OUR AI POLICY.
We sell AI visibility — so we say openly where AI runs on our side and where it does not. This page is our AI inventory: which systems we use, what for, who approves them and how we label content.
This translation is provided for convenience. Only the German version is binding.
Why this page exists
Since February 2025, Art. 4 of the EU AI Act (Regulation (EU) 2024/1689) has required companies using AI to have sufficient AI literacy — to know what the systems they deploy do and where their limits are. From 2 August 2026, the transparency duties of Art. 50 apply on top: anyone talking to an AI must be able to recognise it, and AI-generated content must be identifiable as such. A documented, public policy is the simplest evidence of that — and the most honest answer to whether a human or a machine writes here.
Which AI systems we use
1) Surfaced Analytics (our own tool): rule-based detection of which AI systems send visitors to a website — no generative model, just a deterministic analysis of referrers, parameters and crawler signals. 2) Technical checks (AI readiness): purely technical, deterministic checks of the publicly available HTML — with no AI model involved. 3) Writing support in the Complete plan: Claude (Anthropic) helps us research and draft the FAQ answers and page texts. Every text is written, revised and owned by a human before it reaches a client. 4) AI assistants in internal development: we use AI assistants for code, research and analysis; every result is reviewed by a human.
What is never invented by an AI
The technical blocks — llms.txt, robots.txt, meta tags, H1, Organization and FAQ schema, about section — are assembled by the dashboard without a language model, deterministically from the company data you entered yourself: same input, same result, every line traceable to its source, nothing invented. The texts in the Complete plan are written by a human; AI only helps with research and drafting, and every statement is checked against your company profile. Because selling AI visibility while having an AI invent the content delivers exactly the interchangeable text the AIs already know.
How we label AI-generated content
Should we publish text, images, audio or video generated entirely or in substantial part by an AI, we label it right at the content — visibly, not buried in the small print. The same applies to client deliverables: if a building block were exceptionally produced AI-assisted, it would say so in the document. At present there is no fully AI-generated content on surfaced.at.
No chatbot on this website
There is no chatbot and no AI assistant talking to you on surfaced.at. If you write to us via the form or by email, a human answers — usually Leopold himself. Should we ever change that, the bot will identify itself as an AI in its first message, as Art. 50(1) requires.
What data goes into AI systems
Our AI visibility tests use publicly available data only: company name, website, industry, market and language — exactly what a potential customer would ask the AI as well. Personal data from contact requests, credentials and confidential client documents are not entered into AI systems. No automated decision-making with legal effect within the meaning of Art. 22 GDPR takes place: every assessment that reaches a client has been reviewed by a human. Details on data processing are in the privacy policy.
Responsibility and competence
Responsibility for the use of AI at Surfaced lies with management; it approves every system deployed and reviews the results. We work with AI systems daily and track model updates, crawler behaviour and regulation continuously — that is our core business. We only put a new system into operation once it is clear what data it sees, where that data is processed and how its results can be verified. This policy is updated whenever our use of AI changes.
Limits of this page
This policy describes our actual use of AI and is not legal advice — neither for us nor for you.
Last updated: September 2026 · Surfaced · kontakt@surfaced.at