HaycionVisibility is a number, not a feeling
Buyers now ask an AI assistant who to consider before they open a website. Whether a client is in that answer used to be a matter of anecdote: someone tried it once, and the client was or was not there. A Haycion workspace turns it into measurement. Two instruments, run on a cadence, compared over time, with the fixes generated from what they find.
The discoverability audit
The first instrument looks at the client’s site the way an AI assistant does. It checks 16 discoverability signals: whether the site can be crawled, whether it says plainly what the company does, whether its structured data exists and is valid, whether an AI-readable description of the site is present, whether pages carry the titles and descriptions an assistant relies on, and the rest of what decides whether a company is legible to a machine that reads literally. Each signal gets a status and, where it fails, a fix. The signals combine into a discoverability score from 0 to 100.
The score is a baseline and a target. Most companies start lower than they expect, because a site written for humans skimming leaves out what an assistant needs stated.
AI citation tracking
The second instrument asks the question directly. The workspace takes the topics the client should own, writes the questions a buyer would ask about them, in a buyer’s words rather than the client’s, and asks ChatGPT and Claude. For each question it records whether the client was cited, which of the client’s pages were cited, and which competitors were cited instead, with their pages.
From those probes come the numbers the client reads: the citation rate per topic and overall, the top cited pages, and the change since the previous run. Run it monthly and the trend is a monthly number.
The competitor visibility comparison
Being cited is relative. The comparison puts the client and their competitors side by side on the same questions: who is cited, how often, for what. It is the visibility equivalent of the Strength Matrix, and it answers the question clients actually ask, which is not “are we visible” but “who is winning the answer instead of us, and why”.
From measurement to fixes
Every failed signal and every lost citation becomes a recommendation, ranked by priority with an effort estimate. Where the fix is an artefact, the workspace generates it: an AI-readable description of the site, structured data for the organisation, its products and its FAQs. Each artefact is validated, marked valid or warned, and packaged in a downloadable bundle for the client’s web team. The playbook guide covers how recommendations become action items.
The client’s agent and its guardrails
Measurement tells you where the client stands. The client’s agent changes it. From the workspace, the agent publishes an accurate, structured public profile of the client that AI assistants can read, built only from a fixed list of public fields, saying only what you and the client approved. The client’s domain must be verified before anything is published, and nothing is published until you publish it. The client agents page describes the setup.
What the client receives
- A discoverability score out of 100, with 16 signals and a fix for each failure.
- Citation checks in ChatGPT and Claude on the questions their buyers ask, with the citation rate, cited pages and the change since last run.
- A competitor visibility comparison showing who is cited instead.
- Prioritised recommendations with effort estimates, and a validated fix bundle.
- A public profile that AI assistants can read, published only when you say so.
- An AI visibility report to send every month, under your name.
How it shows its work
The questions asked are listed with the answers received. Every signal in the audit is shown individually with its status. Every artefact is validated before it is offered. Every citation is tied to a page. The number the client watches move is built from probes they can read themselves, so when it moves, you can say exactly why: which signal was fixed, which page was cited for the first time, which competitor dropped out of an answer.
What a result looks like
A first run on a mid-market software company commonly returns a discoverability score in the fifties and a citation rate that surprises the client, in either direction. The audit lists the missing description and structured data; the fix bundle supplies them. A month later, with the artefacts installed and the public profile published, the second run shows the score up and the citation rate moving, with the competitor comparison naming who lost ground. That page is the monthly report, and the reason the client renews.