Does DeepSeek respect robots.txt?

DeepSeek has not published documentation on this. Without a named agent, there is no reliable way to confirm or target it.

Should we block DeepSeek?

You cannot do it precisely without a documented agent. Decide based on the known crawlers of other providers, and revisit if DeepSeek publishes guidance.

Is it worth the effort?

Direct effort is limited, because there is little to act on. The indexability and clarity work done for other engines already covers most of it.

Engine guide: DeepSeek

How to get cited by DeepSeek

DeepSeek is the hardest assistant on this list to optimise for, because its maker documents almost nothing about retrieval. The honest approach is to work on what is known and test the rest.

The short answer

DeepSeek's chat app can search the web when a user turns search on, and it lists the pages it read. DeepSeek has not published a crawler user agent, a search partner or publisher guidance. So the practical route is indirect: be indexed well across major search engines, write pages with plain, self-contained answers, and test DeepSeek directly with search switched on.

What ranked on 26 September 2026: Results for this query are thin and mostly speculative. No primary DeepSeek documentation on web retrieval ranks, because none has been published in the form OpenAI, Google and Anthropic provide.

What is specific to DeepSeek

The model is widely used, including through third-party apps and self-hosted deployments, and many of those have no web access at all.

Search is a user toggle, so most answers come from training data unless the person turns it on
When search runs, the app shows a list of pages consulted, often many more than it quotes
The open-weight models are deployed by other companies, whose products add their own retrieval, or none
The model is strong in Chinese and English, and Chinese-language sources can appear for queries about Asian markets
Training data has a cutoff, so ungrounded answers about your brand reflect how you were described before it

Things that sound right and are not

ClaimThere must be a DeepSeek bot to allow.
None is publicly documented. Guides naming one are guessing. Do not add rules for an agent name nobody has verified.
ClaimDeepSeek answers the same everywhere.
A third-party app running DeepSeek weights may use a different search tool or none. Test the official app and the products your buyers actually use separately.
ClaimIts answer about us is out of date, so we should ignore it.
Stale answers come from training data. You cannot patch that directly, but clear, consistent current descriptions across the web are what the next training run will see.

Where to start

01Run your core prompts in the official DeepSeek app with search off, then on, and compare the two answers.
02Note which domains appear in the list of pages read, since those are the sources its search reaches.
03Make sure your site is indexed in Google and Bing and loads fast from Asia, because retrieval timeouts drop slow pages.
04Keep a short factual about page that states what you do, since that is what training and retrieval both pick up.
05If Asian markets matter, check whether a Chinese-language description of your business exists and is accurate.

Questions

Does DeepSeek respect robots.txt?: DeepSeek has not published documentation on this. Without a named agent, there is no reliable way to confirm or target it.
Should we block DeepSeek?: You cannot do it precisely without a documented agent. Decide based on the known crawlers of other providers, and revisit if DeepSeek publishes guidance.
Is it worth the effort?: Direct effort is limited, because there is little to act on. The indexability and clarity work done for other engines already covers most of it.

Doing this yourself and want a second opinion on what the assistants say about you? Tell me what you are seeing, or read the AEO Playbook for the whole framework.