How GardenPark made it into the body of the Google AI Overview answer on landscape design in Kyiv

01Context and task

gardenpark.com.ua is a landscape design and grounds company (Kyiv and region, in business since 2017): a team of 9, more than 100 completed projects. The niche is competitive and local: dozens of studios, with «Новий Сад», «Украфлора», «Магія Саду» and others in the results.

The initial GEO audit (June 2026) scored 39/100. The weak spots are typical for the niche. The portfolio is photo galleries without facts (location, area, year), and the About page gives generic text with no names or founding year; structured data is nearly absent. To a human the site looked good; to a machine it was almost empty.

Task: make AI engines and Google understand who GardenPark is and take the company as a source for their answer in the Kyiv landscape design niche.

02The bet

Instead of rewriting copy to look nicer, we built what a machine reads: markup, facts about real projects, transparent prices and timelines, a verifiable team.

03What we actually did

We rebuilt the site’s machine layer for identification and quotability — all through an mu-plugin, without touching the design:

• Organisation identity: Yoast Organization upgraded to [Organization, LandscapingBusiness] — NAP (address, phone, email), areaServed “Kyiv + region”, sameAs (Facebook / Instagram / YouTube + the Google entity), foundingDate 2017, founder, 9 employees, 10 areas of expertise (knowsAbout).
• Services: Service schema on 36 service pages (both languages) with prices where they’re fixed (e.g. design — UAH 6,000 per sotka via UnitPriceSpecification).
• Questions and answers: a visible FAQ block plus FAQPage markup on 28 services (home + 27 services, RU and UK) — what’s visible is what’s in the markup.
• Portfolio as evidence: CreativeWork schema on ~44 projects plus 12 case studies with facts (location, year, area, style, works) from the client questionnaire — so that instead of “just photos” AI sees specific projects.
• People: Person schema for four named experts (role, education, experience, social profiles) plus photos — real E-E-A-T instead of an anonymous “we’re a team of professionals”.
• Hygiene for machines: llms.txt, a consistent hreflang with x-default, clean meta descriptions (emoji and caps removed), security headers. Performance: page caching (TTFB ~1s → ~70ms) and local WebP conversion (−65% image weight).

We don’t claim a single causal factor: the result below comes from the combination of markup, a confirmed Google entity, and substantive content.

04Evidence

The main piece of evidence — GardenPark in the body of the answer Google AI Overview, not just in the list of blue links. Two real screenshots of live results (Google, Ukrainian, Kyiv region, July 2026):

Google AI Overview · query: “де замовити ландшафтний дизайн в києві” · July 2026

A commercial query. In the body: “…популярними є Новий Сад, Украфлора та GardenPark”; a separate item “GardenPark — ландшафтний дизайн та благоустрій ділянок під ключ” plus a gardenpark.com.ua source card.

Google AI Overview · query: “ландшафтний дизайн території” · July 2026

An informational query. gardenpark.com.ua is the source of the definition in the answer body (the “gardenpark.com.ua +2” chip) plus a source card in the panel.

Google Knowledge Graph — the brand as an entity

In parallel, GardenPark is present in Google Knowledge Graph (the machine entity identifier kgmid /g/11g81srdht, name “GardenPark — Ландшафтний дизайн”). It’s the same entity we referenced in the sameAs of the organisation markup — so the site and the Google entity are stitched together. For AI search that’s a confirmed “who this is”, not a guess.

05Other engines — the honest picture

Google AIO isn’t the only engine. For the same commercial query “де замовити ландшафтний дизайн в києві”, GardenPark is also named in the answer body by ChatGPT:

ChatGPT · query: “де замовити ландшафтний дизайн в києві”

In the list of recommended companies: “GardenPark — Ландшафтний дизайн — повний цикл робіт від концепції до озеленення” plus a GardenPark source chip. Screenshot of the ChatGPT session as provided.

Perplexity — still a growth area. We ran seven niche queries through Perplexity (the sonarmodel, API): the site lands in the source pool for 2 of 7. That’s a weaker signal than Google AIO or ChatGPT and, honestly, a work front rather than a finished result.

The column shows the site’s place among the sources pulled by the API (sonar), not its position in the app’s answer body. Snapshot of 31.07.2026; the set of sources changes between sessions. ChatGPT was measured separately (above — in the body); Gemini wasn’t measured in this snapshot.

06Limitations

• The case focuses on presence in the AI answer plus machine identity, not traffic. Google Search Console figures aren’t included in this snapshot; when we add them we’ll show the trend separately.
• Perplexity is still weak (2 of 7 in sources, not in the answer body). That’s the next front, not a finished result.
• Personalisation. The set of sources in AI Overview and Perplexity changes with region, language and session — these are snapshots, not a constant.
• Attribution. We don’t attribute everything to one factor; it’s the combination of markup, the Knowledge Graph entity, and substantive content.
• The 39/100 start is the audit’s baseline score, a reference point rather than a result.

07Conclusion

In a niche where competitors show galleries without facts, GardenPark became the company Google AI Overview and ChatGPT name in the body of the answer — for the informational query and for the commercial one, where the user is directly looking for someone to hire. What worked wasn’t looks but machine identity: a confirmed Knowledge Graph entity, organisation/services/FAQ/portfolio markup, and substantive facts. Being present inside the AI answer is a discipline separate from classic rankings, and in a local, YMYL-like niche it delivers visibility where the customer is looking today.