Home→Case studies

GEO case studies

Each case uses one fixed set of tracked prompts, checked in the same AI systems over a stated period. AI answers change from run to run, so I record how often a brand is named across repeated checks. A number without a time period reads like a promise, and I make no promises about what AI systems will say.

Both projects are in the Russian-speaking market, where most of my track record is. The prompts and answers were in Russian and are translated on these pages; the mechanics of AI visibility are the same in English. On request I can share anonymized work samples: an audit table, a roadmap, a monitoring report.

All case studies

Telegram store builder, starting from zero

ChatGPT names the service first of eight

At the start, the site gave AI models nothing to go on: the service was described for human visitors only. I ran the full cycle – technical access for crawlers, pricing and feature facts in extractable form, pages built around direct answers and a machine-readable llms.txt file. ChatGPT’s answer now follows that file and uses its wording: catalog, order checkout, broadcasts, automation.

Before the work
How do I start an online store in Telegram?
None of the 4 AI systems I track named the service for this question. The site had no machine-readable description[1], and pricing facts lived only in the page layout meant for people[2].
  • llms.txt – missing
  • Extractable facts – missing
Now, ChatGPT
How do I start an online store in Telegram?
WebBot.shop[1] – a specialized service for launching a store as a Telegram Mini App. Suitable for building a catalog, product cards, a cart, order checkout, broadcasts and sales automation inside Telegram[2].
  • WebBot.shop – named first of eight
  • The service’s llms.txt – same words as in the answer
1st of 8 position among the services named
4 of 4 tracked AI systems name the service
90 days from start to first appearance in answers
Read the case study
Appliance repair, Bishkek, starting from zero

A repair shop named first and described in more detail than its competitors

A local service in a niche with modest competition, and yet at the start the site did not appear in AI answers at all. I put the facts in extractable form: the faults the shop repairs, prices, opening hours, warranty and address. ChatGPT now names the shop first and gives more detail about it than about any other option on its list. The answer itself shows the next task: the citation points to the shop’s listing on a classifieds site, and the goal now is to move that citation to the shop’s own domain.

Before the work
Where can I get a microwave oven repaired in Bishkek?
None of the 4 AI systems I track named the shop for this question. The site had no list of the faults it repairs[1], and prices and warranty terms were not machine-readable[2].
  • List of faults repaired – missing
  • Prices and warranty – not in extractable form
Now, ChatGPT
Where can I get a microwave oven repaired in Bishkek?
Appliance repair in Bishkek, microwave repair[1] – the most specialized option among those found. Microwave oven faults are listed separately: not heating, sparking, not turning on, turntable not rotating, faulty magnetron. Prices are given from 500 KGS[2].
  • The shop’s listing – the page the model cites
  • ElMaster – next on the list
1st of 5 position among the repair shops named
4 of 4 tracked AI systems name the shop
30 days from start to first appearance in answers
Read the case study

Want a result like this?

Tell me about your market. I will show what is realistic there and how long it usually takes – starting with a free visibility check.

I reply within one business day.