Google has published developer documentation for something it has never offered openly: an API that returns full web search results. It is called the Web Search Service, the docs went up quietly (both pages say last updated 9 September 2026), and the catch is in the first sentence, which says it “allows programmatic partners to develop websites and applications that retrieve and display Google Search results.”
Programmatic partners. Not you, not me, not anyone with a Cloud project and a card on file. I run my own tooling against Google’s open search APIs, so I read the new reference end to end. Here is what is actually in it, and what it means if you build on this stuff.
What the docs actually say
The overview and the REST reference are short, but concrete:
- One endpoint,
GET https://websearchservice.googleapis.com/v1:search, described as performing “a full web search”. REST or gRPC. - Three prerequisites: a Google Cloud project, an API key, and “a designated client_id associated with the partner agreement”. The client id format is literally
partner-[name]-[product]-[feature set]. No agreement, no access, and nothing anywhere explains how you get one. - The request requires the end user’s IP address. Region code is optional; the IP is not.
- Queries support normal search operators, language and region restriction, SafeSearch, and recency filters (days, weeks, months or years).
- Responses are ten blue links, up to 20 per page: title, display URL, snippet, MIME type, plus total-results and corrected-query info. No ranking position field, no ads, no AI Overview, no SERP features.
- No pricing, no quotas, no application process, and, notable given the year we are having, not one mention of AI, grounding or Gemini.
That required end-user IP is the detail I would underline twice. An API that must be told which human it is answering is specified for serving live users on a page, not for firing batch queries from a server. The rank trackers and SEO tools hoping this is their lifeline should read that field twice: the schema itself is shaped to exclude them.
Why this matters: the door that closed in January
The context is the piece most coverage touches only lightly. In January 2026, Google announced changes to its web search products: the Custom Search JSON API, the one anyone could pay $5 per thousand queries for, was closed to new customers and will be discontinued on 1 January 2027. The Programmable Search Element got capped at 50 domains, which ends its life as a full-web workaround. Enterprises were pointed at Vertex AI Search, and anyone needing full-web results was pointed at a contact form.
The Web Search Service is, as far as anyone can tell, what sits behind that contact form. I say “as far as anyone can tell” deliberately: the January post and the new docs do not mention each other, so the link is inference, widely drawn but confirmed nowhere. That is the fact worth being precise about when this gets repeated as settled.
Figures from Google's Web Search Service reference, Custom Search JSON API overview, and the January 2026 Programmable Search announcement.
What I am doing about it
This is not abstract for me. The bulk URL inspection tool on this site does its checking through the Custom Search JSON API, inside the documented 10,000 queries a day, and that API now has sixteen months to live. My plan is the plan I would give a client: keep running on it until the shutdown, do not architect anything new on it, and treat the partner API as unavailable until Google publishes an application path, because a door with no handle is a wall.
Two speculations worth refusing while I am here. First, that this API feeds AI answer engines: the docs say nothing about grounding, and Google routed grounding buyers to Vertex AI Search in January, so writing “Google built this for AI” is inventing a fact. The way AI products actually reach your site is still crawlers and licensing, which is why ChatGPT seeing your site remains a robots and rendering question, not an API question. Second, that regulators forced this open: an official full-web API gated by private contract is closer to the opposite, access on Google’s terms with Google choosing the counterparties.
For site owners the practical takeaway is small but real: another generation of search-adjacent products, from niche engines to research tools, will now be built either on Google-by-contract or on the independent indexes. Where your pages surface increasingly depends on indexes you never look at, which is the same argument I made in the AI search visibility guide for auditing more than just Google.
Sources
- Web Search Service overview, Google
- Web Search Service REST reference, Google
- Custom Search JSON API overview (discontinuation notice), Google
- Updates to our web search products, Google, January 2026
- SEJ’s report of 11 September 2026, which surfaced the docs