Is the Google Custom Search JSON API being discontinued?

Yes, on a specific timeline. Google's own documentation states the Custom Search JSON API is already closed to new customers and that existing customers have until January 1, 2027 to transition to an alternative. Current customers keep their existing access and pricing until that date, but nobody can sign up for a new key, and every customer needs a migration plan before the deadline.

What is the best alternative to the Google Custom Search API?

There is no single best alternative, because the Custom Search JSON API returned actual Google results at a low, transparent price, and no successor does both. If actual Google results matter most, Serper or SerpApi are the closest replacements. If independence from Google's terms of service matters more than exact result parity, Brave Search API runs its own index. If the goal is pre-processed, agent-ready content rather than a raw results list, Tavily or Exa are built for that job.

Is Serper or SerpApi better for AI agents?

Both return Google SERP data through a REST API, so the choice comes down to price transparency and tooling. SerpApi publishes full pricing, with a free 250-searches-a-month tier and paid plans starting at $25 a month for 1,000 searches, plus an official MCP server. Serper advertises 2,500 free queries on signup, but its own pricing page was not publicly loading at the time of writing, so its per-query cost cannot be confirmed from a first-party source, and it has no official MCP server, only community-built ones. Confirm current Serper pricing directly before budgeting for it.

What is the difference between Tavily and Exa?

Both are AI-native search APIs built to hand an agent clean, ready-to-use content, and both ship official MCP servers. Tavily is priced in credits ($0.008 per credit on pay-as-you-go, 1,000 free credits a month) and is commonly used as a general-purpose research tool for agent workflows. Exa is priced per 1,000 requests ($7 for standard search, more for deep search) and uses a neural, embeddings-based index plus dedicated company and people research tools, which makes it a common choice for prospect research agents specifically.

How do AI agents read the full content of a web page?

A search result is a title, a URL, and a short snippet, not the full page, so most agent pipelines add a step that fetches the URL and converts it to clean, LLM-readable text. Common options are Jina Reader (a free, URL-prefix reader with paid tiers for higher rate limits), Firecrawl (a scraping and crawling API with a free tier and paid credit plans), and, for teams already calling the Claude API, Anthropic's server-side web fetch tool, which retrieves and cites page content directly for no charge beyond standard token costs.

Google Custom Search API Alternatives: Serper vs SerpApi vs Brave vs Tavily vs Exa

Serper, SerpApi, Brave, Tavily, and Exa compared for AI agent search, plus exactly what changes when Google closes the Custom Search JSON API on January 1, 2027.