DeepSeek Opens Search Engine To Developers

DeepSeek, a search engine company, has opened its in-house search engine to developers, according to Search Engine Journal. This move allows developers to access the engine's capabilities, including filtering out SEO spam and low-quality AI-generated content. The company's changelog states that the engine handles search on its website and in its API, indexing hundreds of billions of web pages and taking in hundreds of millions of search requests daily.

The Search API returns results with an authority tier and a relevance score, which may help developers evaluate the quality of the content. Each result also includes a snippet and a summary, as well as the option to request page content. The API accepts site: queries, which limit results to one domain, and allows filtering by language and date published.

DeepSeek's engine is designed to filter out GEO manipulation, a practice where content is shaped to be cited by AI answer engines. However, the company has not defined what GEO manipulation is or described how its detection process operates. This lack of clarity may raise questions about how the engine evaluates and filters content.

The company's privacy policy, last revised in February, states that it integrates third-party APIs to provide search services and shares users' input keywords with these companies. A Chinese-based search API company called Bocha is listed as the official web search supplier of DeepSeek in a GitHub README, but DeepSeek does not reference Bocha in its launch pages.

The Search API and Fetch API, which provide a single Markdown page, are subject to the target site's availability and access policies. The EU summary of the data used to train DeepSeek-V4 states that the crawler is designed to respect robots.txt instructions and other standard web protocols, but it does not identify which crawler this refers to.

The Known Agents bot directory lists a DeepSeekBot user agent, described as an AI training data crawler. However, it is unclear if this crawler is used for the search index, and the launch pages do not provide information on what user agent Fetch uses. This lack of transparency may raise questions about how DeepSeek's engine operates and how it evaluates content.

With a funded DeepSeek API account, individual site: searches are relatively inexpensive, costing under $0.01. Each site: search can return up to 50 pages from a website, along with an authority tier and a relevance score for that query. However, if a result is missing, it may be due to the page being outside the index, caught by filters, or excluded for some other reason.