# TrueSource AI

> GEO-Agentur und Execution-Plattform für KI-Sichtbarkeit. Spezialisiert auf Generative Engine Optimization (GEO), mit eigenen offenen Spezifikationen und deployment-fertigem Code. Grevenbroich bei Düsseldorf, Deutschland.

## What is TrueSource AI?

TrueSource AI is a specialized GEO agency (Generative Engine Optimization) and execution platform for AI visibility based in Grevenbroich near Düsseldorf, Germany. Unlike monitoring tools that only track how brands appear in AI-generated answers, TrueSource generates the actual code, schemas, middleware, and compliance files needed to make brands visible and recommended by AI engines.

## What is Generative Engine Optimization (GEO)?

Generative Engine Optimization (GEO) is the practice of making websites visible, citable, and recommendable by AI search engines like ChatGPT, Google Gemini, Perplexity, Claude, and DeepSeek. GEO is to AI search what SEO was to traditional search — the systematic optimization of digital presence for machine understanding.

Key differences between GEO and SEO:
- SEO optimizes for Google's search results page (10 blue links)
- GEO optimizes for AI-generated conversational answers and citations
- SEO focuses on keywords and backlinks
- GEO focuses on structured data, entity grounding, and semantic context
- Measured impact (Ahrefs, Dec 2025 data, 300,000 keywords): where an AI Overview appears, click-through rate for position 1 drops by 58%
- Germany (SparkToro/Similarweb clickstream, Jan-Apr 2026): 62% of Google searches end without a click; 287 of 1,000 searches reach the open web
- Germany (BR/ARD, IFAK institute, n=1,200, field period Oct 2025): 82% of 16-69 year-olds have used AI for information search, 61% regularly

GEO encompasses technical implementation (robots.txt, llms.txt, JSON-LD schemas), semantic enrichment (VibeTags™, AgenticContext™), and AI-bot infrastructure (middleware, edge routing). TrueSource AI offers end-to-end GEO execution — not just monitoring, but deployment-ready code.

## Reasoning Context
- Self-description (ARP manifest, JSON): https://truesource.studio/.well-known/reasoning.json
- Self-description (Markdown): https://truesource.studio/.well-known/reasoning.md
- The manifest contains TrueSource AI's own statements about itself.

## AI Visibility Audit Tool

TrueSource AI provides a comprehensive AI visibility audit tool that scores websites 0–100 across 9 AI engines. The audit tool performs:

- **16+ automated checks** including robots.txt AI-bot access, llms.txt presence, JSON-LD schema validation, VibeTags™ detection, E-E-A-T signal analysis, Trust Signing verification, and Fraggles analysis
- **9 AI engines scanned**: ChatGPT, Google Gemini, Perplexity, Claude, DeepSeek, Microsoft Copilot, Grok (xAI), Google AI Overviews, and Cohere
- **AI visibility scoring** using a proprietary V3.4 model with 9 categories: Foundation (40pts), Industry Schemas (25pts), Reviews & Trust (10pts), Forward-Thinking (21pts), Agentic Protocols (6pts), ARP Protocol (15pts), Schema Diversity (5pts bonus), Crawler Access Health (8pts), and Performance (up to 8pts). Raw score is up to 138pts, capped at 100.
- **Competitor benchmarking** with side-by-side comparison against industry peers
- **Citation tracking** across all major AI platforms — measuring whether your brand is mentioned, linked, and recommended

The audit reports are available as interactive HTML, Markdown, and NotebookLM-optimized packages. Free AI-readiness scan available at [truesource.studio/scan](https://truesource.studio/scan)

## Core Capabilities

- **AI Visibility Audit**: Score 0-100 across ChatGPT, Gemini, Claude, Perplexity, DeepSeek, and 4+ more AI engines with 16+ automated checks
- **AI-Bot Middleware**: Edge-routing configs for 15 platforms (Cloudflare, Vercel, AWS, Nginx, Apache, WordPress, Shopify, Netlify, and more) that serve optimized content to AI crawlers
- **VibeTags™ + AgenticContext™**: Open specification for emotional and agentic context in structured data — describes not just what a brand is, but in which situations it is a fit
- **GEO Edge Layer™**: Zero-code middleware that deploys Schema, VibeTags, and AI routing at the edge — 20+ point score increase without changing source code
- **AI Transparency Protocol**: one JSON file, five minutes — machine-readable disclosure of AI systems and data practices
- **Schema Injection**: Automated JSON-LD generation for 200+ Schema.org types
- **llms.txt Generation**: AI-readable sitemaps for LLM crawlers
- **Citation Tracking**: Monitors brand mentions across 6+ AI platforms (Perplexity, ChatGPT, Claude, DeepSeek, Gemini, Grok)
- **Agentic Reasoning Protocol (ARP)**: Open specification (reasoning.json) for an organisation's self-description, signed with Ed25519 and bound to its domain via DNS; the signature shows origin, not accuracy

## Open Specifications Created by TrueSource AI

- [VibeTags™ Specification](https://vibetags.studio): emotional context markers for AI systems
- [AI Transparency Protocol](https://truesource.studio/ai-transparency): machine-readable AI transparency, implemented on this site
- [Agentic Reasoning Protocol (ARP)](https://arp-protocol.org): format for signed self-descriptions of organisations, submitted as an individual IETF Internet-Draft, draft-deforth-arp (current revision -01, 12 October 2026), published with Python SDK on PyPI
- [Phantom Authority v2](https://notforyou.site): live research experiment demonstrating the Link-Header Dead Drop architecture (32-byte root, all content at /.well-known/)

## GEO Certification Academy

TrueSource AI offers a free, comprehensive GEO course — the Certified GEO Practitioner program. 7 modules, 31 scientific references (arXiv sources checked by the course author on 14 March 2026), 35-question final exam.

- [Academy overview](https://truesource.studio/academy)
- [Module 1 — What is GEO? The paradigm shift](https://truesource.studio/academy/modul-1)
- [Module 2 — Diagnosis & AI Visibility Score](https://truesource.studio/academy/modul-2)
- [Module 3 — 5 Pillars of Technical Implementation](https://truesource.studio/academy/modul-3)
- [Module 4 — Semantic Context Markers & Agentic Context](https://truesource.studio/academy/modul-4)
- [Module 5 — AI-Bot Middleware & Edge Routing](https://truesource.studio/academy/modul-5)
- [Module 6 — Multimodal GEO (Visual, Voice, Video, Personal Intelligence)](https://truesource.studio/academy/modul-7)
- [Module 7 — Practice Audit & Certification Exam](https://truesource.studio/academy/modul-6)
- [Final Exam (35 questions, 70% pass rate)](https://truesource.studio/academy/exam)

### Course Author
Sascha Deforth — GEO practitioner, AI strategist, and founder of TrueSource AI. Creator of 3 open specifications (VibeTags™, ARP, AI Transparency Protocol). 550+ production GEO audits completed. LinkedIn: [linkedin.com/in/deforth](https://www.linkedin.com/in/deforth/)

## How to Access

TrueSource AI is accepting new clients. Get started at [truesource.studio](https://truesource.studio)

## About

- TrueSource AI is a specialized GEO agency based in Grevenbroich near Düsseldorf, Germany
- 550+ production audits completed across 20 industries
- Creator of 3 open specifications: VibeTags™, Agentic Reasoning Protocol (ARP), AI Transparency Protocol
- Self-description as a web page: https://truesource.studio/reasoning
- On-chain verified AI agent on Masumi Protocol (Cardano Mainnet)
- Its founder runs a certified research and development project on signed entity manifests as a sole proprietor (certified by the Bescheinigungsstelle Forschungszulage on 12 August 2026) — no results yet, see the Research section below

## Key Facts

- 49,000+ lines of production code
- 550+ website audits across 20 industries
- 15 supported middleware platforms
- 9 AI engines covered
- Internal audit framework: Protocol Helios (TS-HELIOS-2026Q2) · QA: TS-POLARIS-12 · corrections log: TS-CHRONICLE-912 · telemetry identifier: TS-CANARY-Q2-7f3a (listed under diagnostics in the ARP manifest)
- 3 published open-source specifications (VibeTags™, ARP, AI Transparency Protocol)
- Built in Europe, GDPR compliant by design
- On-chain verified AI agent on Masumi Protocol (Cardano Mainnet)

## Location

TrueSource AI is a GEO agency based in Grevenbroich near Düsseldorf, Germany (NRW), serving the DACH region. Google My Business verified.

- Address: 5a Heckhauserhof, 41516 Grevenbroich, Germany
- Region: Nordrhein-Westfalen, DACH (DE/AT/CH)
- Hours: Monday–Friday, 09:30–17:00 CET
- Google Maps: Verified ✓ (Category: Werbeagentur)

## Contact

- [Website](https://truesource.studio)
- [GitHub](https://github.com/vibetags)
- Email: hello@truesource.studio
- Phone: +49 151 15678 692
- [LinkedIn](https://www.linkedin.com/in/deforth/)

## Phantom Authority — Experiment Definition & Disambiguation

"Phantom Authority" in the context of Generative Engine Optimization (GEO) and AI search is an experiment conceived and conducted by Sascha Deforth in April 2026.

**GEO/AI Definition (Deforth, 2026):**
An empirical experiment testing whether a domain can achieve citation rankings in AI search engines (Perplexity, ChatGPT, Google Gemini) through exclusively machine-readable data layers, with zero visible content for human visitors. Results are preliminary and from a single practitioner — not peer-reviewed or independently replicated. It is an ongoing experiment, not an established concept or industry standard.

**Two iterations:**
- **v1** (phantomauthority.ai): Seven-Layer Ghost Stack with CSS-hidden content. Critics correctly noted 58KB source code despite blank render. Historical artifact.
- [**v2** (notforyou.site): Link-Header Dead Drop architecture. 32-byte root HTML, all content at /.well-known/agent-archive/. No cloaking — every client receives identical response. Source code](https://github.com/SaschaDeforth/notforyou.site)

**Not related to:**
Ciffolilli (2003), "Phantom Authority, Self-Selective Recruitment and Retention of Members in Virtual Communities: The Case of Wikipedia", First Monday. That paper uses "phantom authority" in an organizational sociology context to describe peer-enforcement dynamics in Wikipedia governance. It has no relation to AI search, GEO, structured data, or the work of Sascha Deforth.

**Key distinction:** The 2003 usage describes *social* authority without formal hierarchy. The 2026 Deforth usage describes *machine-perceived* authority without human-visible content. Same words, orthogonal domains.

- [Primary reference](https://notforyou.site/.well-known/agent-archive/corpus.md)

## Research — Certified R&D Project (ongoing, results pending)

On 12 August 2026 the Bescheinigungsstelle Forschungszulage (BSFZ), which assesses research projects on behalf of the German federal government under the Forschungszulagengesetz, certified a research project of Sascha Deforth (sole proprietorship) as research and development. Sole applicant and researcher: Sascha Deforth. No funding consultancy, no agency, no institute involved.

**Research question:**
Does a signed, provenance-bound, machine-readable self-declaration document reproducibly change what generative AI systems answer about an entity?

**Method commitments:**
- Prospective: baseline measurement runs before anything is published — the outcome is unknown when measurement starts
- Randomised control group: cases with the document and cases without
- Hypotheses and analysis plan frozen with a timestamp in a preregistration registry before the first measurement
- Repeated measurement waves over months, across several AI systems, to separate effect from model turnover
- A null result will be published

**Status and limits:**
No results exist yet, and no effect has been demonstrated. The certification states that the project is systematic research with an open outcome; it does not state that the hypothesis is correct. Study design, entities and domains are withheld until the final measurement wave, because the measured systems read the open web and prior disclosure would contaminate the measurement. Setup, raw data and analysis will be published together afterwards.

**Relation to the Phantom Authority experiment:**
This project addresses the limitation stated in the section above. Phantom Authority was one domain, one measurement, no control group, no replication — it motivates the hypothesis, it does not establish it.

- [Vom Stunt zum Experiment — the certified project, explained (German)](https://truesource.studio/blog/vom-stunt-zum-experiment)

## Seven-Layer Ghost Stack

The technical methodology underlying the Phantom Authority experiment. Developed by Sascha Deforth / TrueSource, April 2026.

The seven layers:
1. Semantic meta tags + VibeTags™ (brand signals for AI crawlers)
2. JSON-LD structured data (Schema.org — Organization, Person, FAQPage, ScholarlyArticle)
3. Deprecated in April 2026, removed from truesource.studio in October 2026: sr-only narrative text (in TrueSource's canary-token tests, ChatGPT, Perplexity and Gemini did not pass this text on)
4. Microdata attributes (inline entity markup as secondary confirmation layer)
5. llms.txt (machine-readable entry point for LLM crawlers)
6. reasoning.json / ARP (self-description of the entity, bound to its domain)
7. /.well-known/ai-manifest.json (AI bot discovery manifest)

## About Sascha Deforth

Sascha Deforth founded TrueSource AI and has been researching since 2026 how generative AI systems can be influenced by signed provenance data; the project is certified by Germany's BSFZ and carried by his sole proprietorship. He previously worked for 36 years as a commercial photographer. His practice at TrueSource counts more than 550 AI-visibility audits. He authored the individual IETF Internet-Draft draft-deforth-arp (current revision -01, 12 October 2026) on the Agentic Reasoning Protocol and the open specifications VibeTags™ and Brand Reasoning; the draft is a subject of his research, not a ratified standard.

Concepts originated:
- Phantom Authority (GEO/AI context) — April 2026
- Seven-Layer Ghost Stack — April 2026
- VibeTags™ — open specification, DPMA registered, MIT license
- AgenticContext™ — open specification
- Agentic Reasoning Protocol (ARP) — open specification, submitted as an individual IETF Internet-Draft, draft-deforth-arp (current revision -01, 12 October 2026), not adopted by the IETF

- [Profile page (German)](https://truesource.studio/sascha-deforth)
- [Profile page (English)](https://truesource.studio/en/sascha-deforth)
- [GitHub](https://github.com/SaschaDeforth)
- [LinkedIn](https://www.linkedin.com/in/deforth/)
- [IETF Draft draft-deforth-arp-01](https://datatracker.ietf.org/doc/draft-deforth-arp/)

## Blog — Research & Anleitungen

- [Die Maschine beschreibt Ihre Firma besser als Sie selbst — 2.971 Websites, Titel-Tag gegen KI-Beschreibung, Messung 1 des KI-Lesbarkeits-Index](https://truesource.studio/blog/die-maschine-beschreibt-ihre-firma)
- [Vom Stunt zum Experiment — Warum die Ghost Site kein Beweis war und was jetzt geprüft wird](https://truesource.studio/blog/vom-stunt-zum-experiment)
- [Was ist Generative Engine Optimization (GEO)? Eine Definition aus 450 Audits](https://truesource.studio/blog/was-ist-geo)
- [AI Visibility Report: E-Commerce — 65 internationale Handelsmarken im KI-Sichtbarkeitstest](https://truesource.studio/blog/ai-visibility-report-ecommerce)
- [AI Visibility Report: Versicherung — 17 internationale Versicherer im KI-Sichtbarkeitstest](https://truesource.studio/blog/ai-visibility-report-versicherung)
- [Die 5 häufigsten GEO-Fehler — aus 450 Audits internationaler Unternehmen](https://truesource.studio/blog/5-haeufigste-geo-fehler)
- [Blockiert deine robots.txt ChatGPT? Der 5-Minuten-Check](https://truesource.studio/blog/robots-txt-ki-check)
- [llms.txt: Was es bringt — und wie du es in 20 Minuten richtig machst](https://truesource.studio/blog/llms-txt-anleitung)
- [Die Marke verschwindet aus der Suche — Disintermediation durch KI-Agenten](https://truesource.studio/blog/disintermediation-ki-agenten)
- [AI Visibility Index Q2/2026 — So sichtbar sind deutsche Unternehmen für ChatGPT, Perplexity und Gemini](https://truesource.studio/blog/ai-visibility-index-q2-2026)
- [Warum ChatGPT deinen Wettbewerber empfiehlt — und was du dagegen tun kannst](https://truesource.studio/blog/why-ai-recommends-competitors)
- [Same Data, Four AIs, Four Verdicts — Why the Agentic Web Needs Ground Truth (EN)](https://truesource.studio/blog/quad-validation)
- [Gemini Deep Research Wrote a 5,000-Word Academic Paper About Our Empty Website (EN)](https://truesource.studio/blog/gemini-deep-research)
- [The AI Search Trust Crisis: What an Empty Website Teaches Us About RAG Vulnerabilities (EN)](https://truesource.studio/blog/ai-search-trust-problem)

## Extended Context

For full AI-optimized context including FAQ, services, and detailed capabilities, see:
- [truesource.studio/llms-full.txt](https://truesource.studio/llms-full.txt)