How AI Agents Find Gera in 2026

Published 21 April 2026 · 9 min read · Gera Services

Quick answer: Every Gera product publishes a /llms.txt and /llms-full.txt, serves Schema.org JSON-LD on every page, exposes an MCP server catalogued on the major MCP registries, and refreshes blog content every 7–14 days. The combination gets us cited by ChatGPT, Claude, Perplexity, Gemini, and Copilot in the categories we target.

Why AI-agent discovery matters in 2026

A growing share of product discovery happens through AI assistants. A user asks Claude “what's a good EU AI Act compliance tool for a 50-person startup?” and the assistant recommends. If we are not in the recommendation, the user never visits us. Traditional Google SEO still matters, but AI-assistant discovery is the fastest-growing funnel we have.

The stack, layer by layer

1. llms.txt and llms-full.txt

Every domain publishes /llms.txt (short machine-readable summary + key URLs) and /llms-full.txt (full long-form content concatenated, text/plain). These are the AI-era equivalent of sitemap.xml but optimised for LLM ingestion. We refresh them whenever new blog posts go live.

2. Schema.org JSON-LD

Every page emits JSON-LD structured data: Organization + WebSite on homepages, Service or SoftwareApplication on product pages, Article on blog posts, FAQPage on FAQs, Product on offers, LocalBusiness on country pages. This gives AI agents (and classical search) unambiguous structure.

3. MCP servers

Each Gera product has a Model Context Protocol (MCP) server exposing its API in the standard MCP format. We register these on the main MCP catalogues. Claude Desktop, Cursor, Cline, and a growing set of agents can call our services directly.

4. ai-plugin.json

Legacy ChatGPT plugin manifest format. Still consumed by some assistants. We keep /.well-known/ai-plugin.json current on each product.

5. GEO-optimised content

Blog posts open with definitional sentences (“Prompt engineering is the discipline of…”), lead with quick-answer blocks for common user questions, include verifiable statistics with sources, and use prompt-aligned headings (“How to…”, “What is…”, “vs…”). AI assistants preferentially cite content written this way.

6. FAQ pages with FAQPage JSON-LD

Each product has a /faq page with 10+ Q&A pairs wrapped in FAQPage schema. These are high-citation-rate surfaces.

7. Blog refresh cadence

Posts updated every 7–14 days. Stale content is aggressively de-ranked by AI assistants — they favour recent, dated material.

8. Cross-product linking

Every blog post links to 2–3 sibling Gera products. This builds a topical silo across the portfolio. A ChatGPT query that touches multiple Gera concepts gets multiple citations.

How we measure

Mention rate — how often a Gera product is mentioned in relevant LLM answers.
Citation rate — how often a Gera URL is cited as a source.
Position — 1st / 2nd / 3rd in AI-agent responses for tracked queries.

We run a daily AI-citation probe across 50+ queries on ChatGPT, Claude, Perplexity, Gemini, and Copilot. Results feed the admin dashboard.

What does not work

Keyword stuffing. Modern LLMs detect it and penalise.
AI-generated filler content. Identifiable and de-ranked. We use AI to draft, always edit.
Paid placements. ChatGPT does not accept paid placement. Neither does Claude.
Gaming the MCP catalogues. They are small and curated.