How to show up in ChatGPT: why it names some brands and not others
When someone asks it about your industry, ChatGPT doesn't pull brand names out of thin air: it searches the web, reads a handful of pages and summarizes what it read. Here is what is known about how that works, what you can do about it and how we measure it, with the exact model we use written out plainly.
How does ChatGPT decide which brands to name?
When a question involves search, ChatGPT retrieves web pages at that moment, summarizes them and links to the domains it took the information from. If your site isn't among those sources, there is no way for it to name you through that route.
What is known for certain, because you can see it in the answer itself, is this: when it answers with search on, a list of linked sources appears, and the text of the answer closely follows what those pages say. In other words, it retrieves first and writes second, not the other way around.
The practical consequence follows from that: what matters is not so much the model as the set of pages it ends up reading for that specific question. Comparisons, industry lists, directories, specialist publications, forums and, when the brand already has a place there, its own website.
What we won't tell you is how OpenAI scores each source internally or how much weight each one carries. That isn't public, and anyone who gives you a percentage is making it up.
If half of these terms are new to you, they are all explained in the glossary.
What can you do to show up more in ChatGPT?
Be on the pages ChatGPT consults when people ask about your industry, and have direct answers on your own site to the buying questions people ask.
Answer the buying question, not the keyword
A page that opens by answering "which X is right for Y" in two sentences is material for a summary. A page that circles for two hundred words before committing to anything isn't. It's the difference between something that can be quoted and something that has to be interpreted.
Put the facts where they can be read
Prices, service areas, timelines, what's included and what isn't. In the text and in structured data. If that information only exists in a PDF, in an image or behind a form, then for whatever is doing the summarizing it doesn't exist.
Get onto the pages that already get cited
The industry lists and comparisons that show up among the sources are the ones that actually move the needle. Our report tells you which domains the engines are citing in your industry, which is where you start.
Repeat: one answer is not the answer
ChatGPT doesn't answer the exact same way twice. Measuring on a single day tells you nothing. What counts is the trend across many questions and many runs.
How do we measure it?
We send it your set of buying questions with live search on and read the answers. The specific model we query for ChatGPT is openai/gpt-6-sol, with OpenRouter's web search plugin.
Model we query for ChatGPT
openai/gpt-6-sol:online
Via OpenRouter, with the web search plugin turned on (the :online suffix). It isn't the consumer app, it's the model through the API.
Here is the uncomfortable part, and we say it because it seems like the least we can do: we don't query the ChatGPT app. We query OpenAI's GPT-6 Sol model through OpenRouter, with the web search plugin turned on, which OpenRouter marks with the :online suffix.
That means the answer we measure comes from the same OpenAI model, with internet access at the moment of asking, but it doesn't go through the product layer a person uses on chatgpt.com: not its own search, not its interface, and not OpenAI's decisions about what to show and in what order.
When the model doesn't respond, the fallback for ChatGPT is openai/gpt-5.6-sol, so the engine is always an OpenAI model or there is no data point.
What is comparable is what we look for: whether your brand appears among the options recommended for a buying question, in what position, in what tone and alongside whom. And we measure that the same way month after month, which is what makes the series useful.
Does your site let ChatGPT in?
It depends on your robots.txt. OpenAI declares three separate agents, and blocking one is not the same as blocking another: some sites keep out the one that indexes without knowing it.
- GPTBotCrawls your site for OpenAI.
- OAI-SearchBotIndexes your site for ChatGPT search.
- ChatGPT-UserVisits your site when someone asks ChatGPT about you.
It's a much more common mistake than it seems, and it's almost never intentional: an SEO plugin added it, it came with the theme, or the hosting provider left it there. The checker reads your robots.txt agent by agent and shows you the exact rule that decides, not a vague score.
Check my site for freeDo you show up, or does your competitor?
The free check runs 6 buying questions through ChatGPT and the other three engines, and shows you how many times you show up, in what position and who shows up in your place. No card required.
I want my free diagnosis