01What Is Semantic SEO and How Does It Work?

Semantic SEO is the practice of organizing a page so that both text and visual layout match what Google expects. In this episode, Koray Tuğberk explains that Google uses something called centerpiece annotation. This is a visual element that helps Google decide if a page should rank.

Martin Shiplett from Google mentioned this three years ago. Then Google's DOJ leak files and API leak showed more details. There are about 22 different semantic label annotations on a page.

Koray used these ideas on a single page for audio2text.com. The page went from zero to 8,000 clicks a day with almost no text and no changes after publishing. Google uses the layout and design as a sign of originality.

Text alone no longer proves originality, but design still requires human effort.

02Main Features of the Visual Semantics Approach

Koray's approach is about how you organize a page visually, not how much content you put on it. For audio2text.com, he added a giant upload button with a no-signup notification. These two attributes are important for AI answers.

He also wrote very little text because too many tokens dilute the importance of each one.

  • Centerpiece annotation: Google looks for a main visual element to decide if a page fits the query.
  • Visual layout imitates e-commerce: Even an affiliate site should look like a product page to rank for product queries.
  • Annotation attributes: Adding clear labels like "free" and "no-signup" helps Google classify the page.
  • Short content: Fewer tokens make each word more relevant and harder for AI detectors to flag.

Koray also cleans HTML and CSS when using AI tools. AI-generated sites often leave signatures in class names and div hierarchies, which Google can detect.

03How to Set Up a Semantic SEO Page

To create a page using these principles, follow these steps:

  1. Choose an exact match domain name that clearly describes the tool or topic.
  2. Identify the key attributes your audience wants, like "free" or "no-signup", and make them prominent in the design.
  3. Create a large visual element, such as an upload button, that matches what Google expects for the query type (e-commerce, tool, etc.).
  4. Write a small amount of text. Put all important n-grams and entities in the first 2% of the tokens.
  5. If you use AI to design the page, rewrite the HTML class names and CSS to remove AI signatures.
  6. Publish the page and leave it unchanged. Let Google's algorithm gradually understand it.

Koray noticed that every core update increased his clicks. Sometimes he lost 50% of clicks temporarily, but he did not change anything because Google needs time to trust the page.

04How to Use Visual Semantics and Text Structure

Koray uses visual annotations to tell Google what the page is about. For audio2text.com, that means a clear upload button and the text "free audio-to-text tool". He also designed the page to look like a tool, not a blog post.

The text structure matters too. Koray has four criteria for content:

  • Factual vs. opinionated: Factual content starts with "is a" definitions, then predicates, then attributes. Opinionated content is casual and follows no order.
  • Structured vs. unstructured: Some queries need listicles and tables, others just simple paragraphs.
  • Combining factual and UGC: On pages like eHarmony Senior Dating, he repeats questions in a forum style with an expert answering casually.
  • Query deserves a page, heading, or section: He uses this to decide how to organize each query.

This follows Google's AI overview patent, which scores both language modalities and gives a combined score.

05Results and Case Studies from the Episode

The most striking result is the single page for audio2text.com. Koray used an exact match domain and a simple design. The page went from zero to 8,000 clicks a day.

There were zero content changes and zero visual changes during that time. Every core update increased traffic.

Koray also shared a cautionary case study: BrowserStack lost 70% of their traffic in eight months. They have 23,000 referring domains and 300,000 monthly brand search demands. But they published 2,000 AI-generated documents in one day.

Google's scaled content decision trees crushed them. This shows that even strong brands can get penalized for scaling too fast.

He also mentioned that covering 20-30% more topics than competitors makes you a replacement candidate during core updates.

06Tips, Limits, and Mistakes to Avoid

Koray emphasized that freshness matters less than the age of a subfolder. If you create a new subfolder and publish 1,000 URLs in seven days, it is risky. Instead, mature the subfolder for three months.

Then publish with increasing frequency and randomize the numbers. For example, publish 2 today, 27 tomorrow, then 7 and 8.

He also warned about AI detection. AI detectors often work like Turkish police: they assume you are guilty if you pass a threshold. They are more accurate on long content.

Short pages with few tokens are harder to classify as AI. Koray uses this to stay under the radar.

  • Avoid scaling too fast: Even big brands can be penalized for publishing thousands of AI pages quickly.
  • Don't ignore visual design: It is an originality signal that text no longer provides.
  • Don't change a page too soon: Let Google gain confidence over time.
  • Clean AI signatures: Remove default class names and div structures from AI-generated sites.

07Who Should Use Semantic SEO and Why It Matters

Semantic SEO matters for anyone who wants to rank in Google and AI overviews. It is especially useful for affiliate sites, free tools, and local businesses. Even lawyers can create free tools like accident photo evaluation calculators.

These tools bring back clicks that informational content loses to AI answers.

Koray also applies the same principles to link building. He calls it the "external topical map." You get page rank, mentions, and claims from external publications. These claims help build a consensus for AI answers, like "biggest law firm in Texas."

He also creates exact match domains, subreddits, and secondary sites to feed AI with content. For example, he makes books with exact match titles, like "Truck Accident Love Your Texas." These books rank as book sites and also support the main site.

In the future, AI will keep changing search. Using visual semantics and careful design is one way to stay ahead.