{ case study 02 }
SEO Decay Audit
Classic SEO decay is measured in clicks. The new one is measured in citations: the page still ranks, but the AI answer consumed the click. Sixteen pages across four major content sites, scored on both axes.
{ the build }
Still ranking.
Still losing.
Problem
Every SEO tool on the market measures decay the same way: rankings slipped, so traffic slipped. That model assumes the only way to lose a click is to lose the position.
It no longer holds. A page can sit at position three, hold it all year, and still bleed traffic, because the AI answer above it already gave the reader what they came for. The ranking is intact and the click is gone. Measured on the old axis, that page looks perfectly healthy right up until the quarter it isn't.
Approach
Score both axes at once. Classic decay on one, citation decay on the other, and let each page land where the two put it, which sorts every page into one of four verdicts, where a single health score would have averaged the two axes into something misleading.
The discipline that makes it publishable: every number is labelled with where it came from. Measured, estimated and inferred figures are marked as such on the page itself, so a reader can tell which conclusions rest on hard data and which rest on a reasonable assumption. An audit that hides its own provenance is just an opinion with a chart on it.
{ how it runs }
Four sites,
two axes.
A fixed sample, scored the same way every time, then published where anyone can check the working.
The sample
Sixteen pages drawn from four major content sites, a fixed sample named in the audit itself. Small enough to verify by hand, wide enough that the pattern isn't one publisher's bad quarter.
16 pages 4 sites Fixed sample
Classic axis
The familiar measurement: position and click trajectory over time. This is the axis every existing tool already reports, included so the new one can be read against a baseline people trust.
Rank data Click trajectory DataForSEO
Citation axis
Sixty-four prompts, four per page, put to AI answer engines to see which pages get cited, which get absorbed uncredited, and which never surface at all. That gap between ranking well and being cited is the whole finding.
64 prompts Citation scoring AI visibility
Verdict & publish
Each page lands in one of four verdicts based on where the two axes put it, with the provenance of every figure labelled beside it. Built in Astro and served as static files, which is why the hosting bill is zero and stays zero.
Astro Four verdicts Netlify
{ what it produces }
Scored,
and showing it.
{ more work }