Google is shutting down the Custom Search API on January 1, 2027: How to check indexation moving forward

Google is shutting down the Custom Search JSON API-the interface programs used to fetch Google search results. The official documentation states outright: "The Custom Search JSON API is closed to new customers" and "Existing Custom Search JSON API customers have until January 1, 2027 to transition to an alternative solution." New sign-ups are already cut off, and existing customers have until January 1, 2027 to transition.

There was no standalone announcement. The deprecation notice slipped into the documentation on January 20, 2026, as confirmed by archive snapshots: the January 20 snapshot still shows the footer reading "Last updated 2025-08-28" with zero warnings, while the January 24 snapshot displays the notice and an updated date of 2026-01-20. Neither the developer blog nor the Google Search blog published an announcement.

This isn't Google's first move in this direction. The related Custom Search Site Restricted JSON API was shut down on January 1, 2025, when users were similarly redirected to Vertex AI Search. Around the same time, Google killed the ability to create new whole-web search engines: Programmable Search Engine documentation now explicitly states that the open-web search mode can no longer be created for new engines.

What Google offers instead

For searching your own sites, Google pushes Vertex AI Search, which it has rebranded to Agent Search. Official pricing sits at $1.50 per 1,000 requests for the Standard tier and $4 per 1,000 for Enterprise. However, this is search for your own data and sites: indexing a site requires verified domain ownership. It doesn't return Google's open-web SERPs. Furthermore, authentication changes: Programmable Search Engine ran on an API key, while Agent Search defaults to OAuth 2.0.

Google offers no public product for searching the open web. The deprecation page simply invites developers to contact them and express interest. No product name, no page, no pricing, and no terms. Finding out what it costs or who qualifies is impossible from the outside.

One paid method to tap into Google's web index remains: Grounding with Google Search via the Gemini API, offering 5,000 free requests per month followed by $14 per 1,000. But it doesn't return raw SERPs-it returns a generated answer with links. The annotations provide only URLs and titles, omitting snippets and rankings. That's not enough for index checking. You won't see whether a page matched a site: query; you'll just get a summary.

The JavaScript-based Programmable Search Element remains free. The JSON API is shutting down, not the entire Programmable Search Engine-don't confuse the two. That said, the paid, ad-free Element is also closed to new customers, though it lacks an official shutdown date for now.

Bing did the exact same thing, just earlier

Microsoft shut down the Bing Search API and Bing Custom Search API on August 11, 2025, following a May announcement. Its replacement, Grounding with Bing Search in Azure AI Agents, costs $14 per 1,000 calls. Microsoft's documentation explicitly states that developers lack access to raw search results: the tool returns context for model answers rather than standard SERPs.

In short, over the span of a year and a half, both major search engines eliminated direct access to their search results. What's left are closed partner deals reserved for select major enterprise players, alongside paid intermediaries.

The impact on index checking

Here's a crucial caveat usually glossed over in "how to replace Custom Search API" guides. The most obvious official route-the Search Console API-is fundamentally useless for our use case. The URL Inspection API documentation states that the checked URL must reside within a property you have verified in Search Console. Link builders check the indexation of third-party links on donor sites where they have zero Search Console access. By design, Search Console is useless for this, independent of any rate limits. (By the way, those limits are modest anyway: 2,000 requests per day per site.)

That leaves the site: operator in regular search. Google's stance here is unambiguous: its automated traffic guidelines explicitly classify rank-tracking and index-checking software under automated queries, which are met with reCAPTCHA challenges. Mass verification via standard SERPs remains an ongoing cat-and-mouse game against search engine defenses, not an authorized workflow.

What about Yandex?

The familiar "Yandex XML" is gone as a standalone service: xml.yandex.ru redirects to Yandex Search API, and its documentation has moved to Yandex AI Studio. Official pricing lists 488 rubles per 1,000 synchronous daytime requests and 30.5 rubles per 1,000 deferred requests. The sixteenfold price difference looks tempting, but the cheap tier comes with major strings attached: deferred requests are accepted only between 00:00 and 06:00 Moscow time, take at least five minutes to process, and results expire within twelve hours. Index-checking in Yandex is also offered as a standalone service by platforms in our ratings, such as LinkBooster.

There is no free tier; getting started requires a funded billing account with a linked bank card. Hard limits include up to 250 results per query, 400 characters, and 40 words per query. Index checks require url: and site: operators directly in the query text, as the web search method offers no dedicated fields for this. Yandex also issues a specific warning that the Search API processes requests outside of information systems certified under Federal Law 152-FZ.

Yandex doesn't offer a dedicated index-checking API either. Webmaster API works only for verified sites, monitoring important pages is capped at 100 URLs per site, and the "pages in search" history exposes a single host-level metric rather than granular page data. Other Russian engines aren't worth mentioning: Mail.ru has no search API, its developer portal retains only legacy "My World" social widgets, and Rambler Search migrated to r0.ru with no public API.

What the remaining options cost

Paid intermediaries that scrape search results fill the gap. Pricing verified on November 1, 2026, per 1,000 requests: DataForSEO starts at $0.60 for the deferred queue and $2 for live mode; Serper ranges from $1 on the entry-level plan to $0.30 at scale; Apify Google Search Scraper sits at $1.80; and SerpApi ranges from $25 on the minimum plan down to roughly $9 for enterprise volumes.

Brave Search API charges $5 per 1,000 requests with $5 in free monthly credits. However, it returns its own index rather than Google's, making it useless for verifying Google indexation. Exa and Tavily are similarly unsuitable, as they provide semantic search over proprietary databases rather than access to Google.

A blanket caveat for all these numbers: no independent benchmarks exist measuring how accurately these intermediaries replicate live search results. Every accuracy percentage they publish is self-reported.

Which ranked services are affected

We audited all eight active services on November 1: homepages, checker pages, FAQ sections, API docs, blogs, and Telegram channels. Not a single service publicly admits that its index checker runs on the Custom Search API. Almost none disclose their underlying mechanics either.

There is one exception. SpeedyIndex explicitly details its architecture: checks run against live search results using proprietary infrastructure with distributed proxies, executing site:URL queries with quoted addresses, claiming over 99% accuracy. Competitors offer vague marketing copy like "real-time checks" or say nothing at all. We couldn't find a public standalone checker service for Link Indexing Expert.

That's the honest takeaway: from the outside, you cannot tell whose index checker will break on January 1, 2027. Legacy keys for the deprecated API could sit behind any service-such keys keep functioning right up until the cutoff, and vendors aren't obligated to disclose their backend.

That these checking layers break in practice is already happening. NeedMyLink indexing reports understated results for a month due to a bug in an external status-checking service: after a recount, the average success rate across 326 projects jumped from 56.3% to 75.8%. The indexer itself operated normally the entire time; it was the measurement tool that failed.

What this means for our ratings: the numbers in the service cards don't change based on this news. But keep the date on your radar. If a service relied on Google's API for checks, the fallout will show up in its reporting accuracy after January 1, 2027. During our next data audit, we will add a question about verification methods to our vendor vetting checklist and explicitly tag services that answer transparently.