diff --git a/README.md b/README.md
index ff259f6..d4bbc30 100644
--- a/README.md
+++ b/README.md
@@ -229,7 +229,9 @@ The legacy `402` with `code: "AUTH010"` means the domain isn't enabled. `extract
 
 #### Preparing a domain
 
-`client.prepare_domain(domain, example_urls)` submits a domain for Extract (`POST /v1/extract/prepared-domains`), and `client.prepared_domain_status(domain)` reads where it is (`GET /v1/extract/prepared-domains/{domain}`). Both return the parsed status object. Pass the domain without a scheme and 1 to 5 example URLs, one per kind of page you want to extract. These calls cost no credits.
+`client.prepare_domain(domain, example_urls)` submits a domain for Extract (`POST /v1/extract/prepared-domains`), and `client.prepared_domain_status(domain)` reads where it is (`GET /v1/extract/prepared-domains/{domain}`). Both return the parsed status object. Pass the domain without a scheme and 1 to 5 example URLs on it. These calls cost no credits.
+
+One example URL works, but several related URLs work better. We recommend at least 3 pages of each page type you need extracted, up to 5 in total; several page types can share the 5, for example 3 product pages and 2 search-result pages. Extract compares the pages of one type against each other, so it can check the fields it found and fill in fields that one page is missing. More pages of a type give more complete, more reliable results for that type.
 
 Preparation takes minutes, and the client does not wait for it. Poll the status until it is no longer `"preparing"`:
 
@@ -243,7 +245,12 @@ client = ZenRowsClient("YOUR-API-KEY")
 try:
     status = client.prepare_domain(
         "www.scrapingcourse.com",
-        ["https://www.scrapingcourse.com/ecommerce/"],
+        [
+            # Three pages of one page type: product pages on the same domain.
+            "https://www.scrapingcourse.com/ecommerce/product/abominable-hoodie/",
+            "https://www.scrapingcourse.com/ecommerce/product/adrienne-trek-jacket/",
+            "https://www.scrapingcourse.com/ecommerce/product/aeon-capri/",
+        ],
     )
 except PreparedDomainError as err:
     # e.g. 409 in_progress, 429 out_of_allowance, 503 overloaded
diff --git a/src/zenrows/client.py b/src/zenrows/client.py
index bae8a5e..62f83c3 100644
--- a/src/zenrows/client.py
+++ b/src/zenrows/client.py
@@ -325,7 +325,13 @@ def prepare_domain(self, domain: str, example_urls: list[str]) -> dict:
         `ready` or `failed`), `state`, `submitted_at`, `updated_at`.
 
         `domain` is the domain without a scheme. `example_urls` are 1 to 5
-        URLs on that domain, one per kind of page you want to extract.
+        URLs on that domain. One URL works, but several related URLs work
+        better: pass at least 3 pages of the same page type when you can,
+        up to 5 (several page types can share the 5, for example 3 product
+        pages and 2 search-result pages). Extract compares the pages of one
+        type against each other, so it can check the fields it found and
+        fill in fields that one page is missing; more pages of a type give
+        more complete, more reliable results for that type.
         Submitting a domain that is already prepared returns its current
         status instead of starting a new run.