Find real-time information from trusted internet sources.
Last updated: Oct 21, 2026
Overview
Web Search allows Harvey to reference real-time information from the internet — including current events, company websites, and newly published materials — to provide more accurate and up-to-date answers. This is especially useful when your work depends on recent legal developments, press releases, or regulatory updates that may not included in Harvey’s internal knowledge base.
For Admins: Enabling Web Search
Web Search is an optional feature which must be enabled for users by a workspace administrator.
To enable Web Search:
Go to Command Center → Configurations → Knowledge.
Toggle on Web search for your workspace.
Once Web Search is enabled, Harvey applies a default web search provider and processing location based on your workspace region:
For US workspaces, the default web search provider is Parallel, and the default processing location is the US.
For EU workspaces, the default web search provider is Parallel, and the default processing location is the EU.
For AU workspaces, the default web search provider is Parallel, and the default processing location is in the US.
Admins may change the web search provider or processing location at any time by navigating to Command Center → Users & Permissions → Roles, then choosing a new provider and location from the Web search permission category.
Once enabled, admins may enable Web Search for all users or specific users within the workspace.
For Users: How to Use Web Search
Important: Web Search is intended for retrieving information from public websites only.
Follow these steps to run a Web Search query:
Select Sources when writing a query.
Toggle on Web Search.
[Optional] Shape web search by specifying sites Harvey should use:
Include only specific URLs (for example, harvard.edu,)
Exclude sites you don’t want (for example, reddit.com)
Write your query.
Select the arrow to run your query.
Harvey reformulates your text into a secure search query and sends it our Web Search providers. From there, they return relevant results, which Harvey uses to produce a response, often including citations or summaries from the retrieved web pages.
Web Search is also available in the Harvey for Word and Harvey for Outlook add-ins, where you can select it from the sources menu in the prompt text box.
Tip: If you often search the same set of websites, save them as a custom Knowledge Source filter so you don't have to re-enter them for each query. Refer to Regional Knowledge Sources Overview for more details.
Data Handling and Security
Data Flow
When users submit a Harvey query with Web Search enabled:
Harvey reviews the user’s request and creates a generalized keyword search query for secure transmission to the enabled Web Search provider. Keyword search queries shared with Web Search providers are designed using system prompts to exclude customer documents stored in Vault, verbatim prompt text, or sensitive or identifying information such content could contain.
All data shared with Web Search providers is encrypted, processed temporarily, and deleted immediately after results are returned.
Web Search Providers
Harvey requires its Web Search providers to comply with strict contractual and technical safeguards:
No data storage, training, or human review: Neither Parallel nor You.com store, train on, or allow human access to Harvey customer data.
Temporary processing only: All Web Search queries are processed briefly and query content is deleted immediately after results are returned.
Secure environment: All processing occurs within our web processor's secure cloud infrastructures. To learn more about the security programs of Harvey’s web search providers, see:
Regional data processing: Web Search queries are processed in the United States or the EU, according to the processing location selected by your workspace admin. Both Parallel and You.com can support either region.
End-to-end encryption (TLS 1.2+ in transit, AES-256+ at rest)
Parallel offers a more robust and reliable search index for Harvey's specific use cases, with greater flexibility for Harvey to rank and prioritize sources based on a user’s research objective.
Parallel also offers improved support for returning exact source text snippets referenced in responses, enabling clearer and more transparent citations.
When Web Search is enabled for a query, Harvey reviews the user’s request and creates a generalized keyword search query for sharing with its Web Search providers. This derived query is scoped solely to identifying and retrieving relevant sources from the public internet. This derived query is designed to exclude customer documents stored in Vault, verbatim prompt text, or any sensitive or identifying information such content could contain. Once an individual web search is complete, Harvey’s Web Search providers immediately delete the query.
Yes, both follow the same processing guidelines as detailed in our FAQ above, Are document or file uploads sent to Web Search providers?
Web Search operates under the same zero-retention principles as Harvey’s LLM providers:
No data is stored, trained on, or reviewed by humans.
Queries and search results are deleted immediately after use.
Workspace admins concerned with sharing sensitive information with Harvey’s Web Search providers can:
Limit use to approved users and their use cases
Configure a pop-up warning to remind users not to include sensitive organizational or personal information (submit a feature request to enable and customize this message as desired)
Harvey will continue functioning normally, but it will not access or reference real-time web information.
Customers who have previously enabled Web Search will continue to be able to power their web search experiences via You.com if they do not wish to enable Parallel’s web search capabilities. However, Web Search experiences may differ given certain expanded performance capabilities provided by Parallel.
We will default to Parallel in this case since Parallel offers a more robust end to end search experience for Harvey's use case.
No. Regional processing options apply to Web Search only. Parallel-powered Knowledge Sources are managed through a separate permission and are processed in the United States; there is currently no EU endpoint for them.
Legal and Compliance
Yes. Both are listed as subprocessors under Harvey’s Data Processing Addendum. Harvey requires them to comply with applicable data protection laws and to provide the same level of data protection and information security provided for under Harvey’s DPA with its customer.
Yes.
You.com is certified under the EU–U.S. Data Privacy Framework (DPF). SCCs are maintained as a fallback mechanism, but only one data transfer safeguard (DPF or SCC) applies at any given time.
Harvey has entered into standard contractual clauses with Parallel to facilitate the transfer of personal data from the EEA, UK, and Switzerland to the US.
Other
Once enabled, the Web Search reminder pop-up appears at the start of every new session in Harvey.
You’ll see the reminder:
When you open Harvey in a new browser tab (each tab starts a new session).
When you close your browser or tab and then reopen Harvey.
You won’t see the reminder:
When you refresh the page — your current session continues.
When you navigate between pages within Harvey — the session remains active.
Specification is useful when you already know the types of sources you want to rely on: such as regulatory sites, financial data providers, or industry specific publications, and want to reduce noise in the results.
No. Web Search specification is optional and available when you want more control, without adding steps to the standard workflow.
No. Controls are integrated into the existing experience, so you can apply them quickly or skip them entirely.
When you restrict a Web Search query to specific URLs, Harvey will generally limit results to those domains. However, citations from outside your specified URLs may still appear in two situations:
The included URLs don't contain information relevant enough to answer your query.
The included URLs are not currently present in Parallel's search index.
Not always. It can only reach pages in the provider's search index. Sites or pages that aren't indexed (including some .aspx pages) won't be scanned or appear as citations.
Sources when writing a query.