How to show up in ChatGPT answers: what we did on our own website
Most of our new clients tell us they found us through ChatGPT. Here is what we did on our own site, what does not work and how to check where you stand.
Most of our new clients tell us they found us through ChatGPT. We never planned it as a channel and we have not paid a cent for it. But when we looked at what makes strigy.com different from an average site, it turned out we had done most of the work along the way, doing what we do for SEO anyway. This is the full list, along with what we do not know and what nobody can promise.
In short
- When ChatGPT searches, it reads the web live through its crawler, OAI-SearchBot. If your site blocks it, you will not appear in its search answers.
- AI recommends sites that make it clear who you are, what you do, where and for whom.
- Good SEO does most of the work. For AI you add FAQs, summaries at the top, the same facts everywhere and mentions on other sites.
- Nobody can guarantee a place in the answer, but you can measure it. Ask clients how they found you and track visits from chatgpt.com.
How ChatGPT gets to your website
When you ask someone for a recommendation, the answer depends on what they have heard. ChatGPT is similar, except it “hears” in two ways.
The first is what it has learned. The model was trained on a large part of the internet, but that knowledge has a date. It knows nothing about the site you launched last month.
The second is search. When a question needs fresh or local information, say “who builds booking websites in Split”, ChatGPT searches the web, reads a few pages and builds an answer from them, usually with links to the sources. OpenAI uses a crawler called OAI-SearchBot for this. Their documentation says sites that block it will not be shown in ChatGPT search answers.
For a small business the second route matters far more. You do not have to wait for the model to learn about you one day. It is enough for search to find you and for your site to say clearly what you do.

What we did on strigy.com
None of this is a trick. These are all things a good website should have anyway. The only difference is that AI reacts to them more strongly than Google does.

We opened the door. Our robots.txt does not block any AI crawler. It sounds minor, but plenty of sites block them, sometimes on purpose and sometimes because a security plugin set it up that way. If you do not want your content used to train models, OpenAI separates the two. GPTBot is for training and OAI-SearchBot is for search, so you can block the first and allow the second.
We answered the questions people actually ask. 52 of our pages have an FAQ block, 283 questions with answers in total. Each one is also marked up in the page code with FAQ schema, so a machine knows exactly what the question is and what the answer is. AI assistants like to quote exactly that kind of short, complete answer.
We put the summary on top. Every post starts with an “In short” box of three or four sentences. A reader in a hurry gets the answer straight away, and AI gets a ready summary it does not have to piece together from the whole text.
We wrote an llms.txt. It is a plain text file at strigy.com/llms.txt with facts about the studio: what we do, for whom, where, how long a build takes, how payment works and which pages matter. It is not an official standard and we do not know how much each assistant reads it. But it takes half an hour to write, and whoever reads it reads correct facts.
We made the facts match everywhere. Name, services and location should be the same on the website, the Google profile and social media. When we checked, we found two old listings under our name describing a marketing agency and services we no longer offer. Contradictions like that confuse people and AI alike, so update them or shut them down.
We report changes right away. After every release the site uses IndexNow to tell Bing and other search engines which pages changed. AI assistants take part of their data from search indexes too, so it helps when that data is fresh.
Content AI can recommend
The technical side opens the door, but content earns the recommendation. AI does not recommend a site because it looks nice. It recommends it when it can clearly read an answer to the question someone asked.
For us that means three kinds of pages. Service pages that say what we do and for whom. City pages, such as web design in Split, because people ask about their own town. And case studies with real numbers and a link to the client’s live site, because the best answer to “who builds booking systems” is an example you can open.
Write the way a customer would ask. If people ask what something costs, make sure the page answers exactly that, right in the first sentence. Why that is also good SEO, we explained in why your site is not on Google.
What does not work
This is where promises pile up quickly, so let us be honest.
A place in the answer cannot be guaranteed. Nobody controls what AI will say, not us and not anyone who promises it to you. Answers differ from user to user, depend on how the question is phrased and change from week to week.
Stuffing keywords does not help, and neither do hundreds of AI-written articles. That kind of content says nothing new, and an assistant is looking for exactly what is specific. One post with your real experience is worth more than twenty generic ones.
And the website is not everything. AI also weighs what others write about you, such as Google reviews, directory listings and articles that mention you. That part comes from working with clients over time.
How to check where you stand
You do not need a tool for this. You need about fifteen minutes.
Ask ChatGPT like a customer. Open a new chat and ask five questions your customer would ask, without your company name. Something like “who does … in …”, “how much does … cost” or “what is a good … for …”. Note whether you appear and who appears instead. Repeat it once a month with the same questions and you will see the trend.
Look at Google Analytics. Visits from ChatGPT usually show up as the source chatgpt.com in the traffic acquisition report. The number will not be large, but these are people someone has already recommended you to.
Ask your clients. It is the simplest and most accurate method. We added an optional “How did you find us?” question to our contact form. In a few months we will know how many enquiries really come from ChatGPT instead of guessing.
If you would like us to do this for your site, it is part of the SEO optimisation we offer. Send your website address to info@strigy.com. First we will tell you what AI can currently read about you, and what it cannot.
Questions we hear often
Can anyone guarantee I will show up in ChatGPT?
Nobody can, because nobody controls what AI will answer, and answers differ by question and by user. You can only improve your chances with a site AI can read, clear answers to customer questions and correct facts everywhere.
Do I need an llms.txt?
You do not have to, but it does no harm. It is a simple text file with facts about your business. It is not an official standard and not every assistant reads it, but it takes half an hour to write, and whoever reads it reads correct facts.
Is this different from SEO?
Mostly not. A technically sound site, clear content and speed help both Google and AI. For AI you add FAQs with answers, summaries at the top, the same facts on every profile and mentions on other sites.
Will blocking AI crawlers protect my content?
Partly, and at a price. OpenAI separates GPTBot, which is used for training models, from OAI-SearchBot, which is used for search. If you block GPTBot and allow OAI-SearchBot, their documentation says your content is not used for training, and you can still appear in search answers.
How do I know if clients come from ChatGPT?
Ask them, it is the most accurate way. Visits from ChatGPT also usually show up in Google Analytics as the source chatgpt.com. We added an optional question to our contact form asking how people found us.
Want to know what AI sees on your site?
Send us your website address. We will send back a short overview of what ChatGPT can read about you and what we would fix first.
Start a project