Case Study — offzmi.com

How AI Found a Marketer Niche for $452

How I used AI + MCP-browser to surf niches, deployed a microservice to publish 222 articles, and discovered that Ukrainian marketers are massively searching for presentation tools.

Andrey Rogovsky
7 min read· Updated
Live experiment · April 2026

222

AI articles

$452

Total ad spend

401

Clicks

+1000%

Click growth on "presentation"

11.9%

Top keyword CTR

I used AI + MCP-browser to surf marketing niches, generated 222 articles and deployed a microservice for automatic publishing. The ads revealed: 64.7% of traffic went to presentation-related content. The niche was found by audience behavior, not surveys.

222 articles. $452. 1 niche. Found by clicks, not hypotheses.

offzmi.com is a project for marketers. The question before launch: what tools does the audience actually need? Instead of surveys, I ran AI niche surfing via MCP-browser, generated article content, and ran paid traffic — letting real search behavior give the answer.

What Was Done#

Three tasks defined the entire experiment: find niches via AI, publish content via microservice, validate demand via ads.

The marketing tools market is large and noisy. Without data, it's easy to spend months in the wrong direction. So instead of classic customer discovery — AI niche surfing via MCP-browser, automated generation of 222 articles, and paid traffic as a fast market signal.

// pipeline.js — AI Content Microservice
async function publishArticle(niche, keyword) {
  const content = await generateWithAI({
    topic: keyword,
    audience: 'ua-marketers',
    length: 800,
  })
  await deployToSite(content)
  await trackInAnalytics({ niche, keyword })
}

// MCP-browser niche surfing
const niches = await mcpBrowser.searchTrends({
  market: 'ukraine',
  category: 'marketing-tools',
  intent: 'informational',
})

Article publishing microservice — automated pipeline from generation to deploy

The microservice deployed 222 articles. The ads showed which topic wins.

Experiment Timeline#

Mar 2026 — AI niche surfing

MCP-browser analyzes the Ukrainian marketing market. Clusters identified: presentations, design, marketing, layouts.

Mar 2026 — 222 articles generated

AI generates article content per niche. Microservice automatically publishes to the site.

Mar 16 — Articles_Offzmi_test campaign

Google Ads launched. Unrestricted topic targeting. Conversion: traffic to articles.

Mar–Apr — Data matures

"Presentation" emerges as leader: 209 clicks of 323, CTR 11.9%, +1000% click growth.

Apr 19 — Campaign closed

$452 spent. 401 clicks. Niche identified by audience behavior.

Apr 27 — Analysis complete

Presentations are the top query. Next stage: deeper product for marketers.

One engineer. One microservice. 222 articles. Niche found.

Before vs After the Experiment

BeforeAfter
Niche knowledgeHypothesisConfirmed by data
Content0 articles222 AI articles
PublishingManualAutomated microservice
Top queryUnknown"Presentation" — 64.7% of traffic
Ad signalNoneCTR 11.9%, +1000% clicks
Next stepUndefinedPresentation tool product

How It Works#

Three components. Each solves one problem.

AI + MCP-browser

Niche surfing of the Ukrainian marketing market. Collecting trends, search clusters, and audience behavior.

Generation microservice

AI generates articles per niche → automatic publishing to the site → tracking in analytics.

Google Ads as signal

Paid traffic as a cheaper and faster alternative to customer discovery. A click = confirmed interest.

Cycle from niche to signal

StepWhat happensToolResult
1Surf marketing market nichesAI + MCP-browserTopic clusters
2Generate 222 articlesAI (GPT)222 publications
3Auto-deploy to siteMicroserviceSite populated
4Launch Google AdsArticles_Offzmi_test$452 spent
5Collect behavioral dataGoogle Ads Analytics401 clicks, CTR 11.9%
6Identify winning nicheClick analysis"Presentation" — top

Why AI + MCP-browser, not surveys#

Surveys show what people say they want. Search behavior shows what they actually look for. At $0.82 avg CPC, Google Ads produced 401 behavioral signals in 6 weeks.

The logic is simple: first AI finds potential niches by surfing the market, then content covers those niches, then ads show which one resonates with the audience. This is cheaper and faster than any focus group.

Niche discovery

AI surfs the market

1

MCP-browser analysis

Automated surfing of Ukrainian marketing niches. Collecting trends, search queries, and competitive landscape.

2

Topic clustering

AI groups found topics: presentations, design, marketing tools, layouts. Each cluster is a potential niche.

Content generation

Microservice publishes

1

222 articles by niche

AI generates article content for each cluster. Microservice automatically publishes without manual intervention.

2

Automated deploy

From generation to publication — zero manual work. The entire pipeline is automated.

Validation via ads

A click is the most honest vote

1

Articles_Offzmi_test campaign

$265.83 spent on article traffic. 323 clicks of 401 total — 80.5% of traffic came through the article campaign.

2

Niche found organically

"Presentation" — 209 clicks of 323, CTR 11.9%, +1000% click growth. The audience voted itself.

What the Data Showed#

Four findings, each actionable.

Finding 1 — Presentations: niche confirmed

209 clicks of 323 in the Articles_Offzmi_test campaign went to presentation topics. Search volume +14% month-over-month, clicks +1000%. This is not a hypothesis — it's a behavioral signal.

Finding 2 — Article campaign is cheaper than direct traffic

Articles_Offzmi_test delivered CPC 27% cheaper than the account average. Informational content competes less than transactional queries.

Finding 3 — Audience profile

The data paints a clear portrait of who is searching. This profile directly informs the next campaign's targeting.

Finding 4 — Next step defined

222 articles, $452 and 6 weeks gave a clear answer: Ukrainian marketers are massively searching for presentation creation tools. This is the foundation for the next product decision.

Unit Economics Model#

Built from experiment data. Two scenarios for the next phase.

Current click cost

Articles_Offzmi_test showed CPC $0.82 on informational traffic. For transactional traffic on "presentations" we expect $1.5–2.5 in the Ukrainian market.

Scaling scenarios

Two scenarios at $1,000/mo budget focused on the presentation niche.

What needs to be true for this to work

Three conditions for working unit economics.

Next Steps

The experiment is complete. The roadmap for the next phase.

What $452 Bought#

A structured GTM experiment produces answers, not just users. Here is what each dollar returned:

ModuleBeforeAfterImpact
AI niche discoveryHypothesisConfirmed: presentations — #1 queryDefines product strategy
Auto-publish 222 articlesManual workMicroservice, 0 manual actionsSaves 40+ hours of work
CTR winner identificationUnknown"Presentation" — CTR 11.9%Budget focus
Audience profileHypothesisMarketers 25–34, desktop, Mon–FriPrecise targeting
Trend signalUnknown+1000% clicks on presentation topicGrowing market
CPC benchmarkUnknown$0.82 article, $1.13 overallUnit economics foundation
Next productUndefined directionPresentation toolSaved months on wrong path

$452 · 6 weeks

A marketing agency would charge $10,000–20,000 for market research of this depth — without real behavioral data. AI + paid traffic as research is the AI engineer's unfair advantage.

Before vs After#

AreaBeforeAfter
Niche knowledgeHypothesis: marketers need toolsConfirmed: presentations — CTR 11.9%, +1000% clicks
Content publishingManual, no content222 articles via microservice, automatically
Audience understandingUnknownMarketers 25–34, Mon–Fri 1–6PM, desktop
Product directionBroad marketing marketPresentation tool — clear focus
Next stepUnclearMVP, SEO cluster, price A/B test

Decision Log#

Every methodological choice has a rationale.

Why AI + MCP-browser and not classic research?

Speed and scale. Manual niche surfing takes weeks. MCP-browser automatically analyzes the marketing landscape and produces structured clusters in an hour. AI removes the routine, leaving strategic decisions to humans.

Why 222 articles and not 10–20?

More niche coverage = more behavioral signals. With an automated pipeline, the cost of generating 222 articles is not much higher than 20. But the statistical significance of results is much higher.

Why paid ads and not SEO?

SEO takes 3–6 months. Ads gave 401 behavioral signals in 6 weeks. At $0.82 avg CPC — the cheapest way to quickly validate which niche actually resonates.

Why a microservice and not a CMS?

Scalability and automation. A CMS requires manual action for every publish. The microservice published 222 articles with zero manual steps — from generation to deploy.

Why is "presentation" a real niche and not noise?

Three signals confirm it: CTR 11.9% (highest in the campaign), +1000% click growth, search volume +14% month-over-month. None of them alone is sufficient signal. Together — a confirmed niche.

Takeaways#

1

AI niche surfing is faster and cheaper than any research.

MCP-browser analyzed the Ukrainian marketing market in an hour. Classic research would have taken weeks and thousands of dollars.

2

A click is the most honest audience vote.

CTR 11.9% on "presentation" said more than any survey. People vote with their wallets and clicks, not words.

3

Automated content pipeline is not a luxury — it's a necessity.

222 articles manually is months of work. The microservice did it automatically. Without automation, the GTM experiment scale would be impossible.

4

One niche matters more than broad coverage.

Better to be #1 in "presentations for marketers" than average across all marketing. The data showed the right focus itself.

Timeline#

From zero data to a confirmed niche in 6 weeks.

The experiment ran for 6 weeks.
The data answered all three research questions.
The next phase is already scoped.
The only variable left is {execution}.

Mar 2026 — AI niche surfing

MCP-browser analyzes the Ukrainian marketing market. Clusters found: presentations, design, marketing.

Mar 2026 — 222 articles published

Microservice automatically generates and publishes article content across all niches.

Mar 16 — Ads launched

Articles_Offzmi_test + Traffic campaigns. Unrestricted topic targeting.

Mar–Apr — Data matures

"Presentation" takes the lead. CTR 11.9%, +1000% clicks, trend +14%.

Apr 19 — Campaign closed

401 clicks · $452 · niche identified: presentations for marketers.

Apr 27 — Analysis complete

Unit economics built. Next phase roadmap scoped.

Transferable Methodology#

This GTM experiment structure — AI niche surfing, automated content pipeline, ads as market signal — works for any B2C or prosumer SaaS with measurable search demand.

EdTech

Same structure: AI finds topics, microservice generates educational content, ads show which topic converts. Replace "marketers" with "students" or "HR".

B2B SaaS for small business

Replace articles with niche landing pages. Replace CTR with demo requests. Niche screening logic is identical.

Creator Economy

MCP-browser finds trending topics, AI generates content, ads validate what the audience actually cares about. Without months of manual research.

The methodology is replicable. What changes is the keyword cluster, target audience, and conversion event. The framework — using AI and paid traffic to gather behavioral data before investing in growth — is domain-agnostic.

Building something and not sure which niche to go after first?

I ran this experiment for offzmi.com and turned $452 of ad spend into a confirmed niche, a behavioral audience portrait, and a next product roadmap. The same methodology applies to any product with measurable search demand.

FAQ#

Why AI + MCP-browser and not just Google Trends?

Google Trends shows what is already popular. MCP-browser analyzes the entire marketing landscape — competitors, uncovered queries, adjacent niches — and produces structured clusters. It is not a replacement for Google Trends, it is a deeper layer of analysis. MCP-browser automatically surfs Ukrainian marketing market, collects trends, search queries and competitive landscape, and produces structured clusters in one hour. Manual niche surfing takes weeks. AI removes routine, leaving strategic decisions to human. Google Trends shows general popularity, but does not give context about competitors, uncovered queries, or adjacent niches. MCP-browser gives full market picture, not just one popularity metric. Speed and scale — key advantages.

Is $452 a realistic budget for this kind of experiment?

Yes, and possibly more than needed. The key signal (CTR, presentation dominance) was visible within the first 2 weeks and approximately $150 of spend. The full 6-week run produced cleaner trend and time-dynamic data. At CPC $0.82 on article traffic, $452 gave 401 behavioral signals — cheapest way to quickly validate which niche actually resonates with audience. Marketing agency would charge $10,000-20,000 for market research of this depth without real behavioral data. AI plus paid traffic as research — unfair advantage of AI engineer. Budget $452 gave confirmed niche, behavioral audience portrait and roadmap for next product. This is effective investment.

How does the microservice publish 222 articles?

Pipeline: AI generates content by keyword and niche, automatic quality check, deploy to site, tracking in analytics. Zero manual steps between generation and publication. Microservice automatically generates article content per cluster, checks quality, publishes to site and tracks in analytics. Entire process from generation to publication is automated. 222 articles manually — months of work. Microservice did this automatically. Without automation, GTM experiment scale would be impossible. Scalability and automation — key advantages of microservice over CMS, which requires manual actions for each publication. This enables rapid hypothesis testing. Automated content pipeline — not luxury, but necessity for GTM experiments.

Why "presentation" and not another niche?

The audience chose it. Neither I nor the AI picked presentations as the winner — 64.7% of traffic organically went there. CTR 11.9% and plus 1000% click growth confirmed the signal. Three signals confirm: CTR 11.9% highest in campaign, plus 1000% click growth, search volume plus 14% month to month. None of them alone is sufficient signal. Together — confirmed niche. Click is the most honest audience vote. CTR 11.9% on presentation said more than any survey. People vote with wallet and click, not words. Data showed the right focus without hypotheses. Better to be number one in presentations for marketers.

What is the next step after this experiment?

MVP of a presentation creation tool for Ukrainian marketers. Then an SEO cluster on organic queries and a paid price A/B test at $5 versus $9 per month on confirmed traffic. Experiment complete. Roadmap for next phase defined. Develop MVP of presentation creation tool for marketer needs. SEO cluster cover queries make presentation, presentation templates AI, Canva alternative. Launch paid A/B test $5 versus $9 per month on UA traffic. Build retention mechanism templates, recommendations, collaboration. Scale article pipeline to adjacent niches design, marketing tools. Only variable remaining — execution. Niche confirmed by data, not hypotheses. Audience voted with clicks.

Resources#

Andrey Rogovsky

Andrey Rogovsky

Senior AI Engineer · GenAI · MLOps · Cloud

25 years of infrastructure. Now I build AI that survives production with MCP + RAG + K8S.

More about the author →
© 2026 Andrey Rogovsky. All rights reserved.|Privacy