How does DeepSeek's web search get its sources?
Very little is published: DeepSeek's API documentation says its Search API is powered by an in-house search engine and returns a URL, site name and date for each result; the Fetch API retrieves a single page as Markdown and is subject to the target site's access policies, so URLs blocked by site policy cannot be fetched. All of this is from DeepSeek's API documentation, checked 2026-11-06.
That page documents the API for developers. Whether web search in the app uses the same engine, how it picks sources and how many it cites is not stated officially: unverified. DeepSeek's app announcement (2025-01-15) only lists "web search" among its features.
What has DeepSeek not published?
| Question | Status on 2026-11-06 |
|---|---|
| The search engine's index coverage and ranking | Not stated |
| Crawler name, and whether it obeys robots.txt | Not stated; a third-party registry calls the crawler DeepSeekBot, but no first-party document was found: unverified |
| How sources are cited in answers | The API documentation gives no citation format |
| Whether the app and the API share one search | Not stated |
We do not guess at what is unpublished. The checklist below only contains things that hold for any web-connected AI and that you can verify yourself.
What can an independent developer do now?
1. Do not block crawlers you do not recognize
The Fetch documentation says access policies affect it. Check whether robots.txt or your CDN's bot rules block unknown crawlers wholesale; how to write it is in How do you write robots.txt for AI crawlers?. The simplest allow rule:
User-agent: * Allow: /
2. Put the body text directly in the HTML
Fetch returns the page converted to Markdown. Whether DeepSeek runs JavaScript is not stated officially, so the cautious approach is to make price, features and a one-line description visible without scripts; see Do AI crawlers render JavaScript?.
3. Say what it is, who it is for and the price in the first screen
The result fields include site name and date, which suggests the site name and update time travel with the result. Put the product name, positioning and update date on the page. This is inference, not an official statement.
4. Make a Chinese page (if you target Chinese users)
If your target users ask in Chinese, the Chinese page needs its own URL and a Chinese title; do not rely on browser auto-translation. This is inference: DeepSeek does not say how it handles multilingual pages.
5. Ask DeepSeek yourself with web search on, three times per question
Note whether it mentions you and which pages it cites. Look at how many of the three runs include you, as described in How do you measure AI visibility? Four metrics and a table you can follow.
How does this differ from ChatGPT and Claude?
| ChatGPT | Claude | DeepSeek | |
|---|---|---|---|
| Crawler names published officially | Yes (OAI-SearchBot and others) | Yes (Claude-SearchBot and others) | Not found |
| Official explanation of what robots.txt does | Yes | Yes | Not found |
| Official statement on the search engine | See the ChatGPT guide | Not stated officially | API docs say in-house search engine |
| What you can do | Allow the crawlers, put body text in the HTML | Allow the crawlers, put body text in the HTML | Allow, put body text in the HTML, test the rest yourself |
ChatGPT details are in How do I get ChatGPT to recommend my product? and Claude in How Do I Get Claude to Recommend My Product?. The Chinese assistants Doubao and Kimi are in How do I get Doubao and Kimi to recommend my product?.
FAQ
Which search engine does DeepSeek's web search use?
DeepSeek's API documentation says the Search API is powered by an in-house search engine (checked 2026-11-06). Whether the app uses the same one is not stated: unverified.
What is DeepSeek's crawler called, and should I allow it?
The official documentation does not name a crawler. A third-party registry says DeepSeekBot, but no first-party document was found: unverified. The cautious approach is not to block unknown crawlers wholesale.
Can I submit my URL to DeepSeek to get indexed faster?
We found no official submission channel: unverified.
Can I be guaranteed a DeepSeek recommendation?
No. The ranking is not published and answers differ each time; you can only watch how often you appear across repeated questions.
Sources
- DeepSeek · Search and Fetch API guide （Checked 2026-11-06）
- DeepSeek · DeepSeek App announcement (2025-01-15) （Checked 2026-11-06）
- kagentx evaluations: the data on the case study pages; each example gives its date.