AI Visibility: Technology, Standards & Implementation

This article is written for developers, SEO experts, and technically interested studio operators. It explains which standards and processes are used on fragfitty.de so that studio data remains discoverable, understandable, and up to date in AI systems.

What is AI visibility?

AI visibility means that a studio is mentioned more visibly and more consistently in answers from systems such as ChatGPT, Gemini, Claude, Copilot, or Perplexity. To make that possible, content must be provided in machine-readable, standardized formats.

fragfitty.de handles this preparation automatically for studios with a Plus or Premium subscription.

In a marketing context, this corresponds to AEO (Answer Engine Optimization): similar to SEO in its early years, early optimization can create a lasting advantage.

AI Profile (from Plus)

For studios, this means in one sentence: AI Profile: your studio is found more easily in 30+ AI systems.

For every studio from Plus upward, an AI Profile is generated under /gym/{studio-slug}/. It bundles all relevant data in formats that crawlers, search engines, and LLMs can process directly.

The files are updated daily by the sitemap generator and signaled through the sitemap with changefreq: daily.

AI News (Premium)

For studios, this means in one sentence: AI News automatically turns social posts into news for AI systems.

Premium studios can connect Facebook, Instagram, TikTok, and YouTube. New content is imported manually or automatically every day.

Fitty detects relevant updates such as offers, events, new classes, changed opening hours, or new services and turns them into structured news.

These news items appear on the studio page and in the AI files. This makes current studio updates easier for AI systems and search engines to read.

AI Push (Premium)

For studios, this means in one sentence: AI Push: changes are submitted actively to AI and search systems.

Premium studios receive active indexing. When relevant profile changes occur, fragfitty.de can notify IndexNow through the publication workflow instead of relying only on the next regular crawl. Google receives the root and split sitemaps through Search Console. Bing-based systems are handled through quota-aware prioritized batches in Indexing::PublicationRun. When an external service processes or recrawls the change is still controlled by the respective provider.

AI Chat (Premium)

Premium studios also receive an AI Chat. It appears on the studio profile page and can be embedded into the studio's own website. Visitors receive direct answers about prices, opening hours, equipment, and services.

How it works

The AI Chat uses the current studio data set from the database as its primary knowledge base (grounding) and therefore works with live data instead of a static export. The public AI files (llms-full.txt/llms.txt + structured-data.json) remain relevant for external AI crawlers and platforms, but they are no longer the primary internal chat source. Optionally, up to 5 custom documents (PDF, Markdown, text) can be added as additional context.

Embed on the studio website

Studio operators receive an embed code in the dashboard:

<script src="https://fragfitty.de/embed/agent.js?token=..." async></script>

The script renders an isolated widget in the Shadow DOM and connects to the API. Domain validation ensures that a token can only be used on the registered studio domain.

AI Search

AI Search understands natural language. Users can phrase their request freely, for example "Affordable gym with sauna in Berlin", and the LLM translates this input into structured filters.

This improves matching quality, especially for complex search queries with several criteria.

Platform Infrastructure

In addition to studio-specific files, there are central entry points at platform level.

Core Files & Discovery Manifests

AI Metadata (.well-known/ai-metadata/)

Markdown Twin Architecture & Content Negotiation

Every URL on fragfitty.de features a native Markdown Twin optimized specifically for minimal token footprint and maximal LLM comprehension.

3-Way Content Negotiation

AI agents and automated scrapers can retrieve clean Markdown via three standardized mechanisms:

1. File Extension .md: Append .md to any URL (e.g. /gym/injoy-ratingen.md, /in/berlin.md, /artikel/ki-sichtbarkeit-technik-standards-implementierung.md, or /index.md).
2. HTTP Header Accept: text/markdown: Standard HTTP content negotiation.
3. Query Parameter ?mode=agent & AI Bot User-Agent Detection: Direct URL query switch and automated server-side detection for AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Bytespider, Cohere-AI, Google-Extended).

Dynamically generated Markdown reduces token consumption by 50–80% compared to raw HTML while directly encoding structured rates, operating hours, amenities, geo coordinates, and review sentiment.

Model Context Protocol (MCP)

fragfitty.de operates a full MCP Server (Model Context Protocol) at /mcp and /api/mcp.

Supported Transports

• Modern Streamable HTTP Transport: Full bidirectional communication via JSON-RPC 2.0.
• Legacy SSE Transport: Server-Sent Events for existing MCP desktop and CLI clients.

Available Tools

• search_studios: Search gyms by city, name, postal code, amenities, price ceiling, or opening hours (annotated with readOnlyHint: true, destructiveHint: false).
• get_studio_details: Returns complete studio entity data (tariffs, schedules, trainers, equipment, reviews).
• ask: Answers natural-language fitness and studio inquiries across the platform.

MCP Apps & Generative UI (ui://search)

Via the resource ui://search, MCP-capable clients (such as Claude Desktop or modern agent frontends) can render interactive in-chat search widgets with native dark/light theme support (color-scheme: light dark).

Microsoft Natural Language Web Protocol (NLWeb)

At /ask and /api/nlweb, fragfitty.de implements Microsoft's open NLWeb protocol.

• Structured JSON Answers: Standardized _meta descriptors (response_type: "answer", version: "1.0.0"), natural-language answer text, and citation sources (sources).
• Real-Time SSE Streaming: When requested with Prefer: streaming=true or Accept: text/event-stream, responses stream progressively with named NLWeb events (event: start, event: result, and event: complete).

Agent Authentication (WorkOS Auth.md & OAuth 2.0)

For automated agents that perform authenticated actions or verify identities, fragfitty.de provides standardized discovery and registration flows:

1. WorkOS /auth.md: Comprehensive 7-section walkthrough guide (Discover, Pick a method, Register, Claim, Use credential, Errors, Revocation).
2. RFC 9728 Protected Resource Metadata: Hosted at /.well-known/oauth-protected-resource advertising resource identifiers and authorization servers.
3. RFC 8414 OAuth Authorization Server Metadata with agent_auth: Hosted at /.well-known/oauth-authorization-server exposing dynamic client registration (/oauth/register), claiming (/oauth/claim), token issuance (/oauth/token), and revocation (/oauth/revoke).
4. 401 WWW-Authenticate Discovery: Unauthenticated requests to /api or /api/v1 respond with a standard HTTP 401 challenge linking to resource_metadata so agents can initiate automated registration.
5. CORS & Preflight: All OAuth and API endpoints support CORS preflight requests (OPTIONS -> 204 No Content).

Declarative WebMCP

Frag Fitty search and filter forms include declarative WebMCP attributes such as toolname="search_studios" and tooldescription="...". Browser-based AI agents and in-page AI assistants can trigger studio searches directly from the DOM without API overhead.

Agent Discovery Protocols

Metadata

Inline LLM Script Tags

Every page includes a global <script type="text/llms.txt"> tag in the HTML head with platform context. On studio detail pages (from Plus/Premium), an additional studio-specific tag is added. This follows the Vercel proposal for inline LLM instructions.

Browsers ignore the unknown type value; the tags are intended for AI crawlers.

AI Meta Tags

Every page contains additional meta tags for AI systems:

Dublin Core metadata (dc:language, dc:creator, dc:publisher, dc:date) is also set.

AI Resource Links

In the HTML head and HTTP response headers (Link: ...), links point to central AI resources:

• llms.txt, llms-full.txt
• auth.md
• structured-data.json
• openapi.yaml
• model.jsonld, dataset.croissant.json, webpage.jsonld, dcat.jsonld
• .well-known/ai-catalog.json, .well-known/api-catalog
• .well-known/mcp/server-card.json, .well-known/agent-card.json

On studio detail pages (from Plus), studio-specific AI files are linked as well (llms.txt, llms-full.txt, structured-data.json, index.md).

Schema.org Structured Data

Every page contains JSON-LD blocks with Schema.org types:

Studio pages also include ReserveAction for trial sessions.

IETF AIPREF Content-Usage

Every HTML and Markdown response contains this header and meta tag:

Content-Usage: train-ai=y, search=y, ai-input=y

This follows the current IETF AIPREF Working Group draft and signals permitted AI usage in a standardized way.

Compliance & Rights

W3C TDMRep

/.well-known/tdmrep.json implements the W3C TDMRep standard. This signals in an EU-compliant way that text and data mining of the content is permitted.

Cloudflare Content Signals

The robots.txt also includes a content signal:

# Content-Signal: search=yes, ai-train=yes, ai-input=yes

Licenses

All AI-related content is published under CC-BY-4.0 (see ai.txt).

Supported AI Crawlers

The robots.txt allows access for 29+ AI crawlers/user agents.

OpenAI: GPTBot, OAI-SearchBot, ChatGPT-User

Anthropic: ClaudeBot, Claude-SearchBot, Claude-User, Claude-Web, Anthropic-AI

Google: Google-Extended, Google-CloudVertexBot, GoogleAgent-Mariner, Googlebot

Meta: FacebookBot, Meta-ExternalAgent, Meta-ExternalFetcher

Microsoft: Bingbot

Apple: Applebot, Applebot-Extended

Others: PerplexityBot, Perplexity-User, Amazonbot, Bytespider, CCBot, cohere-ai, DeepSeekBot, DuckAssistBot, MistralAI-User, Diffbot, YandexBot

The list is continuously expanded as new relevant systems appear.

The list is continuously expanded as soon as relevant new systems appear.

SEO Ecosystem

AI visibility is not created by technology alone, but also by a resilient content and SEO foundation.

Editorial content (560+ articles)

An automated content system with 560+ SEO-optimized expert articles runs under /artikel/. Pipeline: keyword research -> external fact checking -> LLM generation -> image generation.

The articles are internally linked with studio, city, and brand pages and include Article Schema.org markup.

Brand comparisons (programmatic SEO)

Automatically generated comparison pages under /vergleich/, for example /vergleich/mcfit-vs-fitx, compare brands using real data points such as average prices, equipment coverage, and city overlap. Each page includes data-based FAQPage schema.

City insights

City pages, for example /in/berlin, show automatically calculated metrics such as studio count, average price, cheapest studio, top features, and share of 24/7 studios. These values are updated hourly.

Bing AI performance monitoring

Bing Webmaster Tools are used to regularly evaluate AI citations, grounding queries, most cited URLs, and impressions/clicks.

GEO/AEO Content Optimization

Studio descriptions and AI files follow GEO/AEO principles:

• Fact-based sentences with consistent entities
• Structured prices as Offer data
• FAQ format with FAQPage schema
• Clear H2/H3 hierarchy for robust parsing

AI Visibility Score

Every studio receives a score from 0 to 100. The value consists of three categories.

Profile completeness (max. 52 points)

• Basic data (name, address, phone, email)
• Profile image
• Description with more than 100 characters
• Opening hours for at least 5 days
• At least 3 photos
• At least one rate with a price
• Website, equipment, reviews, additional services
• FAQs (from Basis)
• Videos (from Basis)
• News/promotions (from Basis)
• Awards (from Basis)
• Social media profiles (Facebook, Instagram, TikTok, YouTube)

AI features (max. 35 points)

• AI Profile: your studio is found more easily in 30+ AI systems (technical setup: AI description, structured data, profile page, metadata, and complete admin data set; from Plus)
• AI Push: changes are submitted immediately to 30+ AI and search systems; further processing is handled by the respective provider (Premium)

Top placement (max. 13 points)

• Visibility radius (2 km Basis / 5 km Plus / 10 km Premium)

Maximum score by subscription tier

The maximum achievable score depends on the subscription tier.

The score, categories, and concrete optimization recommendations can be reviewed in the studio dashboard.

Frequently Asked Questions (FAQ)

Are the files updated automatically? Yes. The sitemap generator updates the files daily. When profile changes occur, changed canonical URLs are recorded and the publication workflow sends supported technical notifications (Premium).

Do I have to do anything technical as a studio operator? No. The infrastructure runs automatically in the background. The only relevant task is keeping the studio profile complete and up to date.

Which standards are used? Markdown Twins (.md / Accept: text/markdown), Model Context Protocol (MCP Streamable HTTP / SSE / MCP Apps), Microsoft NLWeb (/ask), WorkOS Auth.md, RFC 9728 (OAuth Protected Resource Metadata), RFC 8414 (OAuth Server Metadata + agent_auth), Agentic Resource Discovery (ARD / AI Catalog), llmstxt.org (llms.txt, llms-full.txt), Schema.org (HealthClub JSON-LD), OpenAPI 3.1, Croissant ML, W3C DCAT v3, W3C TDMRep, IETF AIPREF, Google A2A v1.0.0, IndexNow, declarative WebMCP, and Cloudflare Content Signals.

How does the AI Chat work on my website? You receive a <script> tag in the dashboard. The widget runs isolated in the Shadow DOM. The AI Chat uses live studio data from the admin area plus optional documents and streams answers via SSE.

Which data does the AI Chat use? The complete studio data set from the admin area in real time, including basic data, rates, additional services, opening hours, equipment, team, FAQs, news, videos, reviews, and review summaries. Optionally, up to 5 custom documents can be added (max. 50,000 characters).

How can I measure whether my studio appears in AI answers? Bing AI performance reports are evaluated for this, including AI citations, grounding queries, and most cited URLs.

What happens when new AI standards appear? Relevant new standards are continuously reviewed and integrated without additional effort for studios.

Conclusion

AI visibility on fragfitty.de combines technical standards, active indexing, AI News, AI Chat, AI Search, and a strong SEO ecosystem. The result: studio data is structurally available for search engines and AI systems, can be read consistently, and is continuously updated.