Short answer
DeepSeek has opened its in-house search engine to developers. Each result carries an authority tier for the page and a relevance score. The company says the engine filters SEO spam, GEO manipulation and low-quality AI content, but it has not explained how the filters work or which crawler feeds the index.
Highlights
- DeepSeek's Search API returns an authority tier from 0 to 4 and a relevance score with every result; how these values are calculated has not been explained.
- According to the company's changelog, the engine filters SEO spam, "GEO manipulation" and low-quality AI content; GEO manipulation is not defined.
- A site: query can list the pages of a domain, but the API does not show why a page is missing from the index.
2 min readAuthor: UNIT Journal EditorEditor-in-chief: Uğur Deniz İlhan
According to Search Engine Journal, DeepSeek opened its in-house search engine to developers on Oct. 17. The company's changelog says the engine filters SEO spam, "GEO manipulation" and low-quality AI content. GEO means preparing content so that AI answer engines cite it.
What does the DeepSeek Search API return?
For every result, an authority field for the page and a relevance score. The authority field runs from 0 to 4, with -1 meaning unknown; DeepSeek's API reference describes tier 3 and above as highly authoritative. The relevance score runs from 0 to 3, combines relevance, authority and freshness signals, and 1.5 or higher counts as high quality. The documentation does not explain how a page receives its tier or score.
Queries can be limited to one domain with site: and filtered by language and publication date. A single site: search can return up to 50 pages. Search and fetch requests each cost $0.75 per 1,000 successful requests.
What do we not know?
The report lists three open points. First, "GEO manipulation" is not defined, so it is unclear which practices the filter catches. Second, the launch pages do not name a crawler or user agent for the index. Third, a robots.txt rule blocking DeepSeekBot may target the crawler a bot directory describes as collecting training data; the launch pages do not make clear whether such a rule affects the search index.
What should sites in Turkey do?
DeepSeek is not a main search channel in Turkey, but apps and agents built on its API can use this index. A few cheap, measurable steps are enough for now:
- Open an account and run site: queries for your own domain; note which pages are in the index and their tiers.
- Do not read a missing page as "blocked": as the report notes, the API does not show the reason.
- Before writing a robots.txt rule, be clear about which crawler you are targeting; training and search may not fall under the same rule.
- Manage the risk through genuinely useful, well-sourced content rather than fear of a "GEO manipulation" rule nobody has defined.
Building measurable visibility in AI answers through GEO (generative engine optimisation) means measuring across several engines, not just one. DeepSeek is a new addition to that list; the findings rest on the first launch pages and may change as the company adds detail.
Frequently asked
- How will DeepSeek's search engine affect my site?
- There is no measurable effect yet. The API may assign your pages an authority tier and relevance score, but how they are calculated has not been explained. A sensible first step is to check with a site: query whether you are in the index.
Sources
- Search Engine Journal ·
Follow UNIT Journal
What's new in search, AI and technology, in your feed every day.
