The 19 AI visibility signals
A Biencité scan reads your site the way a generative AI crawler would, read-only, and checks 19 signals across 3 axes. Here is the full list, how the score reads, and what the measurement does not tell you.
How the score reads
Every signal belongs to an axis, and each axis reads on its own. A site can be perfectly visible and still be uncitable: two different problems, two different answers.
Visible
Can AI crawlers read your site? Access, HTML content, essential information.
Citable
Is your content structured to be quoted in an answer? Structured data, reading guides, entities.
Agents
Can an AI agent understand your services and act on your site? The layer almost nobody has yet.
The 19 signals, axis by axis
What the free scan checks. Each signal names a capability: what your site does, or does not, allow a generative AI to do.
Visible 5
- Secure connection
- AI crawlers allowed in
- Content readable directly by AI
- Essential page information
- Usable site map
Citable 8
- Structured data for AI
- Content structured for citation
- Optimized sharing preview
- Dedicated reading guide for AI
- AI usage preferences declared
- Structured key facts for AI
- Content ready for voice
- Linked entities and profiles
Agents 6
- Agent-readable version
- Capability card for agents
- Agent skills index
- Agent authentication framework
- Exposed service catalog
- Agent access guide
19 signals in total.
The four verdicts
The verdict does not follow from the score alone, it follows a cascade. A site AI cannot read is invisible, whatever the rest of the grade says.
Invisible to AI
AI assistants cannot read your site: its content is out of reach. Right now, you do not show up in their answers.
Visible, but not citable
AI assistants can see your site, but it is not structured to be cited. They see you without being able to recommend you.
Readable and citable
AI can read and cite your site, but the agents layer is missing. You stay behind sites that are ready for AI agents.
Ready for AI agents
Your site is readable, citable and ready for AI agents. You are ahead of the vast majority of sites.
What could not be measured
An unreachable file, a timeout, a server error: the signal is marked unmeasured and leaves the calculation instead of counting as zero. A failed measurement is not an observed absence. Every report therefore states how many of the 19 signals were actually checked.
Why your platform changes the result
Shopify, Wix, Webflow and Framer lock certain files at the domain root. Those signals are marked impossible on your platform instead of counting against you, and never show up in your action plan. You will never get advice your site cannot apply.
What the method does not tell you
- It does not predict a citation. It measures eligibility: the conditions without which a citation is very unlikely.
- It measures neither your brand awareness, nor your backlinks, nor what AI already says about you. That is what citation tracking is for, and it is a separate measurement.
- It does not judge the editorial quality of your writing.
- The weight of each signal is not published: it moves with how generative engines behave. A score compares to itself over time, not to another site’s.
Questions about the method
Does the scan modify my site?
No. The scanner reads your homepage and a few public files (robots.txt, sitemap, llms.txt, agent files), exactly the way an AI crawler would. Nothing is installed, nothing is written.
Why 19 signals, and not 50?
Because each signal we keep matches a capability generative AI actually uses to read, cite or act on a site. Padding the list with criteria that change nothing would inflate the report without improving your visibility.
Does a score of 100 guarantee a citation?
No. A high score means your site meets the technical conditions for citation. What AI then cites depends on your reputation, your sources and the question asked. No serious tool can promise a citation.
Are the signal weights public?
No, deliberately. The signals themselves are all published on this page. Weights move as generative engines evolve, and publishing them would invite optimizing the score rather than actual visibility.