April 3, 20266 min read

Case Study: From 0% to 46% AI Visibility in 10 Days

A detailed breakdown of the GEO Scout case: how a brand moved from zero visibility in Yandex with Alice to 46% AI visibility in 10 days using expert content, FAQ, JSON-LD, and daily monitoring.

Vladislav Puchkov

Founder of GEO Scout, GEO optimization expert

This article is based on the publicly described case published on Sostav on March 31, 2026, but here the goal is different: not a short success story, but a repeatable operating model.

Starting point

At the start, GEO Scout checked its own niche visibility in Yandex with Alice and found a serious gap:

• the brand was not mentioned in core niche answers
• competitors already appeared in recommendations
• the domain was rarely treated as a primary source

This was not a CTR problem or a classic SEO ranking problem. It was a shortlist problem: when AI generated a recommendation set, GEO Scout was missing from it.

That is the essence of the AI dark funnel. The user can form a preference before ever clicking a website. See the AI dark funnel.

What was done

1. Rapid gap audit

The first step was basic diagnosis:

• which prompts actually generated niche answers
• which competitors appeared most often
• which sources Alice cited
• which topics and page types were missing on the site

This turned a vague goal like “improve visibility” into a concrete backlog.

2. A content sprint of 105 expert articles

Within 12 days, 105 expert articles were published.

The key point is not just the quantity, but the structure:

• foundational definitions
• how-to articles
• comparisons and rankings
• industry-specific materials
• research-driven pieces

This is what signals to AI that the brand does not just mention a category, but actually covers the field with depth.

For the underlying logic, see content strategy for GEO from scratch.

3. Each article was built as a citable asset

The content was not written like generic SEO copy. It was designed as reusable evidence for AI systems.

That meant adding:

• exact definitions
• numbers and factual claims
• comparison tables
• direct conclusions in the opening paragraphs
• FAQ sections

These elements are easier for AI to quote or paraphrase than abstract commentary. See what kind of content AI cites most often.

4. FAQ and JSON-LD as the machine-readable layer

The content setup used:

• FAQPage
• BlogPosting
• BreadcrumbList

That helped the AI system extract structure, question-answer pairs, and article context more reliably.

5. Daily monitoring instead of occasional checking

The final critical piece was daily measurement:

• did the brand appear in answers yet
• where did it rank
• did Alice cite the domain directly
• which prompt clusters improved and which did not

Without this, the project would have been a publishing burst rather than a managed feedback loop.

Results

According to the published case data:

• AI visibility grew from 0% to 46%
• Share of Voice reached 21.9%
• Share of Citation reached 54.5%
• first-week growth was x11
• by March 30, 2026, Alice linked to geoscout.pro in 24 out of 51 answers
• the average brand position in answers was between 1.7 and 2.1

This metric combination matters because it shows two different gains:

1. the brand was mentioned more often
2. the brand was cited more often as a source

Those are not the same thing. Mentions show presence. Citation shows authority. For more, see cited sources in AI.

Why the case worked so quickly

Reason 1: the niche had a clear content deficit

When a niche lacks high-quality AI-citable content, visibility can move fast. You do not have to beat everyone at everything. You often just have to close the most obvious structural gap.

Reason 2: Yandex with Alice can respond quickly to accessible, structured web content

For AI systems that rely on current web input, a strong publishing sprint can move much faster than classic SEO.

Reason 3: the effort covered a cluster, not a single keyword

One article about “what is GEO” is not enough to create authority. Definitions, how-to, comparisons, rankings, and proof pages together create a semantic field that AI can trust.

What this case should not be misunderstood to mean

There are three weak takeaways teams should avoid.

Wrong takeaway 1: just publish 100 articles

No. Volume alone is not the lesson. The lesson is:

• factual blocks
• strong topic architecture
• FAQ and schema
• prompt-driven prioritization

Wrong takeaway 2: GEO always works in 10 days

No. Speed depends on:

• the provider
• the niche
• starting visibility
• technical accessibility

Wrong takeaway 3: this only applies to media-like niches

No. The model transfers well:

• SaaS: docs, integrations, comparisons
• e-commerce: categories, product pages, buying guides
• services: case studies, FAQ, About, vertical pages

How to repeat the logic

A simplified repeatable framework:

1. Record a baseline across 20-50 prompts
2. Identify which sources the AI already cites in your niche
3. Close gaps in four formats:
   • definitions
   • how-to
   • comparisons
   • proof / case studies
4. Add FAQ and schema to the highest-value pages
5. Measure changes daily or weekly

The main lesson for marketers

This case matters not because the graph looks impressive, but because it demonstrates a new marketing mechanism:

an expert blog becomes visibility infrastructure for AI, not just an SEO traffic channel.

If a brand systematically creates citable content and tracks how AI systems use it, it starts influencing the shortlists that AI gives to users. That is not just a content metric. It is influence over market choice.