AI Visibility Index Q2/2026: How Visible Are German Companies to ChatGPT, Perplexity, and Gemini
We audited 450 German websites across 20+ industries with 16 automated checks against 9 AI engines. The result: 97% score below 40 on the AI Visibility Index — our threshold for "AI-visible." This report presents the aggregated data — anonymized, but painfully honest.
What Is the AI Visibility Index?
The AI Visibility Index is TrueSource's quarterly benchmark that measures how visible German websites are to AI search engines. Not: How well does a website rank on Google. But rather: Is it found, cited, or recommended by ChatGPT, Gemini, and similar AI engines?
For Q2/2026, we selected 450 websites from over 20 industries — from DAX-listed corporations and mid-market companies to federal agencies. Each website undergoes 16+ automated checks and is tested against 9 AI engines. The result is a score from 0 to 100.
The Index is not a ranking. There are no names, no domains, no identifiable individual findings. What we publish is aggregated industry data — averages, distributions, recurring patterns. The companies remain anonymous. The findings do not.
Methodology: Scoring V3.4
Each website is evaluated across five categories:
- Technical AI Readiness — llms.txt, robots.txt configuration for AI crawlers, sitemap accessibility
- Structured Data — Schema.org coverage (Organization, Article, FAQPage, Product, BreadcrumbList), completeness on subpages
- Content Signals — E-E-A-T indicators, visible authorship, source references, content depth
- AI Engine Visibility — Actual discoverability across the 9 AI engines via query sampling
- Crawler Access — Which AI bots are explicitly allowed or blocked
The 9 AI engines in the audit scoring: ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok, Copilot, Cohere, and AI Overviews.
Threshold: We define "AI-visible" as a score ≥ 40 out of 100. Websites below this level lack the technical foundations to be reliably found, cited, or recommended by AI search engines.
97% of the 450 audited websites score below 40. The average is 19 out of 100 points. This is not an edge case — this is the status quo.
Industry Comparison
Of the 450 audited websites, 287 fall into six clearly defined industry clusters. The remaining 163 are distributed across additional industries with smaller sample sizes and are not reported individually here.
| Industry | n | Ø Score (/100) | llms.txt present (%) | Schema coverage (%) | FAQPage present (%) | GPTBot blocked (%) |
|---|---|---|---|---|---|---|
| SaaS / Tech | 30 | 31 | 15% | 52% | 20% | 10% |
| E-Commerce / FMCG | 65 | 24 | 8% | 41% | 14% | 28% |
| Automotive | 42 | 21 | 2% | 38% | 8% | 40% |
| Insurance | 87 | 18 | 4% | 23% | 9% | 34% |
| Agencies & Consulting | 38 | 15 | 5% | 12% | 6% | 18% |
| Public Sector | 25 | 11 | 0% | 8% | 4% | 12% |
The range is considerable: SaaS/Tech companies average 31 points — still below the visibility threshold of 40, but significantly above the overall average. At the other end, public sector organizations average just 11 points. None of the six industries reaches the AI visibility threshold as a group.
Notably: The automotive industry blocks GPTBot most frequently (40% of the sample). In the insurance sample (n=87), roughly one third block GPTBot via robots.txt — partly a deliberate decision for data protection reasons, partly an inherited CMS configuration that was never questioned.
The 5 Most Common Mistakes
96% have no llms.txt
The llms.txt file is the equivalent of a robots.txt for AI models — a machine-readable table of contents that tells AI engines which content is relevant. Of 450 audited websites, only 18 have an llms.txt deployed. The rest exists as unstructured text mass for AI crawlers.
78% only use schemas on the homepage
Many websites have Schema.org markup on their homepage — Organization, WebSite, sometimes BreadcrumbList. But on the subpages where the actual expert content lives, schemas are silent. Product, article, and FAQ pages without structured data are difficult for AI engines to classify. The homepage alone is not enough.
Approximately one third block GPTBot via robots.txt
Across the overall sample, roughly 30% of websites block GPTBot, Google-Extended, or other AI crawlers. In some industries — Automotive (40%), Insurance (34%) — the figure is even higher. In some cases this is a deliberate decision, in others it's a CMS default or an inherited configuration from the SEO team that was never questioned.
Only 12% use FAQPage schema
FAQPage schema is one of the strongest levers for AI citations. When a website outputs structured questions and answers, AI engines can embed them directly in their responses — with source attribution. Yet only 54 of the 450 audited websites use this schema on at least one page. An enormous untapped lever, especially for insurance companies, public sector organizations, and consulting firms.
87% have no visible author on expert articles
E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness — is not just a Google concept. AI models also weigh author signals. Who wrote this article? With what qualifications? In our sample, 87% of websites have no visible author with credentials on their expert articles. No name, no title, no link to a profile. For AI engines, that's anonymous content — and anonymous content is rarely cited.
What the Top 3% Do Differently
14 of the 450 audited websites achieve a score of 70 or higher. They span various industries and company sizes. What they have in common:
- Schema.org on every page — not just Organization on the homepage, but Article, FAQPage, Product, BreadcrumbList, and HowTo on the respective content pages. An average of 5+ schema types per domain.
- llms.txt deployed — with clear structure: company description, core products, links to the most important expert pages. Machine-readable, up-to-date, maintained.
- FAQPage on content pages — not just on a FAQ overview page, but embedded in product pages, guides, and expert articles. Everywhere users ask questions.
- Visible authors with credentials — every expert article has a named author with job title, photo, and link to their LinkedIn profile or author biography. E-E-A-T, machine-verifiable.
- Structured data beyond Organization — LocalBusiness, ProfessionalService, MedicalOrganization, EducationalOrganization — depending on the industry. The top websites use the most specific schema types, not the most generic ones.
The top 3% haven't found an "AI trick." They do what structured content should have always been — consistently, on every page, machine-readable. The difference is not in technology, but in discipline.
Conclusion
The AI Visibility Index Q2/2026 paints a clear picture: German companies are largely invisible to AI search engines. Not because the technology is missing — llms.txt, Schema.org, and FAQPage exist as open standards. But because implementation is lacking.
97% below the visibility threshold. Average score 19 out of 100. No industry above 40. This is not a niche problem — it's a structural blind spot.
The good news: The levers are known. Deploy llms.txt, roll out schemas on subpages, use FAQPage markup, make authors visible, don't blanket-block AI crawlers. These are not experimental measures — these are the basics that 97% of audited websites have yet to implement.
The AI Visibility Index is published quarterly. Q3/2026 will show whether the numbers are moving — or whether the blind spot is growing while AI search engines continue gaining market share.
Where does your company stand?
Find out in 60 seconds.
Free AI Visibility Scan — no registration required.
Sascha Deforth is the founder of TrueSource AI and GEO practitioner focused on AI Visibility Infrastructure. He developed the AI Visibility Index to make the gap between SEO optimization and AI visibility measurable. LinkedIn profile →