How to measure whether AI answers actually mention your business
The hardest part of generative engine optimisation is not the work, it is proving it happened. Clicks from AI answers are under-reported and inconsistent, so measurement has to combine citation tracking with the demand signals you already own.
By Praveen Kumar · Founder, Vectura Digital
Build a prompt set before you change anything
List 20 to 40 questions a buyer would realistically ask an assistant. Record today's answers and who gets cited. That is your baseline, and without it any later claim is guesswork.
Track citations, not rankings
For each prompt, record whether your domain appears, which page is cited, and which competitor is cited instead. Re-run monthly on the same day, with the same wording.
Separate assisted demand from direct clicks
Watch branded search volume, direct traffic and self-reported attribution on your enquiry form. Generative visibility usually shows up there first, weeks before referral traffic moves.
Instrument the referrals you can see
Some assistants do pass a referrer. Tag those sources in analytics so the visible slice is at least counted accurately, and treat it as a floor, not the full picture.
Generative engines cite sources, not slogans.
Review quarterly, not weekly
Answer sets are volatile week to week. A quarterly read on the same prompt set tells you whether the trend is real; a weekly read mostly measures noise.
About the author
Praveen Kumar, Founder, Vectura Digital
Add your certifications and years of experience here before publishing.
Add a two-sentence bio here: what you do, who you do it for, and why your judgement on this topic is trustworthy.