SEOAST — home
Get an audit

Insights ·

DeepSeek opened its search to AI agents. Can it read and date your pages?

The short version

On October 17, 2026, DeepSeek opened Search and Fetch APIs built on its in-house search engine to all API users, giving AI agents a new way to find and read web pages. Search results carry a publication date (or null when unknown), an authority tier and a relevance score, and date operators restrict results to pages published inside a window. In an early independent test, every undated result disappeared when a date filter was applied, and the fetcher read raw HTML without running JavaScript. The practical takeaway: make sure each important page states its publication date in machine-readable form and carries its content in the server HTML.

What DeepSeek launched on October 17

DeepSeek announced on October 17, 2026 that its Search and Fetch APIs are open to all API users. Both run on DeepSeek’s in-house search engine. Search takes a query of up to 128 characters and returns up to 50 results. Fetch takes a URL and returns the page’s title and its content as Markdown. Each costs $0.75 per 1,000 successful requests, and failed, empty or timed-out requests are not billed. Calls made through the server-side tools in DeepSeek Harness, Claude Code and Codex are free until billing starts on October 31, 2026 at 16:00 UTC.

Each search result carries fields that matter to site owners. The date field is the publication time in UTC, or null when DeepSeek does not know it. The authority field is a tier from 0 to 4, with -1 for unknown. The rerank score runs from 0 to 3 and, in DeepSeek’s words, combines signals such as relevance, authority and freshness. Queries accept site: and -site: operators, plus after: and before: operators that restrict results to pages published after or before a date.

The documentation says fetching is subject to the target site’s availability and access policies. It does not name the fetcher’s user-agent token, publish its IP ranges, or say whether Fetch reads robots.txt.

SEOAST Impact Rating: A2, plan for it. DeepSeek itself confirmed a new search engine that AI agents can call, and its result format rewards pages with a known date, but nothing changes for your site today and there is no penalty. The Impact Rating is SEOAST’s own editorial rating, not a DeepSeek or Google severity level.

What an early independent test found

Suganthan Mohanadasan, co-founder of Snippet Digital, tested both APIs on launch day and published his findings in Search Engine Journal on October 20. This is one practitioner’s test: 110 calls over a single day from one calling region. He labels the results preliminary.

Three findings matter most for site owners. First, date filters removed every undated result, and about 11% of his unfiltered results had no date. Second, the authority tier behaved like a host-level score: in 16 site: probes, every URL on a given host shared one value. Third, the fetcher did not run JavaScript. Its requests used a Firefox user agent with a Chinese-language header from Huawei Cloud addresses in Hong Kong, and it returned content from three URLs that robots.txt disallowed. He could not confirm whether robots.txt was checked at all.

Why dates, server HTML and robots.txt are the pressure points

A date filter is a hard gate. When an agent asks DeepSeek for pages published after a given date, a page with no known date cannot match, however good it is. Pages that show a date only in an image, in a relative phrase such as “3 days ago”, or not at all give a search engine nothing reliable to read.

A fetcher that does not run JavaScript sees only the first HTML response. If your article body, prices or product details are injected by client-side scripts, an agent using Fetch may receive a shell.

robots.txt is the weakest of the three controls here. On October 21, Google’s Gary Illyes said that to block a specific AI crawler you have to target its user-agent token explicitly. DeepSeek has not published a token for its fetcher, so there is currently no DeepSeek-specific group you can write.

What to check this week

These are page-level checks you can do without any new tool.

  • Give every article and dated page a visible publication date near the top, written as a full date.
  • Add datePublished and dateModified to the page’s Article or BlogPosting JSON-LD, in ISO 8601 format, and keep them consistent with the visible date.
  • View the raw HTML of your key pages, not the rendered DOM, and confirm the main text is there.
  • If content only appears after JavaScript runs, move it to server rendering or static generation.
  • Review your robots.txt with the knowledge that an undisclosed fetcher cannot be addressed by name; use authentication or server-side rules for anything that must stay private.
  • Watch your server logs for Firefox user agents from Huawei Cloud ranges if you want to see whether Fetch traffic is reaching you.

What SEOAST checks, and what it does not

SEOAST does not query DeepSeek, does not query ChatGPT, Perplexity, Gemini or Google AI Overviews, does not track citations and has no AI visibility score. It cannot tell you whether DeepSeek has indexed or dated your page.

SEOAST audits one URL from its server HTML, which is close to what a fetcher without JavaScript receives. It flags pages whose substance appears only after JavaScript runs. Its structured data check reports whether JSON-LD parses and which @types are present, such as Article; it does not validate individual properties like datePublished, so confirm the date fields yourself. It also reads your robots.txt, tests the audited path for Googlebot and reports which AI crawlers on its list are disallowed at the site root. DeepSeek’s fetcher is not on that list because it has no published token.

Questions

What did DeepSeek launch on October 17, 2026?

Search and Fetch APIs on its in-house search engine, open to all DeepSeek API users at $0.75 per 1,000 successful requests each. Search returns results with a date, an authority tier and a relevance score; Fetch returns a page’s title and content as Markdown.

Do undated pages show up in DeepSeek search results?

They can appear in unfiltered results with a null date. In one practitioner’s test, applying an after: or before: date filter removed every undated result.

Can I block DeepSeek’s fetcher in robots.txt?

Not by name today. DeepSeek has not published a user-agent token for Fetch, and its documentation does not say whether Fetch reads robots.txt. One early test saw it return content from URLs that robots.txt disallowed.

Does SEOAST check my publication date?

Not directly. SEOAST reports whether your JSON-LD parses and which @types are present, but it does not validate datePublished or other individual properties.

Sources

Check your own pages

Audit your page