Key points
- DeepSeek released a Search API and a Fetch API on Oct. 17, and each call costs $0.75 per 1,000 successful requests.
- Results carry an authority tier and a rerank score, and site: queries let publishers check how their own pages are rated.
- DeepSeek says the engine screens out SEO spam, GEO manipulation and low-quality AI content, but has not defined GEO manipulation.
- Server-side searches for Claude Code, Codex and DeepSeek Harness users are free until billing starts Oct. 31 at 16:00 UTC.
Why it matters for marketers: Marketers can now see how a Chinese AI engine scores their pages, but the scoring rules, crawler identity and GEO definition are undisclosed, so optimization decisions rest on guesswork.
DeepSeek made its own search engine available to developers through a new Search API, launched Oct. 17 alongside a Fetch API that returns single pages as Markdown. The company's changelog says the engine handles search on its website and API, and that it filters SEO spam, GEO manipulation and low-quality AI-generated content.
For marketers, the notable part is the scoring. Every result includes an authority tier and a rerank score that blends relevance, authority and freshness. Because the API supports site: queries, a team can pull up to 50 pages from its own domain and see how each is rated for a given query. Searches cost under a cent each with a funded account. If a page is missing, the documentation gives no way to tell whether it is outside the index or was filtered.
Several gaps remain. DeepSeek has not defined GEO manipulation or explained how tiers and scores are assigned. The launch pages do not name the crawler behind the index or the user agent Fetch uses. The Known Agents directory lists a DeepSeekBot as a training crawler, so it is unclear whether blocking it in robots.txt would affect search inclusion.
Privacy is another open question. DeepSeek's privacy policy says it uses third-party APIs for search, and a Chinese search API firm, Bocha, calls itself the official web search supplier in a GitHub README. Neither the changelog nor the policy says whether queries still go to a third party.
Watch DeepSeek's documentation for a GEO definition, tier methodology and a named index crawler. Until then, use site: queries to audit your pages, and treat the results as a snapshot rather than a rulebook.
Sources
This brief is our own summary. Read the full reporting at the source.
