Google to shut down Custom Search JSON API by January 2027

What is happening

Google is retiring the Custom Search JSON API, one of the standard ways for programs to request web search results as structured data. According to a post on dev.to by a member of the Serpbit team, new sign-ups are already blocked, and the service will stop working entirely on January 1, 2027.

The post notes that Google's suggested successor, Vertex AI Search, is designed for indexing and querying a customer's own websites rather than the open web. That leaves a gap for anyone whose app, script or AI agent depends on results from the public internet. The field has also thinned elsewhere: Microsoft closed its Bing Search API in 2025, the post points out.

A migration path that reuses existing code

The bulk of the dev.to post is a pitch for Serpbit, the author's own service, which is built to accept the same requests and return the same response format as Google's Custom Search. In practice, the author says, migrating means changing the endpoint address and the API key while leaving parsing code and tests untouched. Google's official Python and Node client libraries are said to keep working, as do LangChain and LlamaIndex with a small adjustment, all of which the team says it tested against the live API.

The post is candid about a few gaps. Results come from the whole web, so the search-engine configuration identifier used to restrict queries to a specific site no longer applies; a site filter parameter is offered instead. The pagemap metadata block is missing from responses, and latency is higher, roughly two seconds per query rather than under one.

Beyond basic search, the author says a single key also covers AI Overviews, Google's AI Mode, Shopping, Maps, News, Images and Autocomplete, plus an MCP server that lets tools like Claude, ChatGPT or Cursor run searches after signing in. Responses are available as JSON or Markdown, and billing applies only to successful queries.

The pricing comparison, with a caveat

The post includes a cost comparison at 100,000 searches per month, drawn from each provider's public pricing: Serpbit at $75 with 1,000 free monthly searches and no card required; Serper at roughly $75 with 2,500 free queries to start; SerpApi at $725 with 250 free per month; Brave Search API at around $500 with a $5 monthly credit that requires a card; and Google Custom Search itself at $485 with 100 free queries per day until the shutdown.

It is worth stressing that this table was published by one of the vendors in it, so the framing favours the author's product. Some of its observations are still useful: Serper is described as a solid choice for new integrations that only need web results, SerpApi charges a premium for covering many search engines, and Brave runs its own independent index, which means its results will not match what users see on Google, a distinction that matters for SEO and rank-tracking workloads.

Why it matters

The Custom Search JSON API sits inside a long tail of integrations: internal tools, monitoring scripts, research helpers and, increasingly, AI agents that need live web results rather than a model's training data. A hard cutoff on January 1, 2027 forces every one of those integrations to be rewritten or redirected, and anyone with production dependencies should start by auditing where the API is actually used.

The bigger picture is that programmatic access to web search is consolidating. Bing's API is gone, Google is steering customers toward a site-search product, and the remaining options are largely third-party resellers whose pricing and longevity are unproven. Developers planning a move should test candidates against their real query mix well before the deadline, verify pricing directly with each provider, and treat vendor-published comparisons, including this one, as a starting point rather than the final word.