How to track brand visibility in Google AI Mode

The short version

AI Mode is Google’s conversational search surface: you ask, it answers, you follow up. It is not covered by rank tracking and only partly by Search Console, so tracking visibility in it means asking the questions yourself, repeatedly, and recording whether your business was named.

The method is the same one used for assistants. The difference is which surface you point it at, and one detail that makes AI Mode distinctive: the conversation continues, and the second question is usually closer to a decision than the first.

What is AI Mode, and how is it different?

AI Mode is a conversational mode inside Google Search. Instead of a list of links, you get a written answer, and instead of running a new search you ask a follow-up in the same thread.

Compared with the other surfaces:

The practical difference from an AI Overview is the thread. An Overview answers one query. AI Mode answers a conversation, and the businesses named in turn three are frequently not the ones named in turn one.

Why existing tools do not cover it

Rank trackers report a position. There is no position in a conversational answer, so any number reported for one is an invention. This is worth being blunt about because tools are already selling “AI Mode rank tracking”, and the number cannot exist.

Search Console reports clicks and impressions for your site. It does not tell you whether AI Mode named you, which businesses it named instead, or what it cited. A mention without a click leaves no trace there at all.

Analytics sees the visit if there is one. Much of the value here is a mention that never becomes a click: the buyer reads a name, remembers it, and searches for it directly the next day. Your analytics records a direct visit with no referrer and no explanation.

That gap is the whole reason measurement exists as a separate exercise rather than a dashboard widget.

What to track instead of rank

The share of answers that named you. Per question type, asked several times, because answers vary between identical asks. One check is an anecdote.

Who was named instead. More actionable than your own number. It tells you which businesses the surface currently trusts in your category, which is a shortlist of who to study.

What was cited. Where sources are shown, record every domain. That list is a to-do list: those are the pages where your business needs to be described correctly.

How the answer changes on follow-up. This is the part specific to AI Mode. The opener gets you a broad list; the follow-up — which of those is cheapest, which can see me this week — narrows it, and the narrowing is where the sale is. A measurement that only asks openers misses the half of the funnel closest to a decision.

Whether the answer left a reason to click. An answer that fully resolves the question is a lost visit even when it names you. Recording this changes how you read a flat traffic line.

How to build the question set

From how buyers ask, not from a keyword export. People type keywords into search boxes and sentences into conversational surfaces, and the phrasing carries the intent.

Group by what the buyer wants:

• general recommendation — who should I use for X in this city
• specific service — who does Y
• price — how much does Y cost, who is affordable
• urgency — who can see me today, who is open now
• trust — who is reputable, what do people say
• comparison — X or Y, which suits my situation

Then write the follow-ups. For each opener, the one or two questions a real buyer would ask next. This doubles the size of the set and roughly doubles its usefulness.

We write ours from the client’s own site and city, and the full method is published here.

Doing it by hand versus continuously

By hand is worth doing once, today. Open AI Mode, ask your five most valuable buying questions, ask a follow-up to each, and write down who was named. You will learn more in twenty minutes than from any article, including this one.

What you should expect to find: a shorter list of names than you assumed, at least one competitor you had not been watching, and at least one answer where Google declines to recommend anyone specific.

What you cannot do by hand is the repetition. Answers drift several points without anything changing, so telling progress from noise needs the same questions asked the same way on a schedule, with the raw answers kept.

What you also cannot do by hand is stay honest about it. Everyone remembers the answer that named them and forgets the three that did not. A recorded measurement does not have that problem.

What a measurement should record

For each question, each repeat, each surface:

1. The exact question asked.
2. The full answer text, kept verbatim.
3. Every business named, in order.
4. Every source cited.
5. The date and the surface.

Point two is the one that matters most in practice. If a percentage cannot be traced back to the text that produced it, it is not checkable, and an unverifiable number invites an argument you cannot win — usually with a client who opened AI Mode once and saw something different.

How AI Mode relates to the other surfaces

They move independently and should never be combined into one figure.

When four assistants were asked the same questions in one market, they named 63 businesses between them. Only four were named by all four. Forty-six appeared in a single assistant only. In the same measurement, one business ranged from 100% of answers on one assistant to 50% on another.

Google’s surfaces are not exempt from this. AI Mode, AI Overviews and ordinary results draw on related but distinct retrieval, and a business strong in one can be absent from another.

The practical rule: one number per surface, always. An average across surfaces describes none of them and gives you nothing to act on.

What moves visibility in AI Mode

The same levers as the rest of AI search, in roughly the same order.

How your business is described in the sources it cites. Where citations are visible, this is not guesswork: the list is on screen.

Consistency across those sources. Conversational answers are assembled from several places at once, and contradictions between them are a reason to reach for a different business.

Whether your pages answer the buying question directly. Retrieval favours passages that stand alone. A page whose answer arrives in paragraph four does not supply one.

Coverage of the follow-up questions. This is the AI Mode specific one. If your pages answer the opener but nothing that comes after it, you appear in the broad answer and disappear at the point of decision.

The full ordered version of this work is in how to improve AI visibility.

A worked example

From a measurement on one local market, business anonymised.

Openers. Named in 38% of general recommendation answers.

Follow-ups. Asked which of those is most affordable, the business disappeared entirely — 0% across every repeat. The two competitors that remained both published prices on their own sites and on two directories.

What the citations showed. For the price follow-up, four sources: two directories, a review platform, a regional roundup. The business appeared on one, with no prices anywhere.

What was done. Prices published on the business’s own page, answered in the first two sentences rather than in a table lower down, and the same ranges added to all four listings.

What moved. The opener share shifted four points and was reported as no change. The price follow-up went from zero to roughly a third.

What was not claimed. Causation. The report said what changed, when, and what moved after, and noted that the movement was concentrated in the question group the work targeted.

Frequently asked

Can Search Console show me AI Mode performance? Not usefully. You may see impressions, but not whether you were named or who else was.

Is AI Mode the same as an AI Overview? No. An Overview sits above ordinary results for a single query; AI Mode is a conversation. They cite differently and they move independently.

How often should I measure? Often enough to distinguish signal from drift and rarely enough that the work has time to land. Every two weeks suits most businesses; monthly is fine for stable categories.

What counts as real movement? Over about five points. Below that is drift, and we label it as no change rather than dressing it up.

Can anyone guarantee I will be named? No. The lists are not published, not stable and not for sale.

A recording template

Whatever you use — a spreadsheet is fine — record these columns. Anything less and you cannot answer the questions you will have in six weeks.

The last column is the one people drop first and regret. A share you cannot trace to the text behind it is a number you cannot defend the first time a colleague opens AI Mode and sees something different.

How often to measure

Often enough to distinguish signal from drift, rarely enough that work has time to land.

• Weekly is too often for most businesses: you will be reading noise.
• Every two weeks suits a market where you are actively working on sources.
• Monthly is fine for a stable category with no active programme.
• After a change, wait four weeks, not two. Corrected listings take time to be re-read.

The one exception is a big visible change on the competitor side — a new entrant, a competitor’s rebrand — which is worth measuring immediately, because that is when the names in an answer move most.

Does AI Mode use the same sources as an AI Overview? Overlapping, not identical, and the conversation changes what gets retrieved on each turn.

Should I measure AI Mode separately from ordinary Google? Yes. They answer different questions about the same business, and combining them produces a number that describes neither.

Where to start

Ask your own questions by hand today — openers and follow-ups. Then decide whether the picture is worth measuring properly.

The free check runs a short measurement on your business and emails you what came back, with no call attached. The sample report is open in full if you want to see the output first, and the method explains how the questions are written and why every figure is reported per surface.

One thing to do tomorrow

Ask one buying question in AI Mode, then ask the follow-up a real customer would ask next. Note whether the same businesses appear in both answers. In most markets they do not, and the second list is the one closer to a decision — which is exactly the half most measurement leaves out.