How I Got eviacharge.pl to Rank in AI Search Answers: A Real GEO Case Study

How I Got eviacharge.pl to Rank in AI Search Answers: A Real GEO Case Study

Last updated: 31 July 2026 | Author: Konrad Kluz | Category: Case Studies

In 8 weeks, eviacharge.pl went from zero AI visibility to the number 1 position in ChatGPT for its target query. Here is exactly what I did, in my own words, as the person who ran the project.

eviacharge.pl is an EV charging installation business based in Poland, run by me, Konrad Kluz, alongside my consultancy Geovise. I built the project for a specific reason: I wanted to test GEO optimization on a real, operating business and see the results before recommending it to other companies. This is that report.

The Problem: Solid Website, Zero AI Visibility

When I set up eviacharge.pl, the website had solid technical foundations. I was happy with the content quality and the code structure. Even so, when I typed "who installs EV chargers in Warsaw" into ChatGPT, eviacharge.pl simply did not exist in the results.

This is a typical AI visibility problem for small and mid sized businesses. Most owners do not realize that AI search and Google search are two separate ranking systems that require separate signals.

Before I started GEO work, the picture looked like this:

• Google ranking: around position 50 (page 5, effectively invisible)
• AI citations across ChatGPT, Perplexity, and Gemini: 0
• Structured data (schema.org): none beyond basic product markup
• robots.txt actively blocking GPTBot and PerplexityBot
• Brand name inconsistency across pages ("eviacharge", "Evia Charge", and "EVIACHARGE" all in use)
• No author or Person schema anywhere on the site

eviacharge.pl was not hostile to AI. It was simply invisible to it. That is the most common state for SMB websites: not penalized, just unconfigured.

The Audit: 5 Gaps in 48 Hours

At Geovise, I run a structured GEO audit that checks a site against five pillars AI models use when deciding what to cite: authority signals, entity clarity, structured data, crawlability, and answer ready content.

For eviacharge.pl, the audit turned up five concrete gaps within two working days:

Gap 1: Crawler access blocked. robots.txt was using a legacy deny all policy inherited from a previous developer. GPTBot, ClaudeBot, PerplexityBot, and GoogleExtendedBot were all blocked. AI models cannot cite content they cannot read.

Gap 2: No entity consistency. The brand appeared in at least four different written forms across the site and in third party mentions. AI models build a knowledge graph of entities, and inconsistency creates ambiguity that reduces citation confidence.

Gap 3: No structured data for FAQs. The site had a dedicated FAQ page, but it was unstructured plain HTML. FAQPage schema was absent across every page. This is one of the highest value GEO signals: 76% of content cited by AI models contains structured list or Q&A content.

Gap 4: No author or organization schema. AI models weight content higher when they can resolve the author as a known entity with verifiable credentials. There was no Person schema, no Organization schema with a verified domain, and no author bio on product or article pages.

Gap 5: FAQ questions written for Google, not AI. The existing FAQ used short, keyword style questions such as "EV charger price Poland." AI models process natural language queries such as "How much does it cost to install a home EV charger in Poland?" The content needed rewriting for conversational intent.

What I Changed: The GEO Implementation

The implementation ran across 8 weeks in three sprints. Here is the complete list of changes, in the order I made them.

Sprint 1: Access and Crawlability (Weeks 1 and 2)

• robots.txt update: Removed the blanket deny rules. Added explicit allow directives for GPTBot, ClaudeBot, PerplexityBot, GoogleExtendedBot, and anthropic-ai. Kept blocks only for archiving bots that are not relevant to AI search.
• llms.txt creation: Wrote and deployed a root level llms.txt file following the emerging standard. It declares the site's purpose, primary entity (eviacharge.pl), key product categories, and the preferred language for AI summarization.
• Sitemap verification: Confirmed the sitemap was accessible and submitted it to all major search consoles, including Bing, which feeds Copilot.

Sprint 2: Entity and Schema (Weeks 3 and 4)

• Entity consistency audit: Standardized the brand name to eviacharge.pl across all on site text, meta tags, alt text, and Open Graph fields. Updated 34 individual instances across 12 pages.
• FAQPage schema implementation: Added FAQPage structured data with 8 conversational questions to the homepage, the main product category page, and the FAQ page itself.
• Organization schema: Deployed schema.org Organization markup site wide, including legalName, url, sameAs links to the verified Google Business Profile and LinkedIn page, and foundingDate.
• Person schema for the founder: Added schema.org Person markup for me as founder on the About page: name, job title, url, and a sameAs link to LinkedIn.
• Author bio structured data: Added visible author attribution with structured markup to every blog post and guide page: name, role, and a one sentence credential statement.

Sprint 3: Content and Internal Linking (Weeks 5 and 6)

• Conversational FAQ rewrite: Rewrote 24 FAQ items across 6 pages. Each answer now opens with a direct, citation ready summary sentence, followed by supporting detail. This is the structure AI models extract when building responses.
• Internal linking for topic authority: Built a topic cluster around home EV charging in Poland, with the main product page as the hub, supported by 4 spoke pages: installation guide, comparison of charger types, grid connection requirements, and government subsidies.
• Hreflang implementation: Added correct hreflang tags for the pl-PL and en-GB versions of the key product pages.

Results: The Numbers After 8 Weeks

I tracked results for 90 days after the implementation ended. Here is what happened.

GEO results after 8 weeks of implementation

The first ChatGPT citation appeared in week 4 of the project. That is faster than the typical 4 to 8 week window I usually see when solid technical foundations are already in place.

For comparison, traditional SEO produces measurable Google results after 3 to 6 months. A business can show up in ChatGPT faster than most people expect, provided GEO is implemented methodically.

The results are verifiable. eviacharge.pl is a live website. You can check the ranking yourself by asking ChatGPT to find companies installing wallboxes in Warsaw, or by trying the queries below in Perplexity.

The query "najlepsza ładowarka EV Polska" now returns eviacharge.pl as a cited source in Perplexity's answer panel. You can verify this yourself: open Perplexity, type the query, and check the citations column on the right.

What the Data Shows

Three patterns emerged from this project that I now treat as validated GEO principles.

1. Crawler access is table stakes. Every blocked bot is a citation you will never receive. Fixing robots.txt alone produced the first indirect AI visibility signals within 10 days, before any schema or content work was complete.
2. FAQPage schema is the highest leverage single change. Of everything I implemented, FAQPage schema correlated most directly with AI citations. It gives AI models a pre formatted, extractable answer instead of forcing them to infer structure from prose.
3. Non English GEO works. Polish language queries are underserved by most AI optimization practitioners. eviacharge.pl now has citations that no English language competitor can replicate, because the queries are in Polish and the content is authoritative in that language.

What This Means for Your Business

If you are an SMB owner or marketing manager reading this, here is the practical summary:

• GEO is not a replacement for SEO. eviacharge.pl had organic rankings before I started. GEO amplifies an existing foundation. It does not substitute for one.
• The technical barriers are low. Five of the eight changes I made required no new content, only configuration: robots.txt, llms.txt, schema markup, entity standardization. A competent developer can implement them in under a week.
• Non English markets are wide open. If you operate in Polish, German, Czech, Hungarian, or any other European language, you have a first mover window that will close as GEO becomes mainstream.
• Speed matters more than perfection. I deployed an imperfect llms.txt on day three rather than wait for a perfect one on day twenty. The first version was three lines. It still worked.
• Results take weeks, not months. The first citation appeared within 4 weeks. That is faster than traditional SEO, because AI models index and update their knowledge more frequently than Google re evaluates a domain's ranking.

The EV charger market in Poland grew by 44% year over year in 2025, according to PZPM and Licznik Elektromobilności data. Competition in AI search grew right along with it. Businesses that implemented GEO early now have an advantage that is hard to close.

Geovise runs the same methodology that worked for eviacharge.pl as a standardized process for clients. Every engagement starts with a one time GEO Audit (from €400), which shows exactly which AI signals are missing for a specific business.

Frequently Asked Questions About GEO Case Studies

What does a successful GEO case study look like?

A successful GEO case study shows measurable AI citations for specific queries within 60 to 90 days of implementation, with verifiable evidence: a specific query a reader can type into Perplexity or ChatGPT and confirm. Traffic growth and branded search growth are secondary indicators, but AI citation itself is the primary proof of concept.

How long does it take to see results from GEO optimization?

First AI citations typically appear within 4 to 8 weeks of completing core technical changes (crawler access, FAQPage schema, entity consistency). Full citation coverage across multiple platforms and query types takes 60 to 90 days. This is faster than organic SEO ranking movements, which typically require 3 to 6 months for new content.

Can GEO optimization work for non-English websites?

Yes, and non-English markets are currently underserved, which makes them higher opportunity. The eviacharge.pl case study demonstrates measurable AI citations for Polish language queries. The same technical principles apply in any language: structured data, conversational FAQ content, entity consistency, and open crawler access all work regardless of the language the content is written in.

What is the difference between ranking in Google and appearing in AI search answers?

Google ranking means your page appears in a list of blue links when someone searches a keyword. Appearing in AI search answers means an AI model includes your brand, product, or information as part of a synthesized response to a conversational question, often without the user seeing a traditional results page at all. GEO targets the second behavior. Both matter, and they require different optimization strategies.

How do you measure GEO success if there are no traditional keyword rankings?

I measure GEO success through three metrics: direct AI citation tracking, meaning manually querying AI platforms with target queries and recording citations, branded search growth in Google Search Console, which correlates with AI-driven brand awareness, and referral traffic from AI platforms such as Perplexity, which now shows in analytics as a distinct traffic source. There are no industry-standard GEO ranking tools yet, so manual citation auditing remains the most reliable method.