
Turn AI Strategy into Operational Reality
We build what your 3LA audit found: AI-readable content, structured data, AI-crawler monitoring and custom agents for Nordic businesses.
Knowing what to do is only the first step. Nettpilot helps Nordic businesses implement the technical protocols, content structures and custom agents the AI era requires. We bring the framework, the tooling and a founder with more than 25 years of hands-on digital practice.
At a Glance
| What we do | Implement Generative Engine Optimization (GEO), structured data and custom AI agents |
| Method | The Three-Layer Approach™: Human (L1), Search (L2) and AI (L3) |
| Starting point | A free 3LA audit (about 45 seconds) or an AI Reality Audit |
| Typical scope | €5,000 – €25,000 |
| Timeline | 4 – 12 weeks |
| Based in | Jevnaker, Norway. Working across the Nordics and globally |
Escaping “Pilot Purgatory”
Many businesses are stuck in a loop of AI experiments that never reach production. You might have a strategy deck, or you have run a Three-Layer Approach audit and seen the red warning lights.
You would not be alone. In the 3LA analysis sample (753 URLs from 403 domains, October 2025 – September 2026), Layer 3 (AI) was the weakest layer on 54% of domains. The sample is self-selected and mainly Norwegian, so treat it as a direction, not a census.
So how do you actually fix it?
- Which AI crawlers actually visit your site, and which pages do they skip?
- How do you stop your chatbot from making things up?
- How do you restructure 5,000 product descriptions so AI systems can retrieve and cite them?
We bridge the gap between marketing strategy and technical development. We do not just advise; we build, deploy and maintain the infrastructure that connects your business to AI assistants, while still serving your human visitors and still ranking in Google, Bing and other search engines.
Core Service 1: Technical AI SEO (GEO)
Fixing Layer 3: making you readable, retrievable and citable for machines, without breaking Layers 1 and 2.
1. Measure First: AI Crawler Log Analysis
We start with your raw server logs, not assumptions. The 3LA Log Analyzer identifies AI crawlers such as GPTBot, ClaudeBot and PerplexityBot by name. It also shows which pages they fetch and where high-visibility pages get near-zero AI citations. We keep three signals separate: crawl (a bot fetched the page), citation (an AI answer cited it) and referral (a human arrived from an AI assistant).
2. Semantic Schema Architecture
We go beyond basic Schema.org markup. We build a connected entity graph that explicitly defines the relationships between your organization, your products, your people and your expertise. The result is consistent, machine-readable facts across every page.
3. Content Chunking and Atomic Content Units
AI retrieval systems read in chunks. We restructure your core pages into Atomic Content Units (ACUs): self-contained, answer-first passages (BLUF) that keep the claim, the source and the qualification together. This makes them easier for RAG (Retrieval-Augmented Generation) systems to retrieve and harder to misquote.
4. llms.txt Deployment and Maintenance
llms.txt is a proposed convention that gives language models a curated, plain-text map of your most important content. It is not a ratified standard, and AI providers have not committed to honoring it. We deploy it because it is low-cost and low-risk, and we keep it automatically in sync with our AI Context Engine™ plugin for WordPress. We never sell it as a ranking lever on its own.
Core Service 2: Custom AI Agents and Assistants
Building your digital workforce.
Off-the-shelf chatbots are often generic and poorly governed. We build custom AI agents grounded in your data, aligned with your values and integrated into your systems. Our own editorial collaborator, Kira, runs on the same principles.
Internal Knowledge Assistants
Stop wasting time searching for files. We build internal assistants on top of SharePoint, Google Drive or your internal wiki.
Example: “What is our policy on remote work in Sweden?” returns an instant answer, citing the exact section of your employee handbook.
Customer Service and Sales Agents
We build customer-facing agents that act as a first line of support or as a shopping assistant.
- Grounded answers: the agent answers from approved sources, cites them, and says “I don’t know” when the source is missing.
- Measured, not promised: no vendor can honestly guarantee zero hallucinations. We test against a question set from your business and report the error rate.
- Multilingual: Norwegian (Bokmål and Nynorsk), Swedish, Danish and English.
- Action-oriented: agents can check order status, recommend products or book meetings through your existing systems.
Core Service 3: Strategy and Training
Preparing your crew. The best tools are useless if your team does not know how to use them. We focus heavily on the human side of implementation.
Leadership Workshops
AI for Decision Makers: a hype-free session on where AI actually adds value to your bottom line, and where it is a distraction. Delivered by video or on site.
Content Team Training
We teach your writers and marketers to write for all three layers at once. Your team learns the Three-Layer Approach™, how to write Atomic Content Units, and how to use AI tools to speed up production without losing quality or voice.
Core Service 4: AI Visibility and Market Intelligence
Know what AI says about you before you try to change it.
- AI Reality Audit: a baseline built from server log data, structured prompt sampling across the major AI assistants and technical analysis. You get a prioritized action list.
- Competitive AI visibility: how you and your competitors are mentioned, cited and described by the major AI assistants. You choose which models to test, so the analysis matches the assistants your customers actually use.
- Deep research reports: citation-backed market and competitor reports produced with AI research tools, then verified by a human. Delivered as a report, a briefing deck, a podcast-style summary or an infographic.
We separate readiness from observation. A high LayerScore means a page is well structured for humans, search engines and AI. It is not the same as an observed citation rate, and we never present one as the other.
How a Project Runs
- Diagnose: free 3LA audit for key URLs, plus an AI Reality Audit when the stakes justify it.
- Baseline: we record LayerScore per page, AI crawler activity from your logs and current AI citations.
- Build: schema, content restructuring, agents and integrations, delivered in short iterations.
- Measure: we re-run the same measurements and report the change: what improved, what did not, and why.
Who We Work With
Our implementation services are for businesses that need execution, not more advice.
E-commerce and Retail
- AI-readiness for WooCommerce and Shopify catalogs
- Product recommendation and shopping agents
- Structured product data for AI shopping experiences
SaaS and Technology
- Internal knowledge assistants
- Documentation optimized for AI crawlers and coding assistants
- API and MCP integrations for AI assistants
Professional Services
- Client-facing AI assistants
- Automated proposal generation
- Research assistants for legal and consulting teams
Agencies
- White-label AI implementation
- Client-team training on AI workflows
- The 3LA Partner Program
Our Tech Stack
We use the right tool for the job and are transparent about it. We deliberately do not list model version numbers here. Models are replaced every few months, and a page full of outdated versions is exactly the kind of stale signal we help clients remove. We always build on current production models and choose per task for cost, speed and quality.
AI Models and Platforms
- Commercial models: OpenAI (GPT), Anthropic (Claude), Google (Gemini), Mistral, Perplexity
- Open-weight models: Meta Llama, Alibaba Qwen, DeepSeek, for on-premise and data-sensitive use cases
- Agent tooling: Model Context Protocol (MCP), Google Agent Development Kit, Vercel AI SDK
Our Own 3LA Tooling
- 3LA Audit: LayerScore with an L1/L2/L3 breakdown and prioritized fixes
- 3LA Log Analyzer: AI crawler tracking from raw server logs
- 3LA Brand Analyzer: competitive AI perception benchmarking
- 3LA Content Optimizer and AI Context Engine™: WordPress plugins for Schema.org injection and
llms.txtgeneration
Development and Infrastructure
- Backend: Python, FastAPI, Node.js
- Frontend: Astro, Next.js, React
- CMS and headless: WordPress/WooCommerce, Payload CMS, Umbraco
- Hosting and deployment: Railway, Vercel, Google Cloud Platform
- Vector and document stores: Pinecone, Weaviate, ChromaDB, MongoDB
Low-Code, Automation and Prototyping
- Automation: Make.com, Zapier, n8n, Notion
- AI-assisted development: Claude Code, Cursor, Windsurf, v0
- Rapid prototyping: Google AI Studio, Emergent, Lovable, Replit, Bolt
Integrations and APIs
- Communication: Postmark, SendGrid, Mailgun, Twilio
- Business systems: Google Workspace, Microsoft Teams, Slack, Notion, Jira, Trello, Salesforce, HubSpot
- E-commerce: WooCommerce REST API, Shopify API, Stripe
Ready to Build?
We take on a limited number of implementation projects to keep quality high and give you direct access to senior expertise.
Typical project scope: €5,000 – €25,000
Timeline: 4 – 12 weeks
Let’s build something that lasts.
– Nickolass Jensen and the Nettpilot team
Frequently Asked Questions
What is GEO (Generative Engine Optimization)?
GEO is the practice of making content easy for AI assistants such as ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews to retrieve, understand and cite. In practice it combines structured data, answer-first content (Atomic Content Units), crawler access and monitoring of what AI systems actually say about you.
Do you guarantee that AI assistants will cite us?
No, and nobody honestly can. AI answers vary between models, prompts and runs. We improve the factors you control (structure, accessibility, entity clarity and evidence) and measure the outcome with logs, prompt sampling and citation tracking, so you see what changed.
Which AI models do you test against?
Always current production models from the major providers, including OpenAI, Anthropic and Google. The 3LA audit samples several models in parallel. In the AI Reality Audit and Brand Analyzer, you choose which models to include. Every report records the exact model and version used, so results stay comparable over time.
How long does an AI implementation project take?
Typically 4 – 12 weeks, depending on scope and complexity. Most projects start with a free 3LA audit or an AI Reality Audit to set a baseline.
What is llms.txt, and do I need it?
llms.txt is a proposed plain-text file that points language models to your most important content. It is not an official standard, and AI providers have not committed to using it. It is cheap to maintain, so we usually include it, but structured data, clean HTML and well-written content matter far more.
Can you build AI agents that don’t hallucinate?
We build agents that hallucinate far less, and we prove it. Answers are grounded in your approved sources with retrieval (RAG), every answer cites its source, and the agent is instructed to decline when the source is missing. Before launch we test against a question set from your own business and report the measured error rate. Context caching makes repeated queries faster and cheaper, but it does not by itself prevent hallucinations.
What languages do your AI agents support?
Norwegian (Bokmål and Nynorsk), Swedish, Danish and English as standard. Other languages on request.
