AI Visibility Management

Who it is for

AI Visibility Management fits a business whose customers ask an assistant before they ask you, and whose website is where the answer should come from.

• Customers choose you through search and AI answers, and nobody can say this week whether ChatGPT, Claude, Gemini or Perplexity names you.
• You are a regulated business and a misstated fee, rate or eligibility rule in an AI answer is a problem the week it appears.
• You have an SEO agency for content and want the measurement and the site engineering held to the same record.
• You have a booking, application or enquiry flow and want to know whether an AI agent can get through it.

What is included each month

The service reads twenty-one sources on a schedule and keeps every row. Search Console, Analytics, Google Ads search terms, Business Profile, Bing Webmaster Tools, Core Web Vitals and the crawl sit beside the AI answers, so the question of whether a change brought anyone in has an answer in the same place.

• AI answers, weekly. Twenty-four questions put to ChatGPT, Claude, Gemini and Perplexity. Named, cited, seen and ignored, and the sources each assistant read.
• Google and Bing. Rank for up to sixty target terms, AI Overview presence and citations, Maps results, the brand results page, and the queries and pages from both consoles.
• What it brought in. Sessions and key events by channel including AI referrers, paid search terms with conversions, and Business Profile calls and bookings.
• The assistant test. Twelve customer tasks run against your site by an AI agent each week, ranked against up to twenty competitors, with the transcript of every failed run.
• Site health. Index coverage, Core Web Vitals, crawl issues, how much of each page an assistant can read, and inbound links that land on missing pages.
• Findings and actions. Each finding carries a stable key and a severity. The action list ranks each finding by the gain we expect from fixing it, and every row links to the evidence behind it.
• Fortnightly review. A senior engineer reads the record, writes the findings and raises the tickets in your Jira, Asana or GitHub.
• Two engineering days. Sixteen hours of senior engineering a month on your codebase to clear the actions we own. Unused hours roll into the following month.

What the record looks like

Every page of the dashboard is rebuilt from the stored rows, so any figure can be traced to the day it was read. The row that matters most is "seen, unused": the assistant's search returned your page and the answer did not use it. That is a page problem, and page problems are engineering work.

A fertility clinic shows how this works. On "which IVF clinic on the Gold Coast should I choose", the assistants read Google reviews and Reddit, which are sources an AHPRA-regulated clinic cannot quote in its own advertising, so the action is to be present and accurate where those sources form. On "IVF success rate for a 38-year-old at this clinic", the assistants answer from the government's YourIVFSuccess data, and one quotes a clinic-level rate the clinic never published; that is a statement about the clinic made in its absence, and the weekly record is where it shows up. On "cost of a first IVF cycle after Medicare", two assistants find the clinic's fee page and do not use it, which the extractability reading traces to a PDF fee schedule inside a JavaScript tab. The fee page is one engineering day.

How a week runs

• Every day at 05:00. Search Console, Analytics, Ads, Business Profile, Bing, the crawl log and the lead term of each target are read and stored.
• Monday. The AI answers, the assistant test, the full term set, Core Web Vitals and the competitor standings run. The dashboard rebuilds.
• Tuesday. The weekly note lands in your inbox: the week's findings, what is claimed by an open action, and what shipped and is still firing.
• Every second week. The review. Findings become tickets with the evidence attached, and the engineering days are planned against the list.

Measurement where accuracy is the first test

A clinic, a lender, an insurer or a school is judged on accuracy before it is judged on reach, and each works under rules about what it may say about itself. The record keeps the full text of every answer, so a misstated fee, success rate, eligibility rule or licence can be found the week it appears, with the source the assistant took it from. The action that follows is usually a page change or a listing correction, and both are in scope.

Working alongside your agency

The service measures and engineers. Content, link acquisition, paid media and PR stay with whoever does them today, and the action list gives them the evidence for their rows. The dashboard has a view per owner, so each party sees its own work and the whole list.