What was measured
Layer 2 measures publicly observable business clarity and evidence — what a visitor or an automated system can see on the site itself, with no login, no CAPTCHA or paywall bypass, no crawler impersonation, and no private Search Console or analytics data.
Seven internal signals were measured: clarity of the responsible company, clarity of the offer, clarity of operating geography, contact and legal business signals, observable evidence of work or experience, depth of offer pages, and structured entity data. An eighth signal — external confirmation across third-party sources — was deliberately not measured and is held as unknown throughout.
This is a benchmark, not a per-company audit. Every site was checked against the same codebook. No scores were assigned, no companies were ranked, and no “AI readiness score” was produced. It builds directly on the Layer 1 technical eligibility study.
Methodology in brief
The Layer 2 sample is not a new draw. It is the deterministic subset of the Layer 1 sample where all eight technical metrics closed without a no. Of the 1,353 sites in the first study, 1,005 met that condition; the remaining 348 carried at least one technical blocker and were excluded by design — technical eligibility first, clarity second.
For each site a defined public scope was reviewed: homepage, an about/company page, a contact/legal/imprint page, two to three service or product pages, the visible header and footer, public HTML and structured data, and the internal links connecting them. Review depth varied: 600 sites were checked across five or more pages, 130 across four, and 275 across three or fewer. A smaller checked scope is weaker ground for concluding a signal is missing, so the most scope-sensitive finding — work evidence — is reported by depth strata, not as a single number.
Because Layer 2 is interpretive, the codebook was validated first. On a 50-site calibration subset, two independent reviewers coded every signal; across 350 rows they disagreed on 13 — 96.29% agreement — with the final value taking the more conservative reading where they diverged. The full pass is a calibrated automated internal-signal measurement, not a manual audit: of 7,035 measured cells, ~9% carry a low-confidence flag, two-thirds of them on structured data.
Key findings
Of the 1,005 technically clean sites, only nine were clearly expressed across all seven clarity signals at once — 0.9%. Those nine are the sites where nothing is left to guess: a stranger, person or model, could read the whole business off the page.
The contrast underneath is plain. The contact layer is almost universal — 84.98% of sites clearly expose contact or legal business signals. The evidence layer is not: even among the most thoroughly checked sites (five or more pages), roughly one in three (32.8%) show no visible proof of work at all.
Most sites land in the middle — clear on two to four signals, partial on the rest; fewer than 5% are clear on six or seven. Reduced to a single scale (clear=1, partial=0.5, absent=0), the sample’s mean clarity score is 0.610, ranging from contact (0.93) down to work evidence (0.33). This describes the sample as a whole — it names, scores and ranks no individual company or country.
How to read the seven signals
Each signal carries one of four values: clearly visible (clear, specific and sufficient in scope), partially visible (present but incomplete, generic or ambiguous), not observed in scope (not found on the checked pages — not proof it is missing in reality), and could not be checked (access, render or fetch limits).
Responsible-company clarity — is it clear which company, brand, legal entity, group or branch runs the site? Offer clarity — which HVACR products, services or categories are offered? Geography clarity — where the company operates. Contact and legal signals — a contact route plus a business/legal signal.
Work / experience evidence — are company-specific proof artifacts visible? Offer-page depth — do offer pages explain the offer specifically? Structured entity data — is useful machine-readable entity/offer data present? “Not observed” is never proof a signal is missing in reality — only that it was not found within the checked scope.
Reading the results
Contact is a solved problem; trust is not. Almost every site offers a way to get in touch, but the ability to be contacted and the ability to convince a reader of real experience are different layers — and the second is systematically weaker.
Work evidence is one of the two weakest areas. Among sites checked across five or more pages, ~32.8% show no project, case study, captioned work photo, certificate or named client anywhere in scope. The rate falls in a clean gradient as depth rises — 81.1% (≤3 pages) → 60.8% (4) → 32.8% (≥5) — so we treat the deep-scope ~33% as primary and the full-sample 49.65% as an upper bound. Structured data looks thin (absent on 30.55%) but is our least confident signal, reported as a soft indication.
The offer is understandable but generic — ~50% name HVACR categories clearly, ~37% only partly. Company and geography are blurred, not absent: almost never entirely missing (0.30% and 4.28%) but clearly shown only on a minority (36.32% and 16.72%). Country-level differences are descriptive, not a ranking; controlling for review depth, work-evidence-absent ranges from ~24% (Italy) and ~26% (Germany) to ~45% (France) and ~49% (Spain), always read next to per-country sample size.
Beyond Layer 2: external confirmation readiness
This benchmark stops at what the site says about itself. It does not measure whether the same business entity is confirmed, without material conflict, across external public sources — Google Business Profile, LinkedIn, industry associations, exhibition profiles, business directories or official registries.
That question belongs to a separate confirmation layer with its own methodology, source list and reliability checks. It is held here as unknown and is the natural subject of a third study.
Entity-clarity hygiene implications
These are not SEO recommendations or a promise of visibility. They are checks that follow directly from the measured clarity gaps, in order of likely return:
→ Show proof of work — projects, case studies, captioned photos, certificates, partnerships, client examples.
→ Make the responsible company unambiguous — the brand, legal entity, group or branch, not just a logo.
→ State the operating geography — not only an address, but the markets or service area.
→ Make offer pages specific instead of generically promotional.
→ Add useful structured data about the company, services and contacts where appropriate.
None of this guarantees rankings or AI citations. What it does is reduce ambiguity — and a clearer site is easier for people, search engines and AI systems to read. IndexDock’s role is to help industrial companies find and close publicly observable clarity and evidence gaps before ranking, traffic or AI-citation outcomes are assessed.
What this study does not prove, and limitations
The study does not prove rankings or changes to them, traffic impact, actual indexing status, actual AI citations or recommendations, business quality, or company trustworthiness. It does not establish that one company or country is better than another, and it does not prove that a signal absent from a site is absent from the real company.
What it shows is narrower and defensible: within the measured sample, publicly observable gaps in entity clarity and evidence exist on Layer-1-clean EU HVACR websites. Layer 2 carries higher measurement uncertainty than Layer 1 because several signals require interpretation (~9% of cells low-confidence). The work-evidence absent rate is scope-sensitive (32.8% deep scope vs 49.65% full sample; deep figure treated as primary). Results are time-bound, and external entity confirmation was not measured in this pass.
Overall results across the sample
Observed “no” results by metric. HTTP 200 success is shown for completeness but treated as a sample quality gate, not a market finding. Language routing is measured against its applicable multilingual subset of 365 sites.
Of the 1,005 technically clean sites, how many of the seven clarity signals are clearly expressed. Most land in the middle; fewer than 5% are clear on six or seven.