How DeepSeek finds and cites brands
In short
DeepSeek’s app can search the web, but the model Agentcites asks has no search of its own. So Agentcites gives it Exa, a separate search index, and says so on every answer. This page explains what that means, which pages DeepSeek can find this way, and what a brand can do about it.
How Agentcites asks DeepSeek
Agentcites sends each of your buyers’ questions exactly as written. Nothing of ours is added: no instruction to mention you, and no hint about what we hope to hear.
The question goes to a DeepSeek model through OpenRouter, a service that reaches many AI providers through one account. DeepSeek’s own app can search, but its model offers no web search to anyone asking it this way. Exa is the only way it answers from the web at all.
Every answer is stored word for word, with every link it gave, the model that wrote it and the search it used. The screens say Exa, not DeepSeek’s own search.
How DeepSeek uses the web
Exa is a search index of its own, separate from Google, Bing and the search inside DeepSeek’s app. When Agentcites asks DeepSeek, Exa finds the pages and DeepSeek writes the answer from them.
This is a real difference, and we would rather name it than hide it. A buyer using DeepSeek’s app may get an answer built on different pages. What Agentcites measures is DeepSeek’s model answering from the web through Exa.
Which robot reads your site
Because the search here is Exa’s, the robot that matters is whichever one builds Exa’s index, not one of DeepSeek’s.
We have not found a page in Exa’s documentation that names a crawler a site can allow or block. So we do not name one here. If that changes, this page will link to it.
What a site can control is the same as for any search index. A page that is public, readable without a login, and not blocked for all robots is a page an index can reach.
What you can do to be cited by DeepSeek
- Keep the pages that answer buyer questions public. A page behind a login or a script that never loads cannot be read by any index.
- Do not block all robots in robots.txt. A rule for every robot also blocks the ones you have not heard of.
- Answer the question in plain words on the page. DeepSeek can only repeat what the page Exa found actually says.
- Be named on other sites, not only your own. Exa may find a review or a list before it finds you.
- Compare DeepSeek with the other three. Because its search differs, a gap that shows only here points at Exa’s index rather than at your page.
Questions
Why not use DeepSeek’s own search?
Its model, asked the way Agentcites asks, offers no web search. The search lives in DeepSeek’s app, and there is no public way to ask that. Exa is the only way to have it answer from the web.
Does that make the DeepSeek answers less useful?
They answer a slightly different question: what DeepSeek’s model says when it reads the web through Exa. That is still worth knowing, and Agentcites labels every one of these answers so it is never mistaken for DeepSeek’s app.
Which robot should I allow for DeepSeek?
We have not found a crawler name for Exa in its documentation, so we cannot point you to one. Keeping the page public and not blocking every robot is what a site can do.
Can I see which of my pages DeepSeek used?
Yes, when it gave a link. Agentcites keeps every link with each answer and marks which are your pages.