It is choosing the sentence it can paste. Whoever wrote that sentence gets the recommendation.

The five reasons a factory gets left out

Reason 01
Your Name Never Appears Next to the Problem
You make gaskets for the food industry, but no sentence anywhere on the web contains your company name and the words "gasket", "food industry" and "sanitary standard" together. The model has no way to associate the two. Fix the association explicitly: write pages where the application and the brand sit in the same paragraph, repeatedly and factually.
Most Common
Reason 02
The Content Exists but Not as Text
The specification is in the catalogue image, the application is in the video, the case study is in the printed folder. The model reads text. Anything living only inside an image, a PDF diagram or a video without a transcript is invisible to it — even when your customers love that exact material.
Format
Reason 03
Zero Independent Mentions
Everything ever written about you was written by you. No sector association listing, no trade publication, no supplier directory, no customer publishing the case. A model treats convergent independent sources as confirmation; a single self-published source is a claim. Two or three real third-party mentions change the picture more than ten extra pages on your own site.
Authority
Reason 04
The Page Sells Instead of Answering
Institutional copy — solidity, partnership, excellence — offers nothing quotable. Compare with a page that states "for CIP washing above 85 °C, EPDM outperforms nitrile; nitrile hardens and cracks in about six months of that cycle". The second sentence can be pasted into an answer. The first cannot.
Editorial
Reason 05
The Crawler Is Blocked or the Page Is New
Some sites block the AI crawlers in robots.txt without anyone deciding to — it came with the template or with a security plugin. Others publish good content that is simply three weeks old and has not been picked up yet. Check the file, and give a new page a few weeks before concluding it failed.
Technical

If your pages are not on Google either, none of the above is your first problem — indexing is. Sort that out first in why my company does not show up on Google, then come back here.

Measure where you stand before hiring anyone

Do not accept a diagnosis you cannot reproduce. This test costs twenty minutes and a spreadsheet, and it gives you the baseline against which any future work has to be measured:

Baseline Protocol

Write Ten Buyer Questions
Never your brand name. Use what the buyer types: "who makes stainless conveyors for meat processing in Santa Catarina", "best supplier of industrial hoses for high pressure". Take them from the calls your sales team actually receives.
Ask Three Assistants
The same ten questions in ChatGPT, Gemini and Perplexity, each in a fresh session with no history. Different engines pull from different sources — one of them may already cite you while another has never heard of you.
Log Three Columns
Cited with a link, mentioned without a link, absent. Also write down who did get cited — that list of competitors is the map of which pages you have to beat, question by question.
Reading the sheet:
Absent in all thirty cells is the normal starting point for a Brazilian factory — it is not a catastrophe, it is a baseline. Mentioned without a link means the model knows your name but has no page to point at, and that gap closes fastest. Cited in one engine and absent in the other two usually means you have one good page and no consistency around it. Repeat the same ten questions every quarter, always in a clean session, or the numbers are not comparable.

To turn this manual sheet into a recurring measurement with tracked wording, competitor share and quarter-over-quarter movement, see how to find out what ChatGPT says about your company and the share of model audit.

What actually changes the answer — and what does not

What changes it: publishing pages that answer one buyer question each, opening with two direct sentences, with numbers, criteria and named trade-offs; getting your name written on independent sources your sector already reads; and keeping it up long enough for the pattern to be consistent rather than a one-off. What does not change it: buying followers, posting more often on social media, redesigning the site, or paying anyone who claims they can insert your brand into a model. There is no paid placement inside the answer.

This discipline has a name — AEO, optimising for the answer rather than for the link list. The name matters far less than the habit: write down what your best salesperson explains on the phone, and publish it as text.

The full method, with the production sequence and the realistic timeline, is in industrial AEO. If the competitor already shows up and you do not, the comparison between the two profiles is in my competitor shows up in ChatGPT and I do not. For the practical checklist of what to publish first, go to how to get cited by AI in B2B. The whole pain sits in my company does not show up on Google or in AI.

FAQ

Why does ChatGPT not recommend my company? +
Because no text available to the model ties your company name to the problem the buyer described. The model quotes what it can read; it does not evaluate which factory is better. Institutional copy about tradition and quality gives it nothing to paste, and with no independent mention there is no second source to confirm you.
Does ChatGPT use Google to answer? +
Not directly. It answers from training memory and, when it needs current information, runs its own web search with its own crawler and index. That is why ranking first on Google does not guarantee a citation, and why a page nobody clicks on Google can still end up quoted in an answer.
Can I pay to show up in ChatGPT answers? +
No. There is no ad slot inside the generated answer and no way to buy a position in it. Anyone selling brand insertion into a model is selling something that does not exist. What works is publishing extractable content and earning independent mentions — slower, and it stays.
Does ChatGPT know small Brazilian manufacturers? +
It knows whoever is written about on the open web, regardless of size. A hundred-employee factory with thirty technical pages and two sector-publication mentions gets cited more than a large group that publishes nothing. Size does not enter the calculation — available text does.
My site is new. How long until an AI cites me? +
Engines with live search can pick up a new page in weeks once it is indexed. Consistent citations across several assistants take between three and six months of steady publishing. Appearing in the training memory itself is slower and depends on the next training cycle — which is why the live-search route is the one worth attacking first.
Do I need a Wikipedia page for ChatGPT to cite me? +
No, and most Brazilian manufacturers would not meet the notability criteria anyway. What that page would provide — an independent source stating who you are and what you make — can come from a sector association listing, a trade publication article, a supplier directory or a customer publishing the case. Two or three of those do the job.
Does blocking the AI crawler in robots.txt hurt me? +
Yes, if your goal is to be cited. A blocked crawler cannot read your pages and therefore cannot quote them with a link. Open yourdomain.com.br/robots.txt and check for rules against GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended — many sites carry these by accident, inherited from a template or a security plugin.
Does posting on social media help me show up in ChatGPT? +
Barely, and never as the main lever. Most social content sits behind logins or in formats the crawlers do not read well, and it disappears from the feed in days. A technical article published on your own indexable domain, or a mention on a sector site, weighs much more than the same text posted as a carousel.

Want to know what the AIs say about you today?

We run your buyer questions across the main assistants, log who gets cited in your place, and hand you the list of pages to publish in order of impact. You get the spreadsheet either way, whether or not we work together afterwards.

Run My AI BaselineFree diagnosis, no commitment