FAQ
Questions, and straight answers
What this site is, where the data comes from, how the AI categories are decided, and what you are allowed to do with any of it.
What this is
The basics
An independent record of the public WordPress plugin directory. It keeps its own copy of every plugin slug the directory has published, fetches the full public metadata for each one, and adds a two-level category taxonomy built by a language model. The site is the front end of that record.
No. WP Analyzer is not affiliated with, endorsed by, or connected to the WordPress project, the WordPress Foundation, Automattic, or any plugin author listed on it. It reads the same public endpoints your own WordPress install reads, and nothing else.
No. There is no account, no login, no export gate and no email capture anywhere on this site. Every figure is on a page you can open.
No. It is a measurement of popularity, or of ratings, or of release count - whatever that board says it ranks. Popularity and quality are different things, and a plugin can sit at the top of the install board and still have had serious security advisories. Judge a plugin on its own page, not on a rank.
Where it comes from
The data
Two public wordpress.org sources: the plugin index, which lists every slug the directory hosts or has hosted, and the plugin information API, which returns a plugin's full public record. Nothing is scraped from a rendered page and nothing private is involved - these are the same endpoints a WordPress dashboard uses to show you plugin search results.
Two clocks. The totals, distributions and leaderboards are recomputed from the stored records once a day, and each section carries a stamp saying when the copy you are reading was built. The stored records themselves are re-fetched in continuous passes, so an individual plugin's row is as current as the last time its slug came up. The method page has the detail.
Because it includes plugins the directory no longer serves. A slug enters this record when it appears in the index and is never removed, so the headline count is every plugin ever published - and a separate figure beside it is the number still live today. Both are on the statistics page, with the closed and not-found split.
Partly because that is genuinely what the directory looks like, and partly because the directory never publishes an install figure below ten. Everything with fewer than ten sites - including everything with none - reports the same floor. The bottom band on the install chart is that floor, not a measurement.
Because it excludes unrated plugins. An unrated plugin is stored with a rating of zero, and averaging across the whole directory drags the figure under two stars - which is a statistic about missing reviews, not about software. Every average here is taken over rated plugins only, and says so where it appears.
The API field for all-time downloads returns nothing for every plugin - it is no longer populated. The column exists in the record and is empty, so no page here ranks or reports it. A plugin download total published anywhere else is worth asking questions about.
How classification works
The AI categories
Tags are free text typed by the plugin's author, with no hierarchy and no definitions - there are 65,252 distinct ones in use, and nothing stops "post", "posts" and "gutenberg" describing the same plugin. That works for a search box and fails for a count: you cannot ask a tag cloud what share of the directory is about security. A two-level taxonomy can answer that.
A language model is given the plugin's name, description and tags, together with the entire taxonomy built so far, and answers in a single call: either a category and sub-category that already exist, or a proposal for a new one. A proposed name is then compared against every existing name and folded onto a close match rather than creating a near-duplicate.
Yes. The honest framing is that consistency has been measured and correctness has not, because there is no ground-truth categorisation of the plugin directory to measure against. What can be said: re-running a plugin through the same model used to change its top-level category two times in three, and after the prompt was rebuilt to carry the whole taxonomy in one call it agreed forty times out of forty. The numbers and the method are here.
Because classification is the one pass that costs a model call per plugin, so it advances in batches rather than over the whole directory at once. 11,364 plugins have been through it so far. The shape of the taxonomy is meaningful; the counts inside it are counts of that sample. The coverage figure is stated at the top of the categories page.
Nobody. There was no starting list. The first plugin through the classifier had an empty taxonomy and proposed the first category; everything since has either matched something that existed or added to it. That is why the category names read the way they do - they are what a model reaching for a label actually reaches for.
Practical
Using it
Yes - please say where it came from and when you read it, because these figures move. If you are quoting an average, a distribution or a category count, read the limits section first; most ways of getting a directory statistic wrong are listed there.
Not publicly, not yet. Everything the site shows is derived from two public wordpress.org endpoints, so the underlying data is available to anyone willing to fetch it - what is here is the reading of it, not exclusive access to it.
If a metric is wrong, it is wrong at the source: every number here is copied from your plugin's own API record, so correcting it on wordpress.org corrects it here on the next pass. If the AI category is a poor fit, that is this site's doing and worth telling us about - placements are published precisely so they can be challenged.
If you want to understand the directory as a whole, start with the statistics - the install distribution is the chart that changes how most people read a plugin count. If you are looking for specific plugins, the rankings are faster.
Still curious?
The method page is the long version of every answer above.