Google Custom Search API Shutdown: 7 Alternatives by Price

Google has closed the Custom Search JSON API to new customers and will switch it off for everyone on January 1, 2027. Here is what it cost, what replaces it, how to keep site-restricted search without a cx, and a function that returns the old response shape.
TL;DR
- •The Custom Search JSON API is closed to new customers, and existing customers have until January 1, 2027 to move, according to Google's API page on 2026-11-19.
- •For search across 50 or fewer of your own domains, Google points you to Vertex AI Search. For whole-web search, you need a third-party search API.
- •List prices per 1,000 queries on 2026-11-19: Serper $1, Exa from $4, Brave $5, You.com $5, Tavily $8, API Pick $15, SerpApi $25 on its entry plan. Custom Search was $5.
- •In our test, 20 of 20 site: queries through API Pick returned only pages from the named domain (100 of 100 results), so the cx site restriction can be replaced by a query operator.
- •API Pick costs three times what Custom Search did and returns 5 results per call, not 10. It suits agents and low-volume tools, not high-volume SERP replacement.
When does the Google Custom Search JSON API shut down?
Google has closed the Custom Search JSON API to new customers, and existing customers have until January 1, 2027 to move to something else. Both statements are on the API's overview page, which we read on 2026-11-19. Its pricing section adds that the current prices apply "only to existing Custom Search JSON API customers until the service discontinuation on January 1, 2027."
What that means in practice:
| Status on 2026-11-19 | |
|---|---|
| New API keys or projects | Not accepted |
| Existing customers | Working until January 1, 2027 |
| Free tier | 100 queries a day |
| Paid usage | $5 per 1,000 queries, up to 10,000 queries a day |
| Google's suggested replacement | Vertex AI Search, for searching up to 50 domains |
| Whole-web search from Google | No self-serve option; Google asks you to register interest through a form |
Source: Custom Search JSON API overview, Google for Developers, read 2026-11-19.
We build API Pick, one of the alternatives below. The prices in this article come from each vendor's own page on the same day, and there is a section on where API Pick is the wrong choice.
What should you migrate to?
Pick by what your Programmable Search Engine actually did: search your own sites, search the whole web for people, or feed an AI agent. The three cases need different replacements.
| What you used Custom Search for | Move to | Why |
|---|---|---|
| A search box over your own site or up to 50 domains | Vertex AI Search (Google Cloud) | Google's own recommendation; indexes and ranks only your sites |
| Whole-web results that look like Google's, at high volume | A SERP API (Serper, SerpApi) | Returns Google result pages as JSON, including 10 results per page |
| Search as a tool for an LLM or agent | An AI search API (Brave, Exa, Tavily, You.com, API Pick) | Returns cleaned snippets or page text sized for a model's context |
If you used a cx to restrict whole-web results to a handful of sites, any of the APIs in the last two rows may work with a site: operator. See the test below before you rely on it.
How much do the alternatives cost per 1,000 queries?
On list price, Serper is the cheapest at $1 per 1,000 queries, Exa starts at $4, Brave and You.com charge $5 (the same as Custom Search), Tavily $8, API Pick $15 and SerpApi about $25 on its entry plan. Every number below was read on the vendor's own page on 2026-11-19.
| Price per 1,000 queries | Free allowance | Results per call | Notes | |
|---|---|---|---|---|
| Google Custom Search (ending) | $5 | 100 queries a day | Up to 10 | Capped at 10,000 queries a day |
| Serper | $1 ($50 for 50,000 credits) | 2,500 queries once | Not stated | Google results; credits valid 6 months; cheaper in larger packs |
| Exa | From $4 (Lite); $7 (Auto) | $10 of credits a month | Up to 10 at base price | Own index; extra results $1 per 1,000 |
| Brave Search API | $5 | $5 of credits a month | Not stated | Own index; prepaid credits |
| You.com | $5 | $10 credit once | Not stated | Same price for 1 or 100 results |
| Tavily | $8 pay-as-you-go | 1,000 credits a month | Not stated | Basic search = 1 credit; cheaper on monthly plans |
| API Pick | $15 | 100 credits once (6 searches) | Up to 5 | Only HTTP 200 responses are charged; credits never expire |
| SerpApi | About $25 (Starter: $25 a month for 1,000) | 250 searches a month | Not stated | Google results; only successful searches count |
Sources, all read on 2026-11-19: Google Custom Search JSON API, Serper, Exa pricing, Brave Search API, You.com pricing, Tavily credits, SerpApi pricing, API Pick pricing. Vertex AI Search is priced per query and per GB of indexed data on Google Cloud's pricing page; we could not read the current figures from that page on 2026-11-19, so they are not in the table.
Two things the table hides:
- Result count changes the real price. Custom Search returned up to 10 results per call. If your code pages through results with
start=11, an API that returns 5 results per call needs twice as many calls for the same depth. - Daily caps go away. Custom Search stopped at 10,000 queries a day. We did not see a daily cap on any of the pricing pages above. Limits are stated per second instead where they are stated at all: Serper's entry pack, for example, allows 50 queries per second.
Can you keep site-restricted search without a cx?
Yes, with a site: operator, if the API honours it: in our test all 100 results from 20 site-restricted queries through API Pick came from the named domain. A cx let you pin a search engine to a list of sites. Most replacement APIs have no cx, so the restriction has to move into the query.
We sent 20 queries of the form site:<domain> <query> to API Pick's POST /api/search/web on 2026-11-18 (UTC), covering documentation sites, news, government, research and forums:
| Domain | Query | Results | On the named domain |
|---|---|---|---|
| docs.python.org | asyncio timeout | 5 | 5 |
| developer.mozilla.org | fetch abort controller | 5 | 5 |
| kubernetes.io | pod disruption budget | 5 | 5 |
| postgresql.org | jsonb index | 5 | 5 |
| react.dev | useEffect cleanup | 5 | 5 |
| nextjs.org | server actions revalidate | 5 | 5 |
| docs.github.com | fine-grained personal access token | 5 | 5 |
| stripe.com | idempotency keys | 5 | 5 |
| learn.microsoft.com | azure functions timeout | 5 | 5 |
| cloud.google.com | vertex ai search pricing | 5 | 5 |
| nih.gov | long covid fatigue | 5 | 5 |
| who.int | measles outbreak 2026 | 5 | 5 |
| bbc.co.uk | bank of england interest rate | 5 | 5 |
| sec.gov | form 8-k item 2.02 | 5 | 5 |
| arxiv.org | speculative decoding | 5 | 5 |
| wikipedia.org | model context protocol | 5 | 5 |
| reddit.com | custom search json api shutting down | 5 | 5 |
| stackoverflow.com | google custom search api 403 | 5 | 5 |
| europa.eu | ai act general purpose ai obligations | 5 | 5 |
| nasa.gov | artemis ii launch date | 5 | 5 |
All 20 calls returned HTTP 200. The median call took 3.5 seconds and the slowest 12.7 seconds. A subdomain counted as on-site (for example www.ncbi.nlm.nih.gov for nih.gov).
To restrict to several sites at once, run one query per site and merge, or use an API with a domain filter parameter (Exa and Tavily both document one).
How do you migrate your code?
Map three fields and keep your calling code unchanged: Custom Search's items[].title, link and snippet become title, url and snippet in most search APIs, including API Pick.
1List where Custom Search is called
Search your codebase for customsearch/v1, cse().list or build("customsearch". Note which calls use cx to restrict sites and which use num or start for paging.
2Map the response fields
| Custom Search JSON API | API Pick /api/search/web |
|---|---|
items[].title | results[].title |
items[].link | results[].url |
items[].snippet | results[].snippet (longer: cleaned page text, not a 160-character excerpt) |
items[].displayLink | Host of results[].url |
items[].pagemap | Not returned |
searchInformation.totalResults | Not returned (count of results) |
queries.nextPage / start | No paging; up to 5 results per call |
3Swap the call behind a compatible function
This function returns the same shape as cse.list(...).execute(), so code that reads items keeps working. The site argument replaces a single-site cx.
import os, requests
APIPICK_KEY = os.environ["APIPICK_API_KEY"]
def cse_list(q, site=None, **_ignored):
"""Drop-in for cse.list(q=..., cx=...).execute(): returns the CSE response shape."""
query = f"site:{site} {q}" if site else q
r = requests.post(
"https://www.apipick.com/api/search/web",
headers={"x-api-key": APIPICK_KEY},
json={"query": query},
timeout=30,
)
r.raise_for_status()
results = r.json().get("results", [])
return {
"searchInformation": {"totalResults": str(len(results))},
"items": [
{"title": x["title"], "link": x["url"], "snippet": x["snippet"],
"displayLink": x["url"].split("/")[2]}
for x in results
],
}
res = cse_list("rate limits", site="docs.github.com")
for item in res["items"]:
print(item["displayLink"], "|", item["title"])Output on 2026-11-19:
docs.github.com | Text: rate-limits-for-the-rest-api.md
docs.github.com | Rate limits for the REST API - GitHub Docs
docs.github.com | Rate limits for the REST API - GitHub Enterprise Cloud Docs
docs.github.com | Configuring rate limits - GitHub Enterprise Server 3.20 Docs
docs.github.com | Rate limits and query limits for the GraphQL API - GitHub Docs4Check it from the command line
curl -s -X POST https://www.apipick.com/api/search/web \
-H "x-api-key: $APIPICK_API_KEY" \
-H "content-type: application/json" \
-d '{"query":"site:docs.python.org asyncio timeout"}'5Run both side by side before January 1, 2027
While your Custom Search key still works, send your 20 most common queries to both and compare the top results. Switch over once the differences are ones you can accept, and remove the old key afterwards.
Where is API Pick the wrong choice?
If you need cheap, high-volume, Google-identical results, API Pick is the wrong replacement. Be clear about the trade-offs:
- It costs three times as much. A web search is 15 credits and $1 buys 1,000 credits, so 1,000 searches cost $15, against $5 on Custom Search and $1 on Serper.
- It returns 5 results per call and has no paging. Pages that showed 10 results, or let users click to page 2, need a different API.
- It is not Google's ranking. If your users expect the order Google shows, a SERP API is closer.
- There is no image search equivalent of Custom Search's
searchType=imageon the web search endpoint.
For a broader comparison of agent-oriented search APIs, see the best web search APIs for AI agents and our Tavily alternatives by price. If you are weighing Gemini's built-in search instead of an API, see Google Search grounding vs a search API. The endpoint itself is documented on the web search API page.
Frequently Asked Questions
Is the Google Custom Search JSON API being discontinued?
Yes. Google's API page says the Custom Search JSON API is closed to new customers and that existing customers have until January 1, 2027 to transition to an alternative solution. Its pricing applies only to existing customers until the service is discontinued on that date.
What is the best alternative to the Google Custom Search API?
It depends on what you used it for. To search 50 or fewer of your own sites, Google recommends Vertex AI Search. To keep raw Google-style results at the lowest price, a SERP API such as Serper ($1 per 1,000 on its entry pack) is closest. For AI agents that need clean snippets, Brave, Exa, Tavily, You.com or API Pick fit better.
How much did the Google Custom Search API cost?
100 queries a day were free. Additional queries cost $5 per 1,000, up to 10,000 queries a day. Those prices now apply only to existing customers until January 1, 2027.
Is there a free alternative to the Google Custom Search API?
Several have free allowances, but none is free at volume. On 2026-11-19, Serper gave 2,500 free queries once, Tavily 1,000 credits a month, Brave $5 of credits a month, Exa $10 of credits a month and You.com a $10 credit. API Pick gives 100 credits once, which is 6 web searches.
Can I still restrict search to specific sites without a cx?
Yes, with a site: operator in the query, if the API you move to honours it. We sent 20 site-restricted queries to API Pick's web search on 2026-11-18 and all 100 results came from the named domains. Test your own domains before you switch, especially small sites.
Sarah Choy is the CEO of apipick. She writes about building production-ready APIs for AI agents and LLM workflows.