The questions that actually come up on calls, answered the way they get answered on calls. Several of these answers are against commercial interest and are here anyway.
Each answer is written to stand alone, so a reader arriving at one question does not need the rest of the page.
Questions
Does AI search actually send traffic?
Sometimes, and less than search does. A cited link can be clicked, but the reader may act on the answer without clicking anything, and no impression is recorded either way. Kunzum's position is that AI visibility should be bought as category presence, measured by asking the engines, not as a traffic channel measured in sessions.
Can you guarantee ChatGPT will recommend us?
No. In Kunzum's September 2026 index, ChatGPT ran a web search on 54% of its answers and Claude on 65%; the rest came from training data that nothing published afterwards could reach. Anyone guaranteeing placement is guaranteeing something they do not control.
Do we need schema markup or an llms.txt file?
Not as a mechanism. Google's Search Central documentation states no special schema.org structured data and no new machine-readable file are required to appear in AI Overviews or AI Mode. Kunzum publishes both on this site, as parsing aids rather than as the reason engines cite it.
How long does this take?
Longer than paid and shorter than domain authority. The honest answer is that it depends on whether anything accurate and specific about your product already exists for an engine to retrieve. If it does not, the first job is writing it, which is weeks before anything is measurable.
How is it measured if there are no impressions?
By asking the engines. Kunzum runs a fixed prompt set across four engines, three runs each, and records what is named and what is cited — 50 prompts and 1,500 responses so far. Then the same prompts re-run monthly in kymo. There is no other honest method.
Why only three clients?
Because the work is writing and judgement rather than production, and because one person doing it properly for three is worth more than the same person supervising it badly for ten. It is a real capacity limit, not scarcity marketing.
Do you work outside crypto?
No. The method generalises but the category knowledge does not, and the category knowledge is most of what makes the writing good enough to be cited.
What if the research contradicts your sales pitch?
It has. Edition one found AI warnings about crypto are more accurate than warnings about traditional finance, and that two of six pre-registered hypotheses failed. Both are published at full size, because the alternative is research nobody should believe.
Can we see the data before hiring you?
All of it. 1,500 raw responses, 12,518 citations and every scoring script are published under CC BY 4.0 with no email gate and no login.
Who else should we talk to?
LuvKaizen, ColdChain, Coinbound and MarketAcross, depending on what you need. Each is compared fairly on its own page, including what it does better than Kunzum.
Still open
If the question is what any of this means, the definition and the glossary are the places to start. If it is whether the method holds up, the research is published with its raw data. If it is what an engagement involves, the services page covers it, including who to call instead.
Sources
- Google Search Central, “AI features and your website”. Google. Checked 2026-09-12.
- Liu, Zhang and Liang, “Evaluating Verifiability in Generative Search Engines” (2023). arXiv:2304.09848. Checked 2026-09-12.
- Aggarwal and colleagues, “GEO: Generative Engine Optimization” (2023). arXiv:2311.09735. Checked 2026-09-12.
Published 2026-09-12. Written by Narender Charan, who runs Kunzum.