Google Custom Search JSON API shuts down January 1, 2027, and Vertex AI Search is not a drop-in. The fastest migration is to Scavio, which answers the same request at https://api.scavio.dev/customsearch/v1: change the host and the key, keep your parser. If you would rather move to a richer response, Scavio also returns Google results as structured JSON from one POST endpoint. The three fields most Custom Search integrations use — title, link, and snippet — map directly to Scavio's organic_results[], and you also get Knowledge Graph, People Also Ask, and AI Overviews at no extra cost. The API is already closed to new customers.
The complete map from every Custom Search parameter and response field to Scavio, with runnable before/after code in Python, JavaScript and cURL and a per-1,000 cost table, is in the Google Custom Search JSON API migration guide.
Keep your code instead: Scavio also answers the Custom Search request unchanged at GET https://api.scavio.dev/customsearch/v1 -- same parameters, same items[] response -- so the migration can be just a host and key change. See the Custom Search compatible API docs.
What is actually changing
In January 2026, Google announced the end of the Custom Search JSON API and of free full-web search on Programmable Search Engine (formerly Custom Search). The dates and limits that matter:
- January 1, 2027 the Custom Search JSON API is retired; existing customers must migrate before then.
- The API is already closed to new customers, so you cannot create a fresh key today.
- Free Programmable Search Engines are being limited to a 50-domain scope, and the search-the-entire-web option is going away.
If you built anything that treats Google as a programmatic data source, RAG grounding, an agent's web-search tool, rank tracking, competitive monitoring, this is a hard deadline, not a soft deprecation.
Why Vertex AI Search is not a drop-in
Google points migrating users to Vertex AI Search. It is rarely a like-for-like swap:
- It is built to search your own corpus (your documents, your data store), not to return live public Google results.
- Pricing starts around $2 per 1,000 queries on the basic tier and climbs with extensions, a different cost model from a simple per-query SERP API.
- It carries enterprise setup and minimums that are overkill if all you wanted was "give me Google's results as JSON."
For teams whose code just needs query in, ranked results out, Vertex AI Search means re-architecting around a product you did not ask for.
What a real drop-in looks like
The migration most developers actually want is to a search API that returns Google SERP data as structured JSON over a single HTTP call. You swap the endpoint and the response parser, keep your query logic, and you are done.
That is what Scavio does. One endpoint returns organic results with link, title, and snippet fields your existing code can map onto the old Custom Search shape, and it returns clean JSON instead of raw SERP HTML.
A minimal call:
curl -X POST https://api.scavio.dev/api/v2/google \
-H "Authorization: Bearer $SCAVIO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"query": "best web search api"}'You get back organic_results[] with link, title, and snippet, the three fields most Custom Search integrations relied on.
How to migrate before the deadline
- Inventory your calls. Find everywhere you hit
customsearch/v1and note which response fields you consume. - Pick a replacement that returns live web results, not a private-corpus product. Confirm it covers the engines you need (Google is table stakes; you may also want Maps, News, or Shopping).
- Map the response. Point your parser at the new JSON shape; for most apps this is a few lines.
- Test on real queries and compare result quality before you cut over.
- Cut over well before January 1, 2027 so you are not migrating during an outage.
If you also pull data from Amazon, YouTube, Reddit, Walmart, or TikTok, consolidating onto one API with a single key removes the multi-integration tax at the same time. See the Google search docs or start free with 50 credits to test the swap.
FAQ
When does the Google Custom Search JSON API stop working? January 1, 2027. It is already unavailable to new customers.
Is Vertex AI Search a direct replacement? No. It searches your own data corpus and is priced and architected for enterprise use, not live public Google results.
What is the closest drop-in? A real-time SERP API that returns Google results as JSON from one endpoint, so you only change the endpoint and the response mapping.