Skip to content
Lazarina Stoy

Lazarina Stoy

Founder, MLforSEO

MLforSEO

At a Glance

Role
Founder of MLforSEO
Training
MSc in the computer science department, University of Strathclyde
Reach
More than 700 marketers trained and 20 conference talks
Community
Founder of Women in Marketing - Bulgaria
Lazarina Stoy runs MLforSEO, where she teaches organic search marketers to apply machine learning to entity analysis, AI search, technical SEO, and reporting. She consults for B2B and enterprise teams and has worked through agencies with brands such as AWS, Skyscanner, Compass, and Extreme Networks.

Automating From the First Job

Stoy took a joint degree in marketing and management, then an MSc in the computer science department at the University of Strathclyde, where her dissertation was on machine learning. In her first agency role in 2020 she began automating routine SEO work such as meta descriptions, image captions, Core Web Vitals audits, and entity analysis.

From 2021 to 2023 she was senior manager for SEO and data science at Intrepid Digital, where she led SEO for teams at a big-tech client and ran the agency's data function. In 2024 she launched the MLforSEO Academy and founded Women in Marketing - Bulgaria.

Entities and Query Fan-Out

In a February 2026 conversation with Gianluca Fiorelli, she explained why entity search suddenly mattered to SEOs. “SEOs have been pushed to really understand how LLM works,” she said, and once they read the patents, entity relationships and Knowledge Graph access stood out as the ways AI search systems cut down on errors.

That is why her courses start with entities and structured data. Our schema markup research looks at one of the machine-readable signals she teaches marketers to check.

The Limits of LLMs in SEO

Stoy uses chatbots every day. “I am 100% a proponent of LLMs,” she said. Her objection is to using them for the thinking. “My pet peeve with the use of LLMs is when we're trying to replace thinking or research or planning, which are steps that we should own.”

For tasks such as entity analysis, she prefers a dedicated extraction API, which returns fuller and more predictable data. She sees LLMs as best at turning one format into another, such as a PDF into a blog post. Our AI answer volatility statistics show why predictable methods appeal to teams that report on AI search, and her MLforSEO training is built on that idea.

Links

← All people

Sources