How the extension catalog is built
Everything in this catalog comes from the public Chrome Web Store, read by our own crawler. This page says how and how often we read it, what the numbers mean, and what we deliberately do not know or do. Last full pass: 2026-10-10, with 363 995 listings in the store.
How and how often we collect
We read every public listing up to 2 times a day, in passes scheduled from midnight UTC. A pass is paced to finish within 6 hours and asks the store for 50 listings per request, so the store sees a slow, steady trickle rather than a burst.
A listing's history gets a new point only when something on it changes. A flat line between two points means nothing moved, not that we stopped looking.
Public pages show up to 90 days of history. Longer history and analysis are in the web app.
What the user count means
The user count is the figure the store prints on each listing. Google does not publish how it is counted, and it is rounded, so treat it as the store's own estimate of an audience rather than a measurement of daily use.
When a store page stops publishing the figure, we keep the last one it published and say so beside it, rather than print a zero the store never stated.
Where permissions and warnings come from
Permissions and site access are what each listing's manifest declares, as the store serves it. The install warnings are the ones Chrome shows for those permissions, matched from Chrome's public documentation.
We do not install extensions, run them or read their code. A broad permission is a fact worth checking, not a verdict on intent.
Which pages we offer to search engines
A listing page is offered to search engines only when the listing is still in the store, its description is at least 80 characters long, at least one pass has observed it, and at least 2 blocks of the page have content of their own: permissions, audience history, versions, publisher or ratings.
One exception to the first condition: a listing that left the store stays on offer for 12 months after it disappeared if it had at least 10 000 users when it went. Its page then says when it left, what it was and what to use instead.
An alternatives page needs at least 3 comparable listings, and a list or category page at least 8 qualifying rows. Pages below these thresholds stay reachable and are marked noindex.
How lists, trends and category norms are built
Each curated list starts from titles that name the job, with its own floor on audience and ratings, and is widened by listings close in meaning to the strongest matches. A listing joins by meaning only when two of those matches agree on it. Everything is then ordered by current audience, which is not a verdict on quality.
Trends rank the 7-day change in users against each listing's audience a week earlier, and leave out listings that had fewer than 1 000 users then. Category shares are stated only for groups of at least 30 listings.
Nobody pays to appear in these lists or to rank higher in them. The order comes from the numbers above.
What we do not know
Why a listing left the store: we record that it disappeared and when, never the reason. How many people use an extension every day. How well an extension does its job, such as the quality of a voice or a translation. Anything that happened before our first pass.
What we do not do
We do not collect, store or publish developers' contact details. E-mail addresses, postal addresses and phone numbers from store listings are not kept, and lists of them are not for sale.
We do not keep the text of listing descriptions, only their length. We do not host, mirror or link to extension packages (CRX files).
AI tool plugin marketplaces
The catalog is gaining a second family: plugins for AI tools, such as MCP servers and plugins for Claude Code, Codex and Cursor, read from public marketplaces and registries under the same rules as the extension catalog. Each marketplace is read at its own polite pace and has its own off switch, and a marketplace that asks us to slow down is slowed down and then left alone for a day.
From plugin manifests we keep business and technical facts only. Contact fields such as the author or owner email, maintainer lists, phone numbers and postal addresses are discarded before anything is stored, and no page shows them.
Figures from different marketplaces are never added together. Each number is named after the marketplace that published it, and a marketplace that publishes no number shows none rather than a zero.
Where to go next
Where the data comes from: how the catalog is built