In short. Use search_domain_filter to select permitted search sources, then check returned URLs and answer claims against the allowed pages.
Definition: As of 17 October 2026, Perplexity's
search_domain_filteris an API setting that limits web search results to allowed domains or excludes specified domains. An allowlist selects the permitted sources for a request; an application requiring answers supported exclusively by those sources should also validate the answer's claims.
Key facts
- Perplexity's Search Domain Filter guide documents a maximum of 20 entries and separate allowlist and denylist modes.
- Perplexity's migration overview states that Sonar Chat Completions support ended on 27 September 2026; synchronous and streaming requests continue through gradual reformulation as Agent API requests.
- The migration instructions move Sonar's top-level search filters into the Agent API's
web_searchtool configuration. - The Web Search reference identifies each returned source through its
idandurlfields.
How is the domain restriction configured?
Perplexity's migration instructions show search_domain_filter as a top-level parameter in a Sonar request. This JSON fragment illustrates an allowlist containing a root domain used in Perplexity's examples:
{
"search_domain_filter": ["nature.com"]
}
For an existing website integration, the domain value is replaced with the website's domain. The setting belongs in the API request sent by the application. The Search Domain Filter guide also shows this parameter in Search API requests.
For the Agent API, the same parameter belongs inside filters on the web_search tool. The following request-body fragment matches the structure in the migration instructions:
{
"preset": "fast",
"input": "Latest renewable energy policy updates",
"tools": [{
"type": "web_search",
"filters": {
"search_domain_filter": ["iea.org", "energy.gov"]
}
}]
}
Perplexity's migration overview recommends moving existing Sonar usage to the Agent API.
What are the filter's limits?
Perplexity's domain-filter guide defines an allowlist as including only the listed domains. A denylist excludes entries prefixed with -. Both modes cannot be combined in the same request.
The guide specifies domains without a protocol and recommends avoiding the www prefix. A root domain includes its subdomains. For a narrower section, the guide supports a domain followed by a path: nature.com/articles matches that section and its subpaths, including on subdomains.
Restricting sources can reduce the result count or return zero results. Combining domain filtering with date or language filters can narrow the available results further. The application therefore needs to handle an empty search response.
How can the returned answer be checked?
Perplexity's Web Search reference describes search_results entries with a canonical url, a title, and a snippet containing text extracted during search. It treats the id and url fields as the source of truth for citations.
For an application with a strict source policy, validation should compare returned URLs with the permitted domain scope. A separate content check should compare factual claims with the allowed pages. Perplexity's migration guide supplies an instruction that asks the model to verify each claim against the cited results and drop unsupported claims. Domain selection and claim verification are distinct checks.
Related terms
FAQ
Can I include several domains?
Yes. The domain-filter guide permits up to 20 entries per request. You must choose either allowlist or denylist mode.
Does listing my domain ensure that every answer claim comes from it?
You should check the claims against the returned sources. Perplexity's migration guide provides an explicit claim-verification instruction for this purpose.
Where does the filter go after migration?
The Agent API's Web Search reference places search_domain_filter inside tools[].filters for a tool whose type is web_search.
Sources
- Perplexity, "Search Domain Filter", undated
- Perplexity, "Migrate from Sonar to the Agent API", undated
- Perplexity, "How to migrate from Sonar", undated
- Perplexity, "Web Search", undated
Check how a page reads to answer engines. Open the GEO score analyzer