CORPUSScores
What an SEO score means, and what it can't tell you.
On September 5 our own homepage scored 98 out of 100 on our own audit. On September 10 the same audit gave the same site 91, with the SEO pillar down from 100 to 75. Nothing about the business changed. We rewrote the homepage copy, and three deterministic checks started failing.
Every number on this page is our own measurement of our own site, taken with the audit that runs for customers. Outside facts carry a numbered source, checked 2026-09-10. Corpus Plan does not promise rankings, traffic, revenue, or AI mentions.
No account or email needed for the first report. Pro is $99/mo for weekly monitoring and up to 10 sites.
What actually broke
Three findings fired, and all three are about length:
SEO_TITLE_LENGTH. Our homepage title had grown to 68 characters. The audit's published band is 15 to 65.
SEO_META_DESCRIPTION_LENGTH. Our homepage meta description had grown to 187 characters. The published band is 50 to 165.
SEO_PAGES_TITLE_LENGTH. An inner page, our immigration citation study, carried a 77-character title.
That is the whole 25-point drop in the SEO pillar. We had rewritten the homepage a few days earlier to describe the portal properly, the new sentences were longer than the old ones, and nobody counted characters. The audit counted them.
Every threshold above is published in advance at /methodology, with the code, the severity, and the exact condition that makes it fire. That is the part a score is good for: it is a fast, repeatable read of whether a page meets a list of rules you can look up.
What a score is
A site score is a summary of deterministic checks against thresholds somebody chose. Run the same page twice with no changes and you get the same number. Change the page and the number moves by an amount you can trace to a specific check.
That makes a score useful for three things: spotting a regression you did not intend, ranking your own to-do list, and confirming a fix landed. It is a smoke alarm and a checklist. Our own 25-point drop is a good example of the first: the copy change was deliberate, the metadata side effect was not, and the score is how we found out.
What a score is not
It is not a prediction of how you will rank, and no site score is. Google's own starter guide says plainly that "there's no guarantee that any particular site will be added to Google's index," and that sites following its Search Essentials are only "more likely to show up."[1] Google also says changes can take anywhere from a few hours to several months to show any effect, and that not every change produces a noticeable one at all.[2]
The tools that produce scores say similar things about themselves. Google's Lighthouse documentation notes that a performance score moves for reasons that have nothing to do with the page, including A/B tests, ad serving, traffic routing, the device used for the test, browser extensions, and antivirus software, and suggests thinking of performance as "a distribution of scores, rather than a single number."[3] The same page explains that the scoring curve is fitted to real-world data from the HTTP Archive, where the 25th percentile is mapped to a score of 50 and the 8th percentile to 90, and that a perfect 100 is "extremely challenging to achieve and not expected."[3]
In other words, the number is a position against a rulebook, not a forecast.
The uncomfortable version, from our own data
Our site scored in the nineties throughout. On 2026-09-09, Google Search Console reported 0 indexed and 16 not indexed pages for plan.corpushq.co, every one of them under the single reason "Crawled - currently not indexed." Google fetched the pages and chose not to index them.
Nothing on the page fixed that, because nothing on the page was broken. Titles were present, canonicals were set, the sitemap was submitted and fetched, the pages were indexable. A page-level score cannot see the things that decide indexing and ranking: whether anyone links to you, whether the site has any history, whether the page adds something the index does not already have.
We publish this because it is the honest reading of our own numbers, and because it is the thing most score reports leave out. A high score means the checkable work is done. It does not mean the job is done.
How to read any site score
Ask what the thresholds are. If a tool will not tell you what makes a check pass or fail, the score is an opinion with a number attached. Ours are at /methodology.
Read the findings, not the number. The number is a sort order. The findings are the work.
Watch the direction, not the absolute value. Two tools will score the same page differently because they weight different checks. Comparing your own score across weeks tells you more than comparing it against someone else's tool.
Do not buy a promise attached to a score. Nobody can sell you a ranking or an AI citation, whatever the dashboard says.
What we changed
We shortened the homepage title and meta description back inside the bands, and shortened the study page title. The copy still says what the portal does; it says it in fewer characters. The fix took about ten minutes, which is the usual size of this kind of work once something tells you where to look.
Check yours
The first look is free and needs no account or email address. You get the five pillar scores, every finding with the reason written out, and the thresholds behind each one.
Sources
References
- https://developers.google.com/search/docs/fundamentals/seo-starter-guide - accessed 2026-09-10. Google: "there's no guarantee that any particular site will be added to Google's index"; sites following the Search Essentials are "more likely to show up in Google's search results".
- https://developers.google.com/search/docs/fundamentals/seo-starter-guide - accessed 2026-09-10. Same guide, "How long until I see impact in search results?": some changes take a few hours and others several months, and not all changes produce a noticeable impact in search results.
- https://developer.chrome.com/docs/lighthouse/performance/performance-scoring - accessed 2026-09-10. Lighthouse performance scoring: the score fluctuates for reasons outside the page (A/B tests, ad serving, routing, test device, extensions, antivirus); performance is better read as "a distribution of scores, rather than a single number"; the curve is fitted to HTTP Archive data with the 25th percentile at 50 and the 8th at 90; a perfect 100 is "extremely challenging to achieve and not expected".
Spotted an outdated source or a claim that needs correction? Contact Corpus Plan and we will re-check it.