Real audit, real fixes, real numbers

We ran this on our own fleet first.

Every number on this page is from a real audit and a real fix, verified live in production — not a mockup. This is what "honest audits" actually looks like end to end.

CASE STUDY — SPECTROSCIENCE.COM
Before / After

What the audit actually found.

BEFORE — AUG 21 AUDIT
Stale internal links (pointing at redirects)69
Missing-alt findings flagged63
Analytics pipeline in productionDEAD
Competing sitemap generators3
Pages flagged thin/keyword-dense31
AFTER — VERIFIED LIVE
Stale internal links0
Real gaps fixed (rest were false positives)2 / 63
Analytics pipeline in productionLIVE
Competing sitemap generators1
Pages fixed with real added content13
What we actually did

Real fixes, not a checklist tick.

C1

Analytics wasn't broken — it was never deployed

The "404 on the analytics script" finding looked like a one-line fix. The real cause: the analytics app had never been deployed to production at all, and its port was already claimed by an unrelated internal tool. Deployed the real app, pointed it at the existing (already-populated) database, fixed the nginx routing.

verified 404→200 in production
C2

69 internal links pointing at redirects, not real pages

Rewrote every stale link to point at its real destination, rebuilt, and promoted to production.

verified live, 0 stale links remaining
C7

63 "missing alt text" findings — only 2 were real

Read every single finding instead of bulk-fixing. ~11 were the scanner matching literal code text, ~15 already had correct dynamic alt text a static scanner can't evaluate, ~5 were legitimately empty until JavaScript sets them, ~10 were admin-only and never crawled. Only 2 were genuine public-facing gaps — fixed with real, specific alt text, not keyword stuffing.

honest triage — no blind bulk "fix"
C8/C9

Thin content and keyword density — fixed with real content, not tricks

Short pages read as "keyword-stuffed" purely because they were short — same keyword count, smaller page. Fixed by adding genuine, specific content (real FAQ answers, real explanations) rather than diluting or removing the keyword.

word count up ~40% on the worst offender, density down proportionally
What we didn't do: 2 of the original 9 issue classes were editorial calls (thin content needing a human voice, not a script) — correctly left for founder review rather than auto-generated. One suspected policy issue (a content-generation script possibly using a non-approved AI vendor) was flagged, not silently "fixed" — that's a decision for a human, not an audit tool.

Want this on your own site?

Try the live demo →
Search Engine Seer — Real Case Study Report (PDF)