Public verification example
The audit found the gap. The site now shows the repair.
Carney.wiki is the first public before-and-after example for VerifyHub. The point is not a flattering score. It is a traceable record of what the crawler found, what changed in the source, and what the same rules measured afterward.
Before
24 / 50 measured points
After
48 / 50 measured points
The measured repair
On carney.wiki, the Hugo head-template repair cut detected missing descriptions 100% from September 1 to 2, 2026, from 143 in the capped baseline to 0 in the full crawl.
The two crawls did not use identical page counts, so the result describes the audit record. It does not prove conversion lift, revenue impact, delivery quality, or customer outcomes. The baseline and after report remain public so the claim can be inspected in context.
Metadata
Repair the system once.
The Hugo head template now uses each page's description or excerpt, then a page-type fallback. Every indexable page also receives a canonical URL and valid JSON-LD.
Discovery
Make similar pages distinct.
Tag, category, and series titles now identify their page type. Their fallback descriptions do the same, so a shared topic does not create indistinguishable search metadata.
Buyer validation
Answer the questions that slow a decision.
The site now publishes trust practices, implementation steps, offer comparison, pricing, and frequently asked questions as direct routes.
Structured evidence
Give machines explicit context.
Articles identify themselves as BlogPosting objects. Lists use CollectionPage. The FAQ publishes FAQPage, Question, and Answer objects.
Content defect
Fix the wrong excerpt, not just the empty ones.
One older automation article had inherited an unrelated conference-scandal excerpt. The source now describes the article it belongs to.
Navigation
Evidence only helps when people can reach it.
The primary navigation now links to this verification record. Trust, pricing, and FAQ routes are linked from the footer across the site.
What the score means
The reported scores use VerifyHub rubric vh-1.0.0. Both runs used deterministic checks only because the assisted-analysis key was unavailable. Positioning, Proof Ops, and credibility remained unscored instead of being guessed.
The after result is the highest deterministic total available under this rubric. Its path-friction rule tops out at 4 of 5, so the measured ceiling is 48 of 50. That is a property of the published rubric, not a claim that the site is perfect.
What the comparison does not prove
The baseline and after runs did not have identical coverage. The baseline stopped at 150 sitemap URLs. The after run covered all 339 URLs. The metadata repair was applied in the shared template and separately checked across every non-redirect HTML page in the build.
The crawler checked published structure, links, metadata, route coverage, and response timing. It did not submit forms, measure conversion, prove delivery quality, or establish customer outcomes.
A verification report should be inspectable.
The summary is useful. The raw reports are the evidence. Read the story behind the repair, open the baseline report, inspect the after report, or review the site's trust and verification practices.