This page is about one tool’s data: the E-E-A-T Analyzer. For how these scores sit alongside the other measures, the distributions in full and the methodology, see SEO Score Benchmarks.
Which trust signal is missing, and how often
The tool takes a URL and runs eleven checks: first-hand experience language, original media, a named author, author credentials, Person schema, outbound citations, an About page, a publish date, HTTPS, contact information and reviews. The percentage below is the share of runs in which the check did not pass.
| Check | Did not pass | Weight |
|---|---|---|
| First-hand experience signals | 96% | Hard fail |
| Person / author schema | 91% | Partial |
| Author credentials / qualifications | 85% | Partial |
| Publish / update date | 84% | Partial |
| Reviews / testimonials | 84% | Partial |
| Named author / byline | 76% | Hard fail |
| Contact information | 55% | Partial |
| About page linked | 47% | Partial |
| Cites authoritative sources | 25% | Partial |
| Original media | 19% | Partial |
| Secure (HTTPS) | 0% | Hard fail |
This table does not look like any other table in the series. On the other tools even the most-missed check stays under half of runs; here the first five items sit between 84% and 96%. On a typical page most of these signals are not present-with-gaps, they are simply absent. The average scan passes 4.4 of the eleven checks and leaves 6.6 open items behind it.
The row at the very bottom of the list sums up this page on its own: HTTPS is missing in 0% of scans — that is, in none of them. Which means the one check on this list that everybody passes is also scored as a hard fail. The reason is simple: HTTPS is provided by the host. Nobody provides any of the other ten on your behalf.
The floor of the dataset
The median is 38/100 and 77% of scans graded F — 1,101 runs. 2 scored A+. The distribution agrees: 60% of scans fall below 40 and only 1% climb above 80. That is the lowest median of the 6 measures with a benchmark page — 52 points below Page Speed, the highest.
Those two ends are the two ends of the pattern running through the whole series. On page speed the median is 90/100; here it is 38/100. The distance is not a difference in competence: speed is something the host, the CDN and the framework solved on your behalf. E-E-A-T is the one part nobody can do for you — put a name on the work, say what you did, show the date, cite the source. Because it cannot be automated, it does not get done.
There is a practical side to that. Scoring 80 on this measure is not an achievement, it is an anomaly: 14 scans managed it. And getting ahead in a race nobody has entered is cheap — an author byline, a date and two sentences of credentials close three gaps that are missing from between 76% and 85% of pages.
Two of the four letters carry the whole loss
The tool distributes its eleven checks across the four letters of E-E-A-T: Experience, Expertise, Authoritativeness and Trustworthiness. The weakest-category ranking shows exactly where the loss collects.
Expertise was the lowest-scoring category in 49% of runs and Experience in 49%, leaving the other two at a combined 1%. The loss is almost entirely in the first two letters: who wrote this, and whether that person has actually done the thing. The authority and trust checks — an About page, contact details, HTTPS — are things a site does once while being set up. The first two letters have to be written again on every page.
What this tool is actually measuring
Reading these figures correctly needs one distinction. E-E-A-T is not a score Google computes — it is a concept from a document written for human search quality raters. No tool can measure it directly. What can be measured is the machine-visible evidence of it: does the page name an author, is a qualification stated, is Person schema present, is a date visible, are sources cited. That is what the score here is — not the page’s actual trustworthiness, but how much evidence of it a machine can find.
That distinction does not make the signals unimportant — the opposite. A language model cannot know whether a page is genuinely trustworthy before it cites a passage from it; this evidence is the only thing it has to look at. You will find the same pattern on the AI visibility side, where the first things missing are also the author byline and the publish date. Two separate tools, two separate check lists, the same conclusion.
Analyze a page’s E-E-A-T signals
Frequently asked questions
What is a typical E-E-A-T score?
In this dataset the median is 38/100 and 77% of scans graded F. The 25th percentile is 25 and the 75th is 47; only 1% of scans (14 runs) reached 80 or above. What that means in practice is that a low score is not putting you behind: even 60 here is ahead of the large majority of the sample. Competition on this measure has not started.
Is E-E-A-T a ranking factor?
Not directly, and the distinction matters. E-E-A-T is not a score Google computes; it is a concept from the guidelines that human search quality raters use when judging pages. No tool can measure it directly. What a tool can measure is the machine-visible evidence of it: is there an author name, is a qualification stated, is Person schema in place, is a date visible, are sources cited. That is exactly what the score on this page is — not E-E-A-T itself, but how much evidence of it a machine can find on the page. The distinction matters twice over for generative engines: what a model can rely on when it cites a passage is only the evidence actually written on the page.
Which E-E-A-T signal is missing most often?
First-hand experience signals — absent from 96% of pages. The tool measures it by looking for language in the text that demonstrates first-hand use: "we tested", "I tried", "in our experience". Its absence does not mean the content is bad; it means there is nothing on the page evidencing that whoever wrote it has actually done the thing. That is precisely what the second E added to E-E-A-T in 2022 was meant to capture, and it sits at the very top of this list.
Why do so many pages grade F?
Because both of the heavily weighted checks measure work a human has to do by hand, and both are usually undone: first-hand experience is missing from 96% of pages and a named author from 76%. The average page passes only 4.4 of the eleven checks and leaves 6.6 open items behind it, of which 1.72 are hard fails. Installing a theme passes none of these checks — every one of them is writing that has not been done.
The sample, the methodology, the anonymization and the CC BY 4.0 licence are all set out in the methodology section on the hub.