A newly published plugin may have the right functionality but very little installation history, few reviews and almost no independent coverage.A plugin can also remain small for years because it serves a narrow use case.
These are not identical situations. But they expose the same discovery problem:
a system may have less public evidence about the plugins it has the greatest difficulty finding.
Three different plugin populations
WP-Repo's repository work has encountered three different sizes. They describe different things, not three estimates of the same number.
- More than 120,000 historical plugin entries in the WordPress.org SVN namespace were observed in a local inventory. An SVN entry does not mean the plugin remains publicly available for installation. WordPress.org documents that a closed plugin's SVN repository may remain accessible even after its public download is withdrawn.
- Roughly 76,000 WordPress.org API-derived plugin records were present in a separate WP-Repo inventory. This is the broader catalog context, not the set selected for the following analysis. The public WordPress.org directory currently describes itself as offering *over 77,000 free plugins*; the published figure and a local database count need not refer to the same moment.
- 26,851 locally stored recordsmet WP-Repo's particular version-declaration filter in the analysis below. This is an explicitly restricted research group, not the number of plugins proved to work on current WordPress releases.
The historical SVN count, directory/API inventory and filtered research set must not be added together or treated as interchangeable.
Sources:
WordPress.org plugin directory and
WordPress Plugin Developer FAQ on closed plugins and SVN access].
What the filtered data actually shows
In November 2026, WP-Repo ran a grouping query against its locally stored WordPress.org-derived plugin metadata. The selection was:
WHERE tested_wp LIKE '7.%'This covers stored
Tested up to values beginning with
7., rather than only records equal to
7.1. It returned
26,851 records. The query grouped them by their stored
active_installs values:
| Reported active-installation class | Records | Share |
|---|
| Fewer than 10 | 13,251 | 49.35% |
| 10+ to 90+ | 5,961 | 22.20% |
| 100+ to 900+ | 3,376 | 12.57% |
| 1,000+ to 9,000+ | 2,582 | 9.62% |
| 10,000+ | 1,681 | 6.26% |
| Total | 26,851 | 100.00% |
Almost half of this selected group, 13,251 records, has no reported installation tier above "fewer than 10". That is the relevant finding for the adoption-signal problem. It is not necessary to assume that all 13,251 are new or that an AI assistant overlooked any particular one.
The stored value
active_installs = 0 maps to WordPress's
Fewer than 10 display. It does
not establish that literally zero websites use the plugin. The other labels are reported thresholds, not exact installation counts. See the
WordPress installer code reference.
Why the method matters
Tested up to is author-supplied metadata, not an independent compatibility certificate. A plugin whose declared test version does not start with
7. may still work on a newer site; conversely, a
7.x declaration does not guarantee reliable operation. WP-Repo selected this narrower group as a
practical research boundary, not as a universal definition of usable WordPress plugins.
The SQL counts matching database rows. The underlying data-import date and independent deduplication check are not established here, so these numbers should not be called a live census of the directory. Nor is this a study of newly released plugins. The result describes the stored research group evaluated in November 2026.
What these counts do and do not prove
The numbers
do show that a substantial share of this defined group lacks a meaningful installation-history signal. They
do not show whether these plugins have the required functions, whether their code is safe, whether they attract little demand or whether a specific AI recommendation system excludes them.
That distinction is essential. Installation history tells us about reported adoption. Discovery must also ask what each plugin does and whether a user can find it by describing a problem.
A recommendation process that leans heavily on adoption and reviews may favor the better-known candidate when both candidates deserve functional examination. Whether that happens, and how often, must be tested separately against a defined inventory.
The dataset measures the size of the low-adoption group. The discovery problem is whether that group gets a fair functional examination.