WordPress SEO: the plugin finishes its list, and its list is the small half

We audited 395 WordPress sites and 287 of their blog posts with our own engine on 20 August 2026. On the 332 sites running an SEO plugin, the things the plugin manages are effectively solved: canonical tags and indexability were clean on 100%. What remains is owned by the theme and the host, and heading order, page weight, render blocking and security headers each failed on roughly half to three quarters of pages.

What we measured, and what the sample is not

395 sites detected as WordPress from their own markup, drawn from a pool of small business and agency websites. 332 of them advertise an SEO plugin in the page source: 214 Yoast, 92 Rank Math, 24 All in One SEO, 2 SEOPress. For 287 of those sites we also found and audited a real blog post, discovered through the site’s own WordPress REST API or the plugin’s post sitemap, so that checks about authors and dates were measured on a page that has an author and a date.

Every page went through the same 36 check audit we run for customers, with no paid endpoints. Not clean means fail or warn. The full aggregate is public.

Three limits, stated up front:

• The sample is not published per site. It comes from a private prospect pool, so publishing a row per domain would publish a list of named businesses alongside what is wrong with their websites. The aggregate is enough to check every figure here.
• “No plugin advertised” is not “no plugin”. 63 sites showed no plugin marker, and some of those will have one with its output suppressed. Treat that group as a weak comparison rather than a control.
• Response time and caching latency are excluded from the published table, because our engine measures them on a single cold fetch from one location. That is directional and it is not a benchmark, so no number is quoted here.

The plugin’s own list is finished

This is the part that deserves saying plainly, because the genre this article belongs to usually does not say it. On the 332 sites running an SEO plugin:

Check	Homepages not clean	Posts not clean
Canonical URL	0%	0.3%
Indexable, no accidental noindex	0%	0%
URL structure	0%	1.4%
Open Graph and Twitter cards	0.3%	0.7%
Language and hreflang	0.6%	1.4%
XML sitemap	2.1%	0.7%
robots.txt	2.4%	1.4%

Seven checks, and across 619 audited pages the total number of failures is small enough to count by hand. Whatever else is true about the SEO plugin market, the configuration problem it was built to solve is solved. Nobody needs another article about installing Yoast.

The three the plugin gives you a box for

Three items on the plugin’s own list behave differently, and the difference is instructive:

Check	Homepages	Posts
Structured data	15.4%	20.9%
Meta description	22.3%	30%
Title tag	19.3%	37.6%

These are the fields the plugin provides an input for and cannot fill in. It can generate a canonical tag because a canonical tag is derivable. It cannot write your title.

The title gap between homepages and posts is the whole story of that distinction. A homepage title gets written once, by whoever set the site up, while looking at the plugin’s preview. Post titles get written three hundred times by whoever was publishing that day, and the plugin’s field sits empty and falls back to the post heading, which was written to look good at the top of an article rather than to work in a result.

That is not a plugin failure. It is the boundary of what automation can reach, and it is the highest value thing on this page that a plugin could ever have helped with.

The checks that do not move between a homepage and a post

Here is the pattern that makes the rest of the data interpretable. Some checks change completely depending on which page you look at, and some barely move at all:

Check	Homepages	Posts	Moves?
Attribution and sourcing	77.1%	8.4%	Collapses
Answer extractability	32.2%	12.9%	Collapses
Text to code ratio	68.4%	39%	Halves
Heading order	56.9%	58.5%	No
Render blocking resources	63%	63.8%	No
Security headers	72.3%	71.8%	No
Caching	53.9%	58.5%	No

A check that changes between two pages on the same site is measuring the content. A check that returns the same figure on both is measuring something the content cannot influence, which on a WordPress site means the theme, the page builder or the server.

This also disposes of a claim we could have made from the homepage run alone, and nearly did. 77% of these homepages name no author, and on posts that figure is 8.4%. WordPress puts the author on the post, the theme prints it, and the check that looks catastrophic on a homepage is a category error: a homepage is not an article and does not have an author. Any audit of WordPress sites that only looks at homepages will report this and several others wrongly.

What is actually left

On the 287 posts, after the plugin has done everything it does:

• Security headers, 71.8%. A server and CDN setting. No plugin claims it and most hosts do not set it by default.
• Render blocking resources, 63.8%. The theme and every plugin that enqueues a stylesheet in the head.
• Heading order, 58.5%. Below.
• Caching headers, 58.5%. The host, or the caching plugin you have not configured.
• llms.txt, 56.1%. Nobody has one. That is the whole industry, not WordPress.
• Page weight, 47.4%. Median HTML document across the homepages was 238KB before a single image, script or stylesheet. The largest was 1.7MB of HTML alone.
• Image formats, 43.6%. Images still shipping as JPEG and PNG where a modern format would be a fraction of the bytes.

Not one of those has a setting in an SEO plugin, and that is not an oversight. They belong to the layer underneath.

Heading order is a theme problem, and it is the cheap one

57% of homepages and 59% of posts had a heading level jump, and it is worth explaining why that number is so high and so stable.

Almost nobody writes an H4 after an H2 on purpose. What happens is that the theme does it for you. A widget title is marked up as an H3 because that is what looked right in the theme’s design. A sidebar block prints an H4. A “related posts” strip prints an H2 after your content has already been to H3. The author writes a clean document and the template interleaves headings from three sources at levels chosen for their appearance.

Google is explicit that heading order does not matter for ranking, and that is accurate. What it does is decide where a page divides into passages, which is what anything quoting rather than linking to your page has to work out first. It is also the single most common defect we find on any sample of professional websites, and it is a template edit rather than a project. Run a post through the heading structure checker and you will usually find the jump is not in your writing at all.

Does the choice of plugin matter?

Median audit score by plugin, across the sites running each:

Plugin	Sites	Median score
Yoast SEO	214	87
Rank Math	92	88
All in One SEO	24	88

One point between them, on a 100 point scale, and the differences that do appear on individual checks are not clean comparisons: sites choose their plugin, so any gap is at least as likely to reflect who chooses what as what the plugin does.

The more useful comparison is against having no plugin at all. Median 88 with one, 86 without. Two points, and again the group without one is weak, because a plugin can be installed with its output suppressed.

Two points is the honest measure of what this decision is worth, and it explains why the plugin comparison genre exists at all: when a choice does not change the outcome, the only thing left to write about is the interface.

What to do with your own WordPress site, in order

1. Keep the plugin. It is doing its job and its job is real.
2. Write the title and description on posts, not just on the homepage. 37.6% of posts were not clean on the title. That is where the plugin was actually going to help and it is waiting for you.
3. Fix heading order in the theme. One template edit, and it is the most common defect on any sample we have ever run.
4. Look at what your host and theme ship. Security headers, caching and render blocking are three checks, one afternoon, and none of them are content work.
5. Then audit a post rather than the homepage. They are different pages with different problems, and the homepage will tell you the wrong thing about half of this.

Run one of your own posts through the audit and compare it against the table above. The checks where you are worse than the sample are yours; the checks where you match it are WordPress.

Sources

• visibility100x, WordPress SEO measurement, 395 sites and 287 posts, collected 20 August 2026. Aggregate CSV. Per-site rows withheld, as described above.
• Google, “SEO Starter Guide”, developers.google.com/search/docs/fundamentals/seo-starter-guide, read 20 August 2026, for the statement on heading order. Discussed at length in our piece on that guide.