Google has been indexing DeepSeek conversations that users shared via public links, making thousands of private chats discoverable through a simple search query.

DeepSeek, the Chinese AI lab whose open-source models have been widely adopted since the release of DeepSeek-R1, has grown rapidly into one of the most-used chatbot platforms globally. The scale of that adoption means the pool of users potentially affected by a sharing-related exposure is substantial.

David Konitzny, an information security specialist at Peec AI, discovered that Google was surfacing pages from DeepSeek's chat.deepseek.com/share/ directory. Anyone could locate these conversations by entering site:chat.deepseek.com/share/ into the Google search bar.

While Google initially limits the number of visible results and indicates that similar pages have been omitted, expanding those results reveals a substantially larger set of indexed conversations.

Scope of the Exposure

The investigation identified approximately 2,000 publicly accessible pages. The conversations were created in 2026, with the highest concentration occurring in April.

Approximately 31% of the analyzed chats were written in Russian, while around 20% were in English. The remainder appeared in multiple other languages, indicating that the privacy issue affected users across various regions.

Why Shared DeepSeek Chats Are Appearing in Search Results

DeepSeek's Share feature enables users to select portions of a conversation and generate a public link that can be distributed to colleagues, friends, or other contacts. The platform does warn users that "anyone with the link" can view the conversation. However, the warning does not explicitly state that search engines may also discover and index the page.

Under standard web practices, search engine crawlers treat any URL that does not actively block them as fair game for indexing. Platforms can prevent this by including a robots.txt directive disallowing crawler access to specific paths, or by embedding a noindex meta tag in the page header. If neither measure is applied to shared-conversation URLs, Googlebot and other crawlers will eventually discover and index those pages.

This distinction is significant. A user may understand that a link recipient can open the conversation but still reasonably assume the chat would remain difficult for others to find. Once a page appears in Google Search, the potential audience is no longer limited to those who received the original URL — anyone using appropriate keywords or search operators may be able to locate it.

Some indexed conversations reportedly contained fragments of commercial source code, workplace assignments, personal correspondence, and discussions involving financial matters. The presence of professional code snippets raises particular concern for enterprise users who may be inputting proprietary or confidential business logic into consumer-grade AI tools without their organization's data governance policies accounting for share-link exposure.

Although users actively chose to create public links, many likely did not anticipate their conversations becoming searchable.

Lasting Digital Footprint

Removing a shared link does not guarantee it will disappear from search results immediately. Search engines periodically revisit indexed pages, meaning a deleted or inaccessible link may remain visible until Google updates its records. Previously stored snippets and cached information may also continue to appear for some time.

This creates additional risk for users who only become aware of the exposure after their conversations have already been indexed.

How to Delete Shared DeepSeek Conversations

Users can review and remove their public DeepSeek links through the platform's settings:

1. Open the Settings section.
2. Navigate to the Data tab.
3. Select Shared Links.
4. Click Manage.
5. Delete any links that should no longer be public.

Users are advised to review the content of all previously shared conversations, prioritizing the removal of links containing sensitive professional, personal, or financial information.

Broader Industry Context

Similar indexing incidents have previously affected ChatGPT and other AI services that allow users to publish conversations through shareable links. Search engine crawlers generally treat any accessible public URL like any other webpage. Unless a platform applies appropriate access controls or uses directives designed to prevent indexing, shared pages may eventually appear in search results.

Konitzny published the findings on July 20, 2026, and Cybernews reviewed them on July 22. According to the reports, DeepSeek had not issued a public response at the time of the investigation.

The incident underscores a broader privacy challenge facing AI platforms. Users frequently provide chatbots with detailed information about professional projects, personal situations, internal company processes, and financial decisions. While many users may understand that shared links are technically public, they often expect them to remain accessible only to those who receive the URL directly.

The DeepSeek case demonstrates that a public AI conversation may not remain hidden simply because the link is difficult to guess. Platforms could mitigate this risk by preventing shared chats from being indexed by default and requiring users to actively opt in to search engine visibility. Until stronger protections become standard, users should treat every shared AI conversation as information that could eventually become publicly searchable.

Source: Coinpaper