Diagnose

Measurement tells you that you lost. This chapter is about finding out why, which is where most of the value in this field actually sits and where most programmes stop.

Start with the citations

Most AI visibility work involves inference: you lose an answer, form a theory about why, act on the theory. The citations remove that step, and almost nobody reads them.

Take the answers you lose, collect every cited domain, and count. The distribution is usually lopsided. A small number of sites, often five to ten, supply most of the answers in a given category. That tally is your citation map, and it is the closest thing this field produces to a work order.

It changes the question from “how do we become more visible”, which has no first step, to “are we on these seven sites”, which does.

Perplexity is the best place to build one, because it cites every answer and shows its sources. What you learn there transfers usefully to engines that retrieve but do not show you everything, which is the argument in the Perplexity chapter and in the sources AI engines trust.

Three constraints, three different fixes

Nearly every case of “we are invisible” reduces to one of three things. They look identical in a score and are straightforward to tell apart once you read the answers.

You are not resolvable

The engine cannot confidently work out who you are. Your name is ambiguous, your details differ between your site and your listings, or nothing on your site plainly states what you do and where.

How to tell: ask an engine directly about your business by name and location. If it hedges, describes a different company, confuses two locations or gets your category wrong, this is your constraint. Worth checking even when the answer looks right, because engines produce fluent wrong answers confidently.

This is the cheapest thing to fix and the most commonly broken, particularly after a rebrand, a move, a merger, or opening a second location.

You are absent from the cited sources

The engine knows who you are, but the material it draws on for your category’s questions does not mention you.

How to tell: your citation map. If the same domains keep supplying the answers you lose and you appear on none of them, this is it. In our own citation data this is the constraint for most local and service brands: a small number of sources they are missing from, not a shortage of pages on their own site.

You are cited but not preferred

You appear in the sources and a competitor gets named in the text.

How to tell: you show up in citations but not in the answer, or you are listed last, or mentioned with a qualifier while a competitor gets a recommendation.

This is the best of the three to have. You are already inside the retrieval set, so the problem is substance rather than presence, and it is the one case where content is unambiguously the right lever.

Who is winning, and why that is the better question

Absence tells you little. Who was named instead tells you a lot: the answer had room for a brand like yours and chose someone else, and the citations show what that choice was made from.

Run the comparison deliberately rather than incidentally. Is AI recommending you or your competitors covers how to read this without drawing conclusions from single answers, which as chapter 2 covered is the standing hazard.

Auditing your own pages

Once the citation map says your own pages are the problem, and only then, it is worth auditing them.

Three questions matter more than a checklist. Can an engine fetch the page at all, meaning your robots rules do not block AI crawlers you actually want reading you. Does the page answer the question near the top, since a retriever taking one passage will take an early clear one over a buried one. And is the claim self-contained, because a sentence that only makes sense after three paragraphs of setup is hard to lift and tends not to get lifted.

That is ordinary writing discipline rather than a trick. Pages that state things plainly and early do well; pages built to hold attention before delivering the point do badly. The AI visibility audit sets out a four-layer version of this.

The order matters

The reason to separate the three constraints is that work on the wrong one produces nothing, and it is expensive to discover that slowly.

Publishing more content when engines cannot resolve who you are changes nothing. Fixing your schema when the cited sources in your category simply omit you changes nothing. The most common failure in this field is skipping diagnosis entirely and going straight to publishing, because publishing feels like progress.

The next chapter takes a diagnosis and turns it into work, in the order that actually moves the number.

Frequently asked questions

Why does AI recommend my competitor instead of me?

Usually because the sources those answers draw on include them and not you. Read the citations under the answers you lose and tally the domains. If the same few sites appear repeatedly and you have no presence on any of them, that is the mechanism, and it points at a finite task list rather than a vague content strategy.

How do I find out why I am not in AI answers?

Separate three constraints. Either engines cannot confidently tell who you are, or the sources they cite in your category do not mention you, or they do cite you but name someone else. Each has a different fix, and work on the wrong one produces no movement at all.

What is a citation map?

A tally of every domain cited across the answers in your category, ranked by how often it appears. It shows which sources actually decide your market, and being absent from the top few is the most common reason a brand is invisible.

How can I tell if my brand is the problem rather than my content?

Ask an engine directly about your business by name and location. If it hedges, describes a different company, mixes up locations or gets your category wrong, you have an identity problem, and no amount of publishing will fix it until that is resolved.

Does being cited mean I am recommended?

No, and the difference is useful. Being cited means a page of yours was used as a source. Being named means your brand appears in the answer a person reads. Cited but not named is an encouraging position with a specific fix: you are already in the retrieval set, so what needs work is the substance, not the presence.