How to measure your brand's presence in AI answers
AI answers have no impressions log and no click, so brand presence has to be sampled by asking the engines. Here is the method, and what it cannot tell you.
Search advertising gave you impressions, clicks and a placement report. AI answers give you none of the three, so measuring brand presence means asking the engines the questions your buyers ask and recording what comes back.
How do I measure whether my brand appears in AI search?
Run a fixed list of buyer questions against each engine on a schedule and record whether you were named. Measurement is a sampling exercise rather than a log lookup, because no engine publishes a report of the answers it gave other people.
Which questions belong in the list?
The questions a buyer would actually type, phrased the way they would type them, including the comparisons. “Best tool for X”, “is X worth it”, “X versus a named competitor” and “how do I do Y” each pull different sources, and a list containing only your category name will tell you almost nothing about purchase intent.
Why does the same question need asking repeatedly?
Answers are non-deterministic. The same question can name three different brands across three runs, so a single run measures luck rather than position. Repeated runs over a stable list, across engines, produce a rate that can be compared over time, and that rate is what share of answer means.
What should I record from each answer?
Three columns: whether you were named, which sources the engine cited, and which claim about you it repeated. The cited sources matter most, because those pages are the material shaping the answer, and they are where influence can be bought or earned.
Does being crawled mean I appeared?
No. A crawl means an engine read a page; surfacing means the engine used what it read. Many pages are fetched and never used, so counting crawls overstates reach considerably and makes a campaign look better than it was.
How is a paid placement traced back to an answer?
Each placement carries a unique traceable marker, and the engines are queried independently to check whether that marker turned up in an answer. Attribution works by matching what the engine said against what was placed, so it requires no cookie, no click and no user identity.
Can I control which websites my sponsored content appears on?
You buy a question rather than a URL. Placements land on publisher pages that already answer the questions you chose and are matched to each page’s actual topic, so the control available is topical rather than a site list you pick from. Every page in the network belongs to a publisher who installed the integration on their own domain, which makes the inventory a known set of sites rather than an open exchange. The advertiser overview covers what targeting looks like in practice.
How are AI-referred visits used in the pilot?
They are a secondary measurement signal, never a delivery channel. A human visitor receives the unchanged publisher response, while an exact known referrer host may contribute a directional analytics event. The header can be forged or removed, so it cannot enter an auction, trigger spend or prove that a placement caused the visit. The AI referral signal entry explains how to interpret that evidence without turning correlation into attribution.
How often should the measurement run?
On a fixed schedule rather than after every change you make. Engine answers move on their own as the underlying pages change, so a measurement taken only when you shipped something cannot separate your effect from the drift, and the comparison you wanted is lost.