Pagination and depth
Page through Google results with page, or fetch up to the top 100 in one call with num. num=100 is a depth bundle that costs 7 credits instead of 10.
Google serves results ten at a time. SerpKite gives you two ways to go deeper: request pages one by one with page, or ask for up to 100 results in a single call with num.
page
page selects the results page, from 1 (default) to 10. Each page is a separate Google fetch and a separate credit.
curl https://api.serpkite.com/v1/search \
-H "Authorization: Bearer $SERPKITE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"q":"espresso grinder","country":"us","page":2}'page works on every paged endpoint (search, news, images, videos, maps, places, shopping, scholar, patents).
Positions continue across pages when you request them in order: if page: 1 holds 9 results, page: 2 starts at 10. This holds for 6 hours and needs the earlier pages of the same request from the same engine. A page requested on its own, or in parallel with the earlier ones, numbers from its slot instead (page: 2 starts at 11), so positions can then skip a number. Providers also repeat a result on neighbouring pages at times, so remove duplicates by link when you merge pages.
Images
/v1/images pages differently: Google’s Images tab answers each page with a block of up to 100 different images, and that whole block is one page and one credit.
page: 1returns positions 1–100,page: 2positions 101–200, and so on. Most queries run out after four to six pages.num(1–100) returns only the firstnumimages of the page, for the same credit. It doesn’t change which images a page holds, sonum: 10, page: 2returns positions 101–110. Leavenumout to get the whole page.- A page past the last one returns empty
resultsquickly and is not charged. Stop paging there.
curl https://api.serpkite.com/v1/images \
-H "Authorization: Bearer $SERPKITE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"q":"red sneakers","country":"us","page":2,"num":20}'num and the depth bundle
On /v1/search and /v1/news, num above 10 fetches several pages in one request and merges them into one results list. SerpKite fetches ceil(num / 10) pages and bills one credit per page that returned results, capped at 7:
num |
Pages fetched | Credits (at most) |
|---|---|---|
| 1–10 | 1 | 1 |
| 20 | 2 | 2 |
| 30 | 3 | 3 |
| 50 | 5 | 5 |
| 70 | 7 | 7 |
| 100 | 10 | 7 |
So the top 100 costs 7 credits instead of the 10 you would pay with ten page calls, and it comes back in one response with consistent positions. Pages that come back empty are free: a long-tail query with results on only two pages costs 2 credits, whatever num you asked for. A cache hit with max_age costs half of what the cached result cost.
curl https://api.serpkite.com/v1/search \
-H "Authorization: Bearer $SERPKITE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"q":"espresso grinder","country":"us","num":100}'Rules:
nummust be between 1 and 100. Values up to 10 behave like 10.numabove 10 requirespage: 1(or nopage). Combining them returns400 invalid_request.- On other endpoints
numabove 10 is treated as 10; usepageto go deeper. /v1/images is the exception: see Images. /v1/reviews returns at most 10 reviews per call (the first 10 of a place), billed 1 credit per 10 reviews actually returned. - Google sometimes returns fewer than
numresults for a query. You are billed for the pages actually fetched, and a page that comes back empty is refunded. - Pages are fetched in order: each starts once the page before it has answered, or 1.5 seconds after that page started, with at most two pages in flight (Google refuses the same query arriving twice at once). Two pages in a row without results, empty or failed, end the walk. Each page adds about a second, so
num: 100takes roughly 10 seconds. If you page withpageinstead, request one query’s pages one after another, not in parallel.
Batch it for rank tracking
The depth bundle combines with the batch lane: num: 100 in a POST /v1/batches request costs 3.5 credits for the top 100. From 23 November 2026, batch depth requests cost 0.5 per page that returned results, so the top 100 is 5 credits (see the changelog). See Batch requests and Rank tracking.
Custom Search pagination
The CSE-compatible /customsearch/v1 keeps Google’s CSE paging: start is the index of the first result (any value up to result 100) and num is 1 to 10 per call. Indexes follow Google’s pages of 10: a call within one page is one credit, and one that spans two pages (start=5) reads both for two. With searchType=image, start counts images in the first block of 100, so start=11 returns images 11–20.
curl "https://api.serpkite.com/customsearch/v1?q=asyncio&start=11&num=10" \
-H "Authorization: Bearer $SERPKITE_API_KEY"
Paging through reviews
/v1/reviews uses a cursor instead of page numbers. Each response includes next_page_token; send it back as page_token to get the next batch, and stop when it is absent.
- The token is opaque and bound to the place and
sortit came from. Send it unchanged with the sameplace_id(orcid/fid) andsort; anything else is a400. - Paging reaches the first 10 reviews of a place at most: Google Maps shows logged-out visitors only those, in its most relevant order, and usually only the first 5; SerpKite reads logged-out pages only. The last page Google shows has no
next_page_token, andsortother thanmost_relevantreturns400(not charged). - A page can hold fewer than
numreviews when Google stops loading early. Itsnext_page_tokenresumes right after it, and you’re billed for the reviews returned.
from serpkite import SerpKite
sk = SerpKite()
reviews, token = [], None
while True:
res = sk.reviews(place_id="ChIJLU7jZClu5kcR4PcOOO6p3I0", num=5, page_token=token)
reviews += res.results
token = res.next_page_token
if not token: # absent on the last page (at most 10 reviews)
break
print(len(reviews), "reviews")Related: Credits and billing, Common parameters.
Last updated: 2026-11-06