AICanary separates two questions: can AI systems read and use the site, and do assistants actually recommend the business in answers?
Published by PX7 Digital so the benchmark can be cited, challenged, and repeated.
Can a crawler fetch the public homepage, and do robots rules allow common AI and search crawlers to inspect it?
Does the raw public response contain useful text without needing a browser-only app shell? AI systems cannot cite copy they cannot read.
Title, description, canonical URL, headings, Open Graph, viewport, language and indexability are checked because assistants still depend on web retrieval quality.
llms.txt, markdown, API catalog, MCP, OAuth, A2A and commerce signals are scored when relevant to the site's goals.
AICanary asks real category questions across supported AI engines and records whether the monitored business appears in the answer.
When the business is absent, AICanary extracts the names that get recommended, creating a practical share-of-answer view.
Reports identify the sources that shape the answer and summarize whether the business is framed positively, neutrally, negatively or not at all.
Assistant answers are non-deterministic. AICanary stores samples and trends them over time instead of treating one answer as a final truth.
The country reports scan public primary domains from the outside. They do not use private data, vendor access, surveys or company participation.
Each country page states the number of companies in scope and the index or selection used. It is a benchmark sample, not every business in the country.
An invisible result may reflect deliberate bot protection. The result still matters: it describes what an AI crawler experiences from the public web.
The Danish benchmark is the anchor report and evidence base for PX7 Digital's AI visibility reviews with Danish leadership teams.
Run the free foundation report first. Use a deep analysis when you need answer evidence across AI engines.
Run the free report