For a business planning its next page, the useful question is what extra help can we give this reader? Below are the observed source patterns, examples to build on and the method behind this same-day sample.
What I checked
The sample contained nine US-English searches and five Turkey-Turkish searches, collected between 06:22:22 and 06:24:18 UTC. These were desktop, country-level results obtained through DataForSEO, not searches from a named city or a personalized browser session.
| Question group | Searches | Examples from the panel |
|---|---|---|
| Website and local SEO costs | 4 | small business website cost; web sitesi fiyatları; local seo cost |
| Starting a website or keeping a profile alone | 2 | do i need a website if i have a google business profile; web sitesi yaptırmak istiyorum |
| Visibility and inquiry problems | 3 | why is my business not showing up on google maps; website traffic but no leads |
| Getting a business recommended by ChatGPT | 2 | how to get your business on chatgpt; chatgpt işletme önerileri nasıl görünürüm |
| Sector-specific website or SEO questions | 2 | plumber website design; seo for cleaning business |
| Service-area pages | 1 | service area pages seo |
The grouping above is my editorial classification. The published CSV contains 126 source-link records, 125 distinct URL strings and 100 hostnames after removing the www prefix. A link can recur across searches, and different pages can share a hostname. These counts describe the captured panel, not market-wide citation shares.
Download the question, date and citation records as CSV to inspect the observations. The public GitHub repository also includes the method, data dictionary and reuse terms.
The source mix changed with the question
For website costs, the captured answer linked to provider cost guides and site-builder publishers. One example was GruffyGoat’s small-business website cost guide. For the US plumbing-design query, the cited set instead included template libraries, design galleries, video and a discussion thread. An image-led question did not produce the same source mix as a budgeting question.
The Turkish Maps question cited Google’s own visibility help, alongside an agency guide and a community discussion. The service-area query linked to specialist explanations, including BrightLocal’s service-area page guide, as well as forums and video.
A citation is evidence of a link appearing in that answer. It is not Google's approval of every statement on the linked page. I kept ordinary organic results separate: a page ranking below the Overview was not counted as a citation unless it also appeared in the answer’s recorded source links.
Before turning local AI-search observations into an AI search optimization plan, use the AI search visibility measurement guide to define prompts, locations, repeat runs, and what counts as a linked citation. Pair that log with a local SEO audit to check business information and page accessibility separately from changes in generated answers.
Useful patterns to borrow—and improve
Cost answers separated building methods, upfront fees and continuing expenses. A better page can go further by comparing the same scope, naming who writes the content and showing renewal costs separately. A large price range alone does not help an owner compare two proposals.
Problem answers organized possible causes and checks. For Maps, the useful distinction is whether the profile is missing, has a management warning, or simply ranks poorly for a general service search. Those situations need different next steps.
Design and service questions benefited from concrete examples. Show a working demonstration, label what is editable and explain the mobile contact path.
| Question to answer | Something useful to add to the page |
|---|---|
| What will the website cost? | The same page and content scope, with first-year and renewal costs separated |
| Why is my Maps profile missing? | A route that distinguishes verification, suspension and local ranking |
| What should my service website look like? | A working mobile example and an explanation of how to request a quote |
These are practical page experiments suggested by the observations. Record when they go live, then revisit the same questions on later dates.
Two claims I would not repeat as rules
In the profile-versus-website answer, a linked website with extensive schema was presented as a local ranking advantage. That is too broad to turn into a promise. Google describes local ranking through relevance, distance and prominence; its official explanation does not promise a ranking boost from adding a particular schema format.
Some pages reviewed for AI visibility blurred search access and training access, or treated special files and publishing frequency as requirements. OpenAI documents OAI-SearchBot and GPTBot separately. Google’s AI optimization guide does not require an AI-specific schema, an llms.txt file or a fixed paragraph length.
The practical response is to check the technical claim against the platform’s documentation before building it into a client recommendation.
Method and scope
This is a selected, same-day desktop sample: all 14 captured searches returned an AI Overview. Results were country-level, with nine US-English and five Turkey-Turkish questions. The two questions about ChatGPT were also searched on Google.
I counted numbered source links in the returned answer text and kept ordinary organic links separate. URL counts use the recorded strings; hostnames omit www. Redirect destinations were not exhaustively resolved. The sample covers recorded Overview links, rather than all Google interface elements or hidden fan-out queries.
No recorded link in this panel pointed to umitkaratay.com. The study does not estimate market-wide citation frequency or measure traffic, leads or the effect of subsequent page changes.
How I would turn the findings into a page test
Choose one buying question your business can answer well. Save the answer and actual source links, read the relevant pages, and identify something useful you can add: a worked cost example, a real service limitation, a documented process or an original comparison.
After publishing, repeat the same questions under comparable conditions on separate dates. Keep unsuccessful captures separate from answers with no citation. Record mentions, clickable sources and identifiable visits as different outcomes.
That repeat-date stage has not been completed in this study. The AI visibility measurement guide explains how to keep the records comparable. If you need the research and page work done for your business, see the defined AI-search support scope.
This is a dated source study, not an audit of your business. If you want research on your own market and offer, the AI audit scope guide explains the evidence and recommendations to expect.