Google Custom Search API is shutting down: what to use instead before 2027

Google’s Custom Search JSON API is going away. It is closed to new customers, and Google lists January 1, 2027 as the date the service is discontinued. If your app calls it today, you need a new source of search results before then.

Google suggests Vertex AI Search for site search and a request form for whole-web search. If your app needs the real google.com page, a SERP API is the closest match, and the code change is small.

What is happening to the Custom Search JSON API?

Google has closed the Custom Search JSON API to new customers and lists January 1, 2027 as its discontinuation date. Google explains this on its Custom Search JSON API overview page, last updated February 18, 2026.

The change was first announced on January 20, 2026, on the Programmable Search Engine blog. That post says your move “needs to be completed by January 1, 2027.”

We checked these pages on October 9, 2026. Google has not said what the API will return after the cutoff, such as an error code. Treat the date as a hard stop.

Who is affected?

Three groups need to act.

• Apps that call the Custom Search JSON API. This is anyone with a key and a search engine ID (the cx value) who sends requests to customsearch/v1.
• Search boxes set to “Search the entire web.” These engines keep working only until January 1, 2027.
• Search boxes that cover more than 50 domains. Google asks these users to contact it about a full web search product.

One group is fine. If you use Google’s free search box on your own site and it covers 50 domains or fewer, Google says that option stays.

Why your Custom Search results were never quite Google

Some teams used this API as a cheap stand-in for google.com. It never matched.

Google’s own help page, Programmable Search Engine vs Google.com, lists the reasons. An engine favors your own sites. Whole-web engines search only part of Google’s index. And they leave out answer boxes, real-time results and personalized results.

The API also had hard limits. Each call returned at most 10 results. A single query could never go past result 100, says the API reference. There were no ads, maps, People Also Ask boxes or AI Overviews. A real results page holds all of those, as we show in SERP features explained.

So the shutdown is also a chance to fix a data problem you may not have noticed.

What Google suggests instead

Google names two paths.

1. Vertex AI Search, for up to 50 domains. This is Google Cloud’s search product for your own sites and files. Google’s docs say it is being renamed Agent Search. It searches the content you give it, not the whole web.
2. A “full web search solution”, by request. For whole-web search, Google asks you to fill in an interest form. It has not posted a public price.

Neither one returns the page a person sees on google.com. If that is what your app needs, you need a different kind of tool.

Other options: SERP APIs and independent search APIs

A SERP API fetches the real Google results page for a keyword, country and device. It returns that page as JSON. Rank trackers, SEO tools and AI agents use them. For agents, our guide on how to give an AI agent live Google search shows a full setup.

An independent search API runs its own index of the web. Brave Search is one example. Its results come from Brave’s index, not Google’s.

Bing is no longer an option. Microsoft retired the Bing Search APIs on August 11, 2025. Our post on what replaced the Bing Search API covers that change.

Custom Search API alternatives compared

Prices below come from each provider’s own pricing page, checked on October 9, 2026.

A fair note on our own row: Serplify, which publishes this blog, only bills a search that succeeds and has no monthly plan. For a wider field, see our comparison of SerpApi alternatives.

How to move a Custom Search request to a SERP API

Often the change fits inside one function. Here is a typical Custom Search call in Python:

import os, requests

res = requests.get(
    "https://customsearch.googleapis.com/customsearch/v1",
    params={
        "key": os.environ["GOOGLE_API_KEY"],
        "cx": os.environ["SEARCH_ENGINE_ID"],
        "q": "best running shoes",
        "gl": "us",
        "num": 10,
    },
)
for item in res.json().get("items", []):
    print(item["title"], item["link"], item["snippet"])

Here is the same search sent to our SERP API, using the request from the quickstart. The standard format keeps organic results, ads and featured snippets:

import os, requests

res = requests.post(
    "https://api.serplify.io/v1/serp/search",
    headers={"Authorization": f"Bearer {os.environ['SERPLIFY_KEY']}"},
    json={
        "keyword": "best running shoes",
        "location": {"code": 2840},
        "language": {"code": "en"},
        "format": "standard",
    },
)
res.raise_for_status()
for item in res.json()["data"]["items"]:
    if item["type"] == "organic":
        print(item["title"], item["url"], item["description"])

The fields map like this:

Three things change for the better. You get the real page, not a filtered copy. You can pick a city and a device, as well as a country. And each item has a type, so you can keep or skip ads, maps and question boxes.

Two things to watch. Results will differ from what your engine returned, so check any tests that expect exact URLs. And if you used Custom Search to search your own site, Vertex AI Search may fit better than a SERP API.

Which option should you pick?

If you were thinking of scraping Google yourself instead, read that post first. Google blocks plain scripts within a few dozen searches. If you want the same results page your users see, a Google SERP API is the most direct swap.

Whatever you pick, start now. Run the new source next to the old one for a few weeks, compare the output, and switch well before January 1, 2027.