Google Search API Cost in 2026: Hidden Fees + the Shutdown

Google closed its Custom Search JSON API to new customers. If you already use it, you have until January 1, 2027 to get off it. That deadline is sitting quietly in the official documentation , one paragraph long, easy to miss.

We think that deadline changes how you should read every number in this post. The question is no longer whether the official Google Search API cost is fair at $5 per 1,000 queries. The question is what you migrate to, and what that will really cost you.

The official API bills $5 per 1,000 queries with 100 free per day and a hard cap of 10,000 per day. Google’s suggested replacements cost more: Vertex AI Search runs $1.50 to $4.00 per 1,000 (and only searches 50 domains), Gemini grounding runs $14 per 1,000. Third-party SERP APIs deliver the actual results page from $0.50 per 1,000.

We run a real-time Google Search API that processes millions of SERP requests, and a good share of our recent signups are Custom Search users working through exactly this migration.

This post is the cost autopsy we walk them through: what the old API really charged, what Google offers now, and the math at every volume.

What the official Google Search API costs in 2026

Let’s put every official option on one table first, then dig into the parts the pricing pages don’t spell out.

Read that last column again. The only product on the list that both searches the open web and survives 2027 costs $14 per 1,000 requests, 2.8 times the price of the API it replaces. And it doesn’t even return a results page.

All prices verified on Google’s own pricing pages in August 2026: Custom Search overview and Agent Search pricing . They can change, so check before committing.

The January 2027 shutdown, explained

The Custom Search JSON API (you may know it as Programmable Search) has been the “official” way to query Google programmatically for over a decade. Google has now closed it to new customers. Existing customers got a migration deadline: January 1, 2027.

Google’s documentation offers two paths forward. Vertex AI Search, which supports up to 50 of your own domains. Or “contact us” for a full web search solution, which is enterprise-sales language for pricing you won’t see on a public page.

Neither path is what most Custom Search users actually bought the product for: programmatic access to search results across the whole web, at a price a small team can afford.

Four hidden costs that were always in the fine print

Even before the shutdown, the sticker price of $5 per 1,000 undersold what Custom Search actually charged. These four mechanics are where budgets quietly leaked.

We bring these up because they set the evaluation checklist for whatever you migrate to:

• How is a billed unit counted?
• Is there a volume ceiling?
• Do you get the real results page?
• What happens on empty or failed calls?

Those same questions apply across every data source, by the way. We apply the per-request billing answer uniformly, whether you’re pulling SERPs, scraping Amazon product data , or extracting Google Maps business listings .

What Google wants you to use instead (and the real math)

Google’s two suggested destinations solve different problems, and neither is the problem Custom Search solved. Here’s what each one is.

The Vertex trap catches people in the migration rush. It’s the path Google’s docs point to, the free trial is generous, and the per-query price looks reasonable. Then you discover the 50-domain limit and the indexing meter, and realize you’ve been sold site search.

Gemini grounding is genuinely useful if what you want is answers. If what you want is data (positions, URLs, titles, the actual page), you’d be paying $14 per 1,000 to have an AI summarize information you never get to see raw.

Google search API cost calculator: 1K to 1M queries per month

Here’s the number everyone searches for. Monthly bill by volume, official options against the main third-party providers. To plug in your own volume: multiply your monthly queries (minus the free allowance) by the per-1K rate, or take the flat plan that covers you.

Official prices verified on Google’s pages in August 2026; third-party rates come from each provider’s published pricing. CSE assumes even daily usage (its 100/day free tier is worth ~3,000 queries a month).

*Serper requires a $50 minimum top-up and credits expire; DataForSEO requires a $50 minimum deposit, and its $0.60/1K rate is async delivery (submit a task, poll later). Live delivery there costs $2.00/1K, which makes the 100K column $200 and the 1M column $2,000.

Our column: the Pro plan at $19.99/month includes 15,000 requests. Ultra at $49.99 includes 50,000, so 100,000 requests means Ultra plus overage: $124.99.

The 1M figure is Mega ($149.99, 300K included) plus 700K of overage at $0.50/1K. Every plan and overage rate is public on the Google Search API pricing page , nothing gated.

Migrating from Custom Search: what actually changes

This is the part we know best, because we built our Google SERP API with autocomplete included around the exact complaints Custom Search users had. Four differences matter most in practice.

Geo-targeting deserves its own sentence. Google serves a different results page depending on the searcher’s country, so our geo-targeted search results API serves every request for the exact gl country and hl language you specify. If your users are in Germany, you see the German SERP they see.

The switch itself takes three steps:

1. Get a free key. The FlyByAPIs free plan includes 100 requests a month, no credit card, and it renews every month.
2. Replace the endpoint and keep your pagination. Both APIs hand back a page of about ten results, so your existing loop maps onto the start offset with almost no change.
3. Point your existing parser at the new response. The payload is flat JSON with organic_results, position, and domain fields, so most parsers need only field-name changes.

Here’s the migration in code. If you’ve called Custom Search from Python, this will look familiar, minus the API key dance and the daily cap:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

import requests

url = "https://google-serp-search-api.p.rapidapi.com/search"
headers = {
    "x-rapidapi-key": "YOUR_RAPIDAPI_KEY",
    "x-rapidapi-host": "google-serp-search-api.p.rapidapi.com",
}
params = {"q": "best wireless earbuds 2026", "num": 10, "gl": "us", "hl": "en"}

data = requests.get(url, headers=headers, params=params).json()["data"]

for r in data["organic_results"][:5]:
    print(r["position"], r["title"], r["domain"])

print(len(data["organic_results"]), "results, one billed request")
# for positions 11+, pass start=10, start=20 ... each page is its own request
print("PAA questions:", data["people_also_ask"][:3])

The response also carries real_query (what Google actually searched after spell correction) and people_also_search_for. There’s a separate autocomplete endpoint included in every plan at no extra cost, which is how several customers do long-tail keyword discovery. Full parameter reference lives in the Google Search API docs .

One thing we do that doesn’t show up on a pricing table: we monitor every endpoint continuously for changes on Google’s side and ship fixes within hours of detection.

SERP markup changes constantly (it’s a big part of why scrapers break ), and it’s the kind of breakage you usually discover in your own error logs.

We’d rather catch it before your pipeline does. The same monitoring runs across our whole suite, from the Crunchbase company data API to the jobs search API and the AI document translator .

Can you still get Google results by API for free?

Yes, within limits, and the limits are the whole story. Every “free” tier here is sized so you can test your integration, and nothing more.

Our advice, and it applies to any provider: use the free tier to validate the response format against your parser, then price the paid plan at 1.3x your expected volume to cover retries and growth.

A free tier that renews monthly (rather than a one-time batch) also gives you a permanent staging environment. That’s why the free tier on our Google search data API renews every month.

If you’re comparing the wider market beyond this migration, we’ve published the full 10-provider teardown in our cheapest SERP API comparison , including the billing-model traps that don’t appear on pricing pages.

Which option fits your migration

Four honest recommendations, depending on what you actually built on Custom Search.

The deadline does the deciding

Here’s the uncomfortable part: doing nothing is also a decision, and on January 1, 2027 it becomes an outage. Migrations always take longer than planned, and rewriting a search integration in December with a hard deadline is nobody’s idea of fun.

The math in this post points one way for open-web workloads. The official route costs $14 per 1,000 and you never see the raw results. The third-party route costs $0.50 to $2.00 per 1,000 and hands you the full results page, deeper than Custom Search ever did.

Start with the Google Search API free plan , point it at your existing parser, and see what breaks. Usually the answer is “less than you’d think”: the response is flat JSON with organic_results, positions, and domains.

P.S. If you’re mid-migration and something about your Custom Search setup doesn’t map cleanly (weird pagination, site-restricted queries, odd locales), write to us. We’ve probably seen your exact case this year, and if we haven’t, we want to.

Oriol.