Comparison Report: Water Leak Insurance Claim
The original live page on bluebot.com was already a strong, publish-quality guide: it scored 82% standalone. The pipeline rewrite carried that fidelity forward and closed the AI-search gaps the original left open, landing at 98/100. This is where the roughly +16 normalized-point lift came from, and which AI queries the page now wins.
Composite Score: Before vs. After
The original was not a weak page. It opened with a definition, used question-style headings, carried dense entity clusters and a full set of statistics, and read cleanly for a YMYL-financial audience: it was already AI-platform-ready on every engine. The pipeline’s job here was narrower than a typical rescue: verify and lock the coverage the original had, add the machine-readable layer it was missing entirely, and turn every figure into an extractable fact. The biggest single gain, Schema & SEO, 0 → 10, came simply because the original page shipped no JSON-LD at all.
Layer-by-Layer Comparison
Two notes on the math. The “before” is a standalone score against 66 available points: Schema was SKIPPED (0 available, no JSON-LD to read), and Entity, GEO/AEO and AI-Citation were capped below their full-mode maximums because no Entity Map, Schema, or Research Blueprint was supplied. The “after” is scored against the full 100-point how-to rubric under the comprehensive length tier. The bars below show each layer as a percentage of its own maximum so before and after are visually comparable.
| Layer | Before | After | Delta | What changed |
|---|---|---|---|---|
| 1: Content Structure | 12 / 15 | 15 / 15 | +3 | 7 ordered steps + “What Is” + Common Mistakes + Conclusion + 6-Q FAQ; scored comprehensive tier (no length-overshoot penalty) |
| 2: Readability | 8 / 10 | 8 / 10 | even | Held strong (FK ~7–8, appropriate for YMYL financial); dense insurance-jargon sentences split where natural |
| 3: Entity Optimization | 6 / 8 | 15 / 15 | +9 | Verified surface-form coverage; 11-entity 3-layer map (3 PRIMARY / 6 SECONDARY / 2 AUTHORITY); density in band; author + 3 products added |
| 4: GEO/AEO Compliance | 8 / 10 | 15 / 15 | +7 | Definition-first lead; every statistic turned into an extractable EAV / quotable fact; 14–16 SPO triples |
| 5: Schema & SEO | — / 0 (skipped) | 10 / 10 | +10 new | Biggest single gain. Original had NO JSON-LD. Added single @graph: HowTo root + step array, author Person, publisher, SpeakableSpecification, 3 Product, FAQPage × 6 |
| 6: E-E-A-T Compliance | 5 / 6 | 10 / 10 | +5 | Named author Matthew Olin, Founder & CEO restored; first-party data; credential context |
| 7: Claim-Safety Compliance | 9 / 10 | 10 / 10 | +1 | All 6 sourced facts + caveats + non-advice disclaimer; product claims hedged (“many insurers offer”, “confirm eligibility”) |
| 8: AI Citation Readiness | 6 / 7 | 15 / 15 | +8 | Speakable + confirmed @type + recency marker; all 9 recovered statistics source-attributed and extractable |
| TOTAL (normalized) | 82% (54/66) | 98% (98/100) | +16 | PASS GOOD → PASS EXCELLENT |
Where the Lift Came From
Four layers account for nearly all of the gain. Note that the “before” was a genuinely strong baseline, so the lift is less about fixing defects and more about unlocking capped layers and adding the machine-readable surface the original never had.
Schema & SEO
The original live page shipped no JSON-LD whatsoever, so the layer was SKIPPED (0 available). The pipeline added a single @graph rooted on HowTo with a one-to-one step array, an author Person (Matthew Olin), the Bluebot publisher, a SpeakableSpecification, three Product mentions (Bluebot WiFi Smart Water Meter, Mini, EcoLink), and a FAQPage with 6 Question nodes, plus a 56-char meta title and 159-char description. This is the difference between a page an engine can read and one it can parse with confidence.
Entity Optimization
Standalone, the entity layer was capped at 6/8 because there was no Entity Map to verify against. The pipeline supplied the 11-entity, 3-layer map (3 PRIMARY: water-leak insurance claim, damage documentation / proof of loss, homeowners insurance policy; 6 SECONDARY; 2 AUTHORITY: State Department of Insurance, insurer claim guidance) with verified surface-form rotation, traceability 0.93, and density landing in band. Adding the named author and the three Bluebot products thickened the entity set without diluting salience.
GEO/AEO Compliance
The original led with a definition and carried the right facts, but they sat as prose. The rewrite kept the definition-first lead and converted every statistic into an extractable EAV / quotable fact (the $13B/yr claim volume, the $11,000 average payout, the 5–10% premium discount, the 43,200 data-points/day Bluebot figure) and expanded the narrative SPO set to 14–16 triples so each answer pattern an AI engine looks for is present and machine-clean.
AI-Citation Readiness
With schema in place, AI-citation signals could finally be confirmed rather than inferred: a SpeakableSpecification for voice surfaces, a confirmed @type graph, an explicit recency marker (“Last updated June 2026”), and all 9/9 recovered statistics source-attributed inline. Every platform (Google AI Overview, ChatGPT, Perplexity, Claude) is rated Ready.
Source fidelity restored. An earlier over-compressed draft (1,074 words) had dropped the named author, every statistic, and the entire Bluebot product layer. This comprehensive rebuild (~3,720 words, ratio 0.86 to the ~4,300-word source) brought back 9/9 statistics, 13/13 internal links, the named author Matthew Olin, and the Bluebot product sections, while keeping every AEO/GEO and compliance gain above.
What This Means for AI Search
The original could be cited for the page’s primary query: “how to file a water leak insurance claim.” But its facts lived in prose and it carried no schema, so for the high-intent sub-questions a homeowner actually asks an AI assistant, there was nothing structured to lift. The rewrite’s definition-first lead, 6-question FAQ, and JSON-LD now surface the page across a much wider query set:
| AI query class | Before | After |
|---|---|---|
| how to file a water leak insurance claim (primary) | ~ cited from prose, no schema anchor | ✓ HowTo step array + speakable + definition lead |
| does homeowners insurance cover water leaks | ✗ answer buried in prose, not extractable | ✓ FAQ Question node + EAV fact + definition-first lead |
| sudden vs gradual water damage (coverage) | ~ comparison present, no structured pairing | ✓ explicit sudden-vs-gradual framing + quotable fact pair |
| how long do I have to file a water leak claim | ✗ no FAQ pair to lift | ✓ FAQ Question node + direct-answer pattern |
| average water damage insurance payout / claim cost | ~ $11k figure in prose, uncited | ✓ $11,000 + $13B as source-attributed EAV facts |
| what is proof of loss / how to document water damage | ~ checklist as bullet list | ✓ PRIMARY entity + step depth + extractable triples |
| do I need a public adjuster for a water claim | ✗ 10–15% fee figure in prose only | ✓ quotable fact + Common Mistakes section context |
| smart water leak detector for insurance discount | ~ Bluebot mention, no Product schema | ✓ 3 Product nodes + 5–10% discount EAV fact (hedged) |
Outstanding Items (non-blocking)
- Readability (+1–2): a few compound sentences in Steps 5–6 still run ~22–25 words. Optional split toward FK ≤ 7, acceptable as-is for YMYL regulatory precision.
- Inline source attribution: the $13B, $11k and 5–10% figures are extractable but would be stronger with named third-party sources (III / NFIP) alongside the first-party Bluebot data.
- Legal review: compliance scanning is pattern-based; product / discount statements are hedged and the non-advice disclaimer is present, but a legal pass is recommended before publishing YMYL financial content.
Generated by the content-comparison-report agent (Phase 7) for the “How to Make a Successful Water Leak Insurance Claim” universal-content-pipeline run. Before = standalone pre-pipeline validation (05-validation-scorecard-pre-pipeline.html, 54/66 = 82%, PASS GOOD). After = post-pipeline validation (05-validation-report.md, 98/100 = 98%, PASS EXCELLENT, comprehensive length tier). Compliance scanning uses pattern detection; no violations detected does not constitute legal clearance.