Report card

Do AI agents pick Typesense?

How often agents choose Typesense when a developer needs search — measured across Claude, GPT, Gemini.

Typesense

#4 of 5 search
4%Pick Rate
95% CI 2%8%

When developers ask an AI agent for search, Typesense is picked 4% of the time — ranking #4 of 5 measured across Claude, GPT, Gemini.

Picked 4% Named, not picked 32% Never surfaced 64%

Beaten by Meilisearch (17%), Algolia (14%), Elasticsearch (8%).

Awareness

640K package downloads / week

Thin presence in the data models learn from — an awareness problem. Get into more code, docs, and developer discussion.

0 docs in FineWeb (10BT sample) · 215 URLs in Common Crawl

In the raw web crawl, but thin in the filtered corpus models actually train on — an awareness gap at the source.

AdoptionDid the pick convert to installs? Public Adoption = npm downloads of your packages. The proof half of the funnel.

2.4M installs / month ▼ 8% MoM

Installs are declining. Worth pairing with the Pick Rate trend to find the leak.

Get the full report

The per-model and per-surface breakdown for Typesense — where it wins, where it loses, and to whom — plus an alert when the Pick Rate moves.

About Typesense

An open-source search engine you can host yourself or use via a managed cloud service.

Typesense is a search engine designed for developers who want fast, typo-tolerant full-text search without the operational complexity of larger search platforms. It exposes an HTTP API for indexing documents into collections and querying them, supporting features like filtering, faceting, and ranking. It fits use cases ranging from site search to product catalogs.

Where to start

Start by running a Typesense server (self-hosted or via Typesense Cloud), then use the JavaScript client library to create a collection, index documents, and issue search queries against the HTTP API. The official API documentation and the bundled examples cover the full workflow.

Install

Links and summary verified from public sources.

The rest of the Search ranking

← Full Search leaderboard · How this is measured