ChatGPT has no directory you can sign up to. When someone asks about a provider, the model draws on what it saw during training and, for current questions, searches the web live. It names businesses that can be found there with clear answers to specific questions, consistent company details and mentions on other sites. There is no submission process and no paid slot in the answer.
How ChatGPT learns about company names in the first place
There are two routes. The first is training: at some point, the model saw a large part of the public web, and that snapshot is static and often months old. The second route is live search. If you ask about a fiduciary office in Kreuzlingen, ChatGPT searches the web, reads a handful of pages and summarises them. Perplexity always does this and shows its sources openly, while Google's AI answers build on the normal Google index.
This leads to the most important point: classic discoverability is the foundation. A page that Google doesn't know about, or isn't allowed to crawl, can't be cited in an AI answer either. AEO doesn't replace SEO, it builds on it.
The difference from a results list is the quantity: instead of ten links, an answer names two or three providers. On top of that, answers vary depending on the session, language and region, because live search finds different pages each time. Exactly how the models weigh things in detail is something nobody outside OpenAI, Google and Perplexity knows for sure, not even the agencies that sell it to you.
There is no sign-up and no advertising slot in the answer
Offers such as We'll register your business with ChatGPT are selling something that doesn't exist. There's no form, no fee and no list you get added to. The fact that OpenAI is experimenting with advertising formats on the side doesn't change this: that would be purchased space next to the answer, not a purchased recommendation within the answer text. Anyone who promises you a guaranteed spot in ChatGPT can't deliver on it.
What you can genuinely control technically is access. Check your robots.txt: if it blocks GPTBot or OAI-SearchBot, ChatGPT won't take your pages. The same applies to PerplexityBot. For Google, Google-Extended controls whether your content is used for Gemini, but not the AI answers in search: those follow the normal Googlebot. Many businesses never consciously set these blocks, they were already in the provider's template.
A second common obstacle: content that only appears after JavaScript has loaded, or that sits behind a login. A quick test is to open the page without JavaScript. Whatever remains blank doesn't exist for most of these systems.
Content that answers a specific question
Models don't cite websites, they cite passages of text. That's why a simple pattern works best: one page per question, the question as the title, the answer in the first two or three sentences, the details below. A page titled Services doesn't provide a citable passage. A page titled What does inheritance advice cost in Switzerland does.
Specific beats general. Price ranges, durations, deadlines, cantonal differences, figures from your own business. The sentence In Thurgau, a deadline of 30 days applies is citable, the sentence We offer individual advice is not. State which point in time a figure refers to, and it will still be useful a year from now.
You don't have to guess which questions these are: they're the ones that come up again and again on the phone and in the first conversation. BlogPilot is built exactly on this pattern, one post, one question, the answer at the top.
Structured data and consistent company details
Structured data following schema.org helps machines classify a page correctly: Organization or LocalBusiness with name, address, phone number and website, FAQPage where there genuinely are questions and answers, Article with author and date. To be honest, no provider has confirmed that schema markup directly increases your chance of being mentioned. What is proven is that it improves how the page is understood by classic search systems, and those systems supply the material for AI answers.
Consistency matters at least as much. Write your company name, legal form, address and phone number the same way everywhere, with the number in +41 format. The sources that Swiss company data typically comes from are the commercial register via Zefix, local.ch, search.ch, the Google Business Profile and industry directories. If three different addresses appear there, the model picks one of them, and not necessarily the right one.
This is unglamorous work, but it pays off. An hour spent aligning five directory entries often achieves more than three new blog posts.
Mentions on sites you don't control
Language models weight what others write about you more heavily than your own self-description. Answers to questions like the best carpenters in eastern Switzerland strikingly often come from list articles, comparison portals and directories, not from the websites of the businesses named. Useful sources therefore include expert articles on industry portals, interviews in regional media, member pages of associations and trade bodies, references on customer sites, and reviews with genuine text.
In practice, that means: try to get into existing lists rather than writing your own. A call to the editorial team of a regional comparison portal is more effective than another post on your own blog. How long it takes for a new mention to show up in answers is unclear. Sometimes it takes weeks, sometimes nothing happens, and there's no guarantee either way.
What doesn't help, and how to check the effect
What doesn't help: stacking keywords, a dedicated AEO page with 200 questions and generic answers, hidden text just for crawlers, and paid placements. Some caution is needed with llms.txt. The file is currently being sold as essential, yet no major provider has publicly confirmed that it actually reads it. Creating one takes ten minutes and doesn't hurt, but it isn't a strategy.
You can only measure this roughly, and it's worth saying so openly. Once a month, ask the five questions yourself that you want to be mentioned for, in ChatGPT, Perplexity and Google's AI mode. Note whether you appear and which sources are cited. That list of sources is your to-do list: that's where you want to be.
It's also worth checking your access statistics for referrals from chatgpt.com and perplexity.ai. The numbers are small, because many users don't click through after getting an answer. There's no ranking tracking like in SEO, and anyone who sells you an exact AI visibility score is calculating a model, not measuring reality.
Frequently asked questions
Can I register my business with ChatGPT?
No. There's no sign-up form and no fee to get listed. What you can do is make sure your content is crawlable, answers specific questions, and appears on third-party sites.
How long does it take for ChatGPT to know my business?
For live search, it's enough for your page to be in the standard search indexes, which takes days to a few weeks. Trained knowledge, on the other hand, lags months to years behind. No provider gives a reliable timeframe, and no one outside can put a number on it either.
Do I need an llms.txt file?
Not as things stand today. No major AI provider has confirmed that it reads this file. It's more important that your robots.txt doesn't block AI crawlers and that your content is readable even without JavaScript.
Does my Google Business Profile help with AI answers too?
Indirectly, and quite a lot for local questions. It provides verified business details, opening hours and review text, which show up in many answers to questions about providers nearby. What matters is that this information matches your website and directory listings.
This page answers a single question. If you want to know how visible your business currently is, the free AI check shows you that in about a minute.