Can buyers find Kata through Google and AI search?
We are testing this on our own business. Here is where we started, how we checked, and what we still need to prove.
Our starting point
Kata Automation builds AI growth systems for service businesses. We help with lead generation, enquiry follow-up and sales conversion. For this experiment, Kata is the business being measured.
We saved the starting evidence on 7–8 September 2026. These are baseline findings. This page does not yet report a gain from our changes.
| Check | What we found |
|---|---|
| Search Console | 3 clicks and 19 impressions in the available daily rows, 20 August to 5 September. We had selected 28 days, but only 17 daily rows were available. |
| Google's AI report | 5 impressions, all on 21 August. These are included in the 19 above, so we do not add them again. |
| Searches for Kata | Our homepage was the first ordinary result in both branded searches we checked. |
| Searches for a provider | Kata was absent from the first page in all four unbranded searches we checked. |
| Indexing | The homepage and enquiry-response service page were indexed. The try-it page was discovered but not indexed. |
Source: Kata's saved Search Console tables, page checks and Google results. We searched on desktop Google from Denmark. These checks do not show our rank across the US. The search-term table was empty. We could not tell which terms led to the clicks.
The starting problem is clear within these checks: Google could find the site, but buyers searching for a type of provider did not see Kata on the first page.
What the AI answers showed
We used four fixed prompts that asked about providers without naming Kata. These are our unbranded checks. We ran each prompt three times in each app below.
| Surface | Kata mentions |
|---|---|
| Google AI Mode | 0 of 12 answers |
| ChatGPT | 0 of 12 answers |
| Gemini | 0 of 12 answers |
This is a small set of observed answers, not an AI search ranking or a share of the whole market.
We also ran two prompts that named Kata, three times each. Google AI Mode linked to our homepage in all six. Gemini described our broad AI growth work. ChatGPT did not consistently connect the name with that work.
When we gave ChatGPT the homepage address directly, it could read and cite the service descriptions. That was a separate check. Finding a page when given its address is different from recommending it to a buyer.
Claude's seven checks stay separate: an installed business skill affected one answer. They do not give us a neutral buyer baseline. Bing measurement was unavailable after setup failed twice. Unavailable does not mean zero.
How we keep the comparison fair
To run your own check, use our AI search visibility guide. It includes the exact prompts, a results checklist and a way to choose the next fix.
- Keep the same six prompts: two name Kata and four do not. Run each one three times in new chats where we can. Record the app used.
- Save the full answers, source links, dates, settings and exclusions. Keep private account records off this public page.
- Track a name mention, a linked citation, a site visit and an enquiry as separate events.
- Record every change and its release date before judging the later numbers.
The prompts ask who can help with AI growth, new leads, follow-up and faster replies. One asks about US businesses. Our location was Denmark or was not shown. Asking about the US does not mean we ran the test there.
Answers can vary by account, settings, place and date. A selected search control does not prove a search ran. Some source chips did not expose a link in our capture. Our own searches may also add later Google impressions.
We follow Google's definitions for its search reports. We keep Google's AI report separate from the answers observed inside each AI app.
We could not verify how many visits or suitable enquiries came from organic or AI search. Those totals remain unknown. Rankings alone do not prove that we won new work.
9 September: the first improvements are live
We published the lead follow-up worksheet and this baseline journal, then expanded our enquiry-response page. We also added pages for managed lead generation and proposal automation, with links from the homepage.
We made small text easier to read, added image dimensions, and made missing addresses return a proper page-not-found response. Company information now includes our official business registration.
We are continuing useful improvements together. We will compare the combined result with the saved baseline, rather than wait to measure each small change on its own. The dated change record will help explain what happened, but it will not prove which change caused a result.
Our next indexing check is 16 September. Outcome reviews are planned for 7 October and 4 November. No ranking, citation or enquiry gain has been measured from these changes yet.
What we will test next
We think clear service pages, useful tools and proof of our work may help buyers and search systems understand Kata. We need to test that idea.
The first content and service pages are live. Next we will build stronger examples from our own work, improve how we measure search visits and enquiries, and add dated results as they become available.
After each approved release, we will check that the pages work and record what shipped. We plan to check indexing after seven days and review outcomes after 28 and 56 days. Those are review windows, not promises of a ranking gain.
If several things change together, we will not claim one caused the result. If a test fails, the record should show that too.