What AI visibility covers
AI visibility is an umbrella for three separate things people tend to blur together. The first is presence: does your brand get named when someone asks an assistant for options in your category. The second is sourcing: does the assistant link to your pages when it answers. The third is accuracy: when it does describe you, does it get your services, location and offer right.
Those are measured differently and fixed differently. Presence usually follows from being mentioned and described across the wider web. Sourcing follows from having pages that can be fetched and reused. Accuracy follows from your own facts being consistent wherever they appear. A report that collapses all three into a single score will tell you that something is wrong without telling you which thing.
Why AI visibility matters
A growing share of the research that used to happen on a results page now happens inside a conversation, particularly for the early questions — what are my options, who does this well, what should I be asking. If the assistant never names you, you are not in that shortlist, and no amount of ranking on the eventual comparison query recovers the ones who never got that far.
The accuracy half matters just as much and gets far less attention. An assistant that confidently tells someone you do not serve their city, or quotes a service you dropped, costs you the enquiry silently. Nobody writes in to correct it. Checking what the assistants say about you is the cheapest part of any AI visibility audit, and it is where most first findings come from.
Where AI visibility goes wrong
The biggest mistake is treating it as a single number to be raised. Assistants answer differently by user, by country, by phrasing and by day, so a headline score is a rough sample dressed up as a metric. It is far more useful as a diary of specific prompts and specific wrong answers than as a chart.
The second mistake is buying tactics sold as shortcuts — pages written to please models, mass-published content, listings bought purely to be scraped. Assistants draw heavily on the same sources search already rewards, so thin material aimed at machines tends to fail in both places at once. The third is ignoring the plumbing: a site that blocks the crawlers, hides its facts in images or renders everything through scripts can be well regarded and still uncitable.
How to act on it
Start by asking. Write down the questions a customer would actually put to an assistant, run them, and save the answers with their sources. You will normally find a mix of absence, competitor dominance and small factual errors, and each of those points at a different job.
Then work on the two things you control. Make the facts about your business consistent and easy to fetch: plain HTML pages, one clear answer per question, the same details on your site and your Google Business Profile. And earn the mentions that put you in the pool at all — real coverage, real listings, work that people describe in their own words. Repeat the prompt check on a schedule so you notice drift, and pair it with citation share once you want a number to trend rather than a list to fix.