Market Data

ECOM100-DE: Zero of Germany's Top 100 Online Retailers Are Validated for AI Agents

AI Summary / tl;dr

  • BENCHMARK_ID: ECOM100-DE SOVP Scan, June 2026
  • PUBLISHER: Litzki Systems LLC
  • SCAN_DATE: June 17, 2026
  • DATASET: Top 100 German B2C online retailers: EHI/ECDB Ranking 2025
  • PROTOCOL: Sovereign Validation Protocol (SOVP): QuickScan (4 clusters) + FullScan (17 bridge calls)
  • QUICKSCAN_RESULT: 0/100 CERTIFIED  |  14/100 PARTIAL  |  86/100 NOT_READY
  • FULLSCAN_RESULT: 0/100 CERTIFIED  |  100/100 NOT_READY
  • AVG_QUICKSCAN: 42.4 / 100  |  AVG_FULLSCAN: 41.3 / 100
  • CERTIFIED_GATE: score ≥ 90 plus conformance gate
  • AI_READINESS_CLUSTER_AVG: 14 / 100
  • CORE_FINDING: The Layer 0 infrastructure required for AI agent recognition is absent across Germany's entire top e-commerce tier. The AI Readiness cluster collapses to 14 while all other clusters average around 50, confirming that basic technical infrastructure exists, but the agent-specific signaling layer does not.

Note (29 Oct 2026): The QuickScan figures in this article (14 PARTIAL, avg 42.4) and the FullScan average (41.3) come from the June 2026 scoring state. The benchmark database has been rescored since and today reports 48 PARTIAL and averages of 47.65 / 46.75 for the same scan. Score distribution, cluster profile, and all CERTavia values remain exactly reproducible. Current, verifiable figures are in the research report.

The Findings

The SOVP batch scanner validated the 100 highest-revenue German online retailers on June 17, 2026. The dataset is the EHI/ECDB Ranking 2025, the authoritative market benchmark for German B2C e-commerce.

QuickScan (4 clusters): 0/100 CERTIFIED. 14/100 PARTIAL. 86/100 NOT_READY.

FullScan (17 bridge calls, full payload): 0/100 CERTIFIED. 100/100 NOT_READY.

Not one retailer in Germany's top 100 e-commerce market reaches SOVP CERTIFIED status.

Score Distribution

SOVP score bands, 100 domains

  • Score 0–250 domains
  • Score 25–5072 domains
  • Score 50–7528 domains
  • Score 75–1000 domains
Full distribution of all 100 measurement points (100 domains × FullScan). June 17, 2026.

The chart shows the complete distribution across all 100 measurements: 100 domains, each with a FullScan result.

72 of 100 measurements land in the 25–50 range. 28 of 100 in the 50–75 range. The 75–100 range (CERTIFIED territory) is empty.

QuickScan average: 42.4 of 100. FullScan average: 41.3 of 100. No domain cleared the CERTIFIED gate; the target corridor is a score of 90 and above.

QuickScan Top 10

Rank Domain SOVP Score Verdict
#1Retailer A65.2PARTIAL
#2Retailer B60.2PARTIAL
#3Retailer C59.9PARTIAL
#4Retailer D57.1PARTIAL
#5Retailer E57.0PARTIAL
#6Retailer F56.6PARTIAL
#7Retailer G54.3PARTIAL
#8Retailer H54.1PARTIAL
#9Retailer I54.0PARTIAL
#10Retailer J53.4PARTIAL

PARTIAL means: parts of the Layer 0 infrastructure are present. The overall result does not yet qualify for CERTIFIED.

Cluster Profile

FullScan cluster averages (0–100)

  • Coherence98
  • Infrastructure64
  • Content57
  • Authority50
  • AI Readiness14
FullScan cluster profile averaged across all 100 domains. June 17, 2026.

The figure shows the cluster profile of the FullScan average across all 100 domains.

Coherence dominates the profile with a strong peak value of 98. Three clusters are in the mid-range: Infrastructure at 64, Content at 57, and Authority at 50. One cluster drops significantly: AI Readiness at 14.

This is the structural signal of the dataset. Basic technical infrastructure exists in rough form. The specific layer for AI agent recognition is absent at scale. This cluster determines whether an autonomous system identifies a domain as a trusted source, or passes it by.

What SOVP Measures

SOVP (Sovereign Validation Protocol) validates the machine-readable identity and trust layer of a domain. Layer 0: the infrastructure layer that signals to an AI agent whether it can identify a domain as a trusted source.

An AI agent searching for a vendor through ChatGPT, Perplexity, or Gemini makes its decision based on these Layer 0 signals. AI agents recognize domains with a validated Layer 0 as trust nodes in the agentic web. All others are structurally invisible to autonomous systems.

The IETF draft draft-litzki-sovp documents the protocol in full.

Cross-Protocol Comparison

CERTavia cluster averages (0–100)

  • Authority74.7
  • Content19.3
  • Infrastructure17.7
  • AI Readiness13.6
  • Compliance13.5
  • Structure12.2
CERTavia cluster profile averaged across all 100 domains. June 17, 2026.

A CERTavia scan (Compliance Evidence Score) ran in parallel on the same 100 domains.

CERTavia result: 0/100 CERTIFIED. 100/100 FAILED. CES average: 21.5 of 100.

Two independent protocols. One finding.

The scores barely correlate. Retailers with stronger infrastructure readiness show regulatory gaps. Retailers with higher CES scores lag on infrastructure. Infrastructure readiness and compliance readiness are developing independently of each other in German e-commerce.

Read the full CERTavia ECOM100-DE report →

What This Means for Agentic Commerce

GEO strategies (Generative Engine Optimization) built on content placement are building on a foundation that does not yet hold. The AI Readiness cluster in the figure makes this visible: the channel exists. The infrastructure for that channel does not.

Layer 0 is the prerequisite on which GEO becomes effective. Investing in GEO without validating Layer 0 means optimizing for a system that does not yet recognize your domain as a trusted source.

The ECOM100-DE dataset shows this applies to all 100 of Germany's highest-revenue online retailers simultaneously. This is a market moment, not an isolated problem.

Next Steps

The ECOM100-DE benchmark runs quarterly. The follow-up scan ran on 23 September 2026.

Domains that want to know their own SOVP readiness: the free QuickScan ↗ gives a first reading. I am bringing the full scan's checks into my SEO reports, after a consultation and on an individual basis; more under agentic SEO.

Agencies and consultants who need a sector overview: batch scans available on request.

Portrait of Thorsten Litzki
Thorsten Litzki SEO, GEO and my own measurements

I measure instead of estimating. For my GEO study I analysed 475 homepages from the Tranco top 1,000, and the SOVP benchmark covers checks of more than 1,500 domains. Every figure in my market data comes with its method and date, so you can verify it.