Skip to main content

External Websites

External Websites lets the Virtual Agent answer questions from trusted public websites, such as vendor documentation or industry references. You add the website domains the agent may use. When a question needs web information, the agent runs a live web search restricted to those domains and answers with citations to the pages it used.

Live search, not indexing

External Websites doesn't crawl, copy, or store website content in the Knowledge Base. Each question is searched live, so answers always reflect the current, publicly indexed version of the site. There's nothing to sync or refresh.

To make internal or login-protected content available to the Virtual Agent, use a Knowledge Integration or upload documents instead.

How it works​

  1. A user asks the Virtual Agent a question.
  2. The agent collects the External Websites available to that user's audience.
  3. The Virtual Agent searches the web only when at least one website is available to the user. It searches in either case:
    • The question is best answered from the web.
    • The Knowledge Base has no answer.
  4. The search is limited to the configured domains. The agent reads the most relevant pages and writes an answer that cites them as sources.

If no relevant pages are found, the Virtual Agent falls back to its normal no-answer handling.

Adding a website​

  1. In the Rezolve.ai console, go to Knowledge Management → Manage Knowledge → External Websites.

  2. Click Create External Websites.

  3. Fill in the fields:

    FieldDescriptionExample
    Website Name (required)A unique, friendly nameMicrosoft Learn
    Web Domain (required)The domain to search. Must be unique, without https:// or a leading www..learn.microsoft.com
    Blacklist AudienceWhen on, the selected audiences are excluded instead of includedOff
    AudienceThe audiences that can use this websiteIT Staff

  4. Save the record.

The website is used as soon as it's saved. The record's workflow status (TO DO, IN PROGRESS, DONE) doesn't affect whether it's used. To stop using a website, delete its record.

Domain matching​

Web DomainWhat is searched
example.comexample.com and all of its subdomains, such as docs.example.com
docs.example.comOnly docs.example.com and its subdomains

A leading www. and the http:// or https:// prefix are ignored when results are matched.

Controlling who can use a website​

Blacklist AudienceAudienceWho gets answers from the website
Off(empty)Everyone
OffOne or more audiencesOnly users in those audiences
OnOne or more audiencesEveryone except users in those audiences

Users who don't belong to any audience can only use websites with no audience set.

Search settings​

In the Generative AI configuration, two settings affect web answers:

SettingLocationEffect
Web Search SpeedBot Responses tabFast runs a single search for quicker answers. Standard runs several searches and ranks page content more thoroughly for better answers.
Web Search PromptAdvanced tabCustomizes how web-based answers are written, for example to add a disclaimer that the answer comes from an external site

Limits and considerations​

  • Public pages only: only pages that public web search can find are used. Intranet sites, sites that require a login, and pages blocked from search engines can't be used.
  • No file ingestion: PDFs and other files on the site are only used if the web search returns them. Nothing is downloaded into the Knowledge Base.
  • Up to 100 websites are considered for each user.
  • Unique entries: each Website Name and Web Domain can be added only once.
  • Answer quality depends on the site: choose authoritative, well-structured sites, and prefer a specific subdomain (for example support.vendor.com) over a broad one.

Troubleshooting​

SymptomLikely causeResolution
The Virtual Agent never uses a websiteThe user isn't in the website's audience, or Blacklist Audience excludes themCheck the Audience and Blacklist Audience settings
Web Domain is rejected when savingThe domain starts with www., contains spaces, or already existsEnter the bare domain, for example example.com
Answers don't include expected pagesThe pages aren't publicly indexed, or are behind a loginUse a Knowledge Integration or upload the content instead
Answers come from a Knowledge Base article instead of the websiteThe Knowledge Base had a relevant answerThis is expected. Web search is mainly used when the Knowledge Base has no answer.

Next Steps​