Google Custom Search API Is Closing: What to Use by January 1, 2027

Google's Custom Search JSON API "is closed to new customers", and existing customers "have until January 1, 2027 to transition to an alternative solution" (Google, checked 28 September 2026). Until then it gives 100 queries a day free and charges $5 per 1,000 after that, which comes to about $485 a month at 100,000 searches. At that volume, the open-web search APIs below list from $50 to $800 a month: DataForSEO $60 (queued), Serper $75, Serpex $50 to $80, Brave $500 and Tavily $800 pay as you go. Google's own suggested replacement, Vertex AI Search, is described for "searching up to 50 domains", not the whole web.

Google's terms checked 28 September 2026 on its Custom Search JSON API overview. Other prices checked 1 October 2026 on each vendor's own pricing page or docs (Serper on 19 October 2026), linked in the table below.

What is changing, and when?

Google's overview page says two things, word for word:

• "The Custom Search JSON API is closed to new customers. Vertex AI Search is a favorable alternative for searching up to 50 domains."
• "Existing Custom Search JSON API customers have until January 1, 2027 to transition to an alternative solution."

So no new projects can start on it, and anything already running on it has to move before 1 January 2027. Counted from 19 October 2026, that's 74 days (13 left in October, 30 in November, 31 in December). Plan for your integration to stop working after that date and you won't be caught out.

Can you still sign up for a Custom Search JSON API key?

No. It "is closed to new customers". If you're starting a project today, pick another API from day one.

If you're an existing customer, the current pricing still applies until you move: "100 search queries per day for free", then "Additional requests cost $5 per 1000 queries, up to 10k queries per day." That daily cap works out to about 300,000 queries in a 30-day month.

Is Vertex AI Search a replacement?

For some teams. Google's words are "a favorable alternative for searching up to 50 domains." If your Custom Search engine covered your own site or a short list of sites, start there. If you used Custom Search to search the whole web, for an app, an agent or a research tool, Google's own description doesn't cover that case, and an open-web search API is the like-for-like move. We haven't priced Vertex AI Search here.

What do open-web search APIs cost compared with Custom Search's $5 per 1,000?

Monthly totals at list prices, before any free allowance, for one plain search per request:

How the numbers are worked out:

• Custom Search: 100 free a day is about 3,000 a month (100 × 30). At 10,000 a month, 7,000 are paid: 7,000 × $5 ÷ 1,000 = $35. At 100,000, 97,000 are paid: $485.
• Serper: credits are valid for 6 months, so the monthly figure is the pack price spread over the months it lasts: $50 ÷ 5 = $10 and $375 ÷ 5 = $75.
• Serpex: the rate depends on how much you top up in total: Starter for top-ups of $5 to $49, Standard for $50 to $499, Scale for $500 to $1,999. At 100,000 a month you'll pass $50 in the first month, so $65 is the realistic figure unless you top up $500 or more. With page content: 100,000 × 3 credits = 300,000 credits, × $0.50 or $0.80 ÷ 1,000 = $150 or $240.
• SerpApi isn't in the table: we checked only its Starter plan ($25 a month for 1,000 searches) and its largest plan (about $1.96 per 1,000), and neither matches these volumes.

What does the migration look like in code?

The old request is a GET with your key, your cx and the query in the URL:

import os
import requests

# Before: Google Custom Search JSON API
resp = requests.get(
    "https://www.googleapis.com/customsearch/v1",
    params={"key": os.environ["GOOGLE_API_KEY"], "cx": os.environ["GOOGLE_CX"], "q": "acme hotel"},
    timeout=30,
)

The Serpex request sends the key as a header and keeps q:

# After: Serpex Search API
resp = requests.post(
    "https://api.serpex.dev/api/search",
    headers={"Authorization": f"Bearer {os.environ['SERPEX_API_KEY']}"},
    json={"q": "acme hotel"},
    timeout=60,
)
resp.raise_for_status()
for r in resp.json()["results"]:
    print(r["position"], r["title"], r["url"], r["snippet"])

The steps:

1. Drop key and cx from the request.
2. Send your Serpex key as Authorization: Bearer <key>.
3. Keep q. Search operators such as site: go straight into the query, so a Custom Search engine limited to one site becomes q="acme hotel site:example.com".
4. POST JSON to https://api.serpex.dev/api/search. GET with query parameters also works.
5. Read results[]: each result has title, url, snippet and position. Update your parsing code to these names.
6. Optionally set include_content: true to get the page content of the top 5 or 10 results as Markdown in the same response. It's best-effort: content arrives for roughly 70% of requested results, and blocked or robots.txt-disallowed pages return content_error instead. It costs at most 3 credits for 5 pages and 6 for 10, and pages that don't arrive aren't charged.

Errors are never charged, and an empty result that couldn't be confirmed as genuinely empty is free. The full checklist, with the response format, is on our Google Custom Search API alternative page.

Which replacement should you pick?

• You need Google's own results page (rank tracking, SEO tools): a SERP API. Serper returns the Google results page as JSON from $1.00 down to $0.30 per 1,000, prepaid with no subscription, with country and language targeting. DataForSEO costs $0.60 per 1,000 if you can wait in a queue and $2.00 live. SerpApi covers the widest range of engines and verticals and is the reference choice for rank tracking, at a higher price per search.
• You search up to 50 of your own domains: Vertex AI Search, on Google's own description.
• You feed web results to an app, an agent or an LLM: a search API built for that. Serpex costs $0.50 to $0.80 per 1,000 and can return page content in the same call. Brave ($5.00) returns results plus LLM context. Tavily ($8.00 pay as you go, $5.00 on Growth) has answer synthesis and mature LangChain and LlamaIndex integrations. Exa ($4 to $7) bills page contents separately at $1 per 1,000 pages.

Where Serpex isn't the best fit: if the product is a faithful copy of Google's results page, use a SERP API. On price per plain search, Serper's largest pack ($0.30) is below our lowest rate, and DataForSEO's queue ($0.60) is below our Standard and Starter rates. The wider comparison is in Search API Pricing in 2026 and on our alternatives page.

What's the migration checklist, with dates?

• By 31 October 2026: list every service that calls googleapis.com/customsearch/v1, its daily query volume, and whether its cx limits search to particular sites.
• By 15 November 2026: decide which kind of replacement you need (results page, up to 50 domains, or open web) and price it at your volume with the table above.
• By 1 December 2026: run the new API alongside Custom Search on real queries and check the results your users rely on.
• By 15 December 2026: move production traffic, and keep Custom Search as a fallback while it still answers.
• Before 1 January 2027: remove the old key and cx from code and secrets. That's Google's deadline for existing customers.

FAQ

When does the Google Custom Search JSON API shut down?

Google says existing customers "have until January 1, 2027 to transition to an alternative solution". New customers already can't sign up.

Can I still get a Custom Search JSON API key in 2026?

No. Google says the API "is closed to new customers".

Is Vertex AI Search the official replacement?

Google calls it "a favorable alternative for searching up to 50 domains." For whole-web search, use an open-web search API.

How much does Custom Search cost until it closes?

"100 search queries per day for free", then "$5 per 1000 queries, up to 10k queries per day." That's about $35 a month at 10,000 searches and $485 at 100,000.

What changes in my code when I move to Serpex?

Drop key and cx, send Authorization: Bearer <key>, keep q, POST JSON to https://api.serpex.dev/api/search, and read results[] (title, url, snippet, position).