How To Predict the Meta Description Google Will Use for Target Keywords Based on Cosine Similarity

Google rewriting your meta descriptions is more common than you might think. According to Ahrefs, Google rewrites meta descriptions over 60% of the time. Interestingly, these rewrites often utilize sentences already present on your webpage. At Go Fish Digital, we’ve uncovered a reliable method to predict which sentences Google is likely to select by analyzing cosine similarity between keyword and sentence embeddings.

How Google Likely Selects What Text To Use As Meta Descriptions

Based on our observation and examples demonstrated below, there is a correlation between the sentence on your page with the highest cosine similarity to your target keyword. It’s likely that Google’s system uses vector embeddings for passages on your page and measures those against the keywords used in a query to find what it thinks is the most relevant text to use as a meta description for your page.

Why Predict Meta Descriptions with Cosine Similarity?

Using cosine similarity to predict Google’s meta description rewrites can significantly enhance your SEO strategy by:

• Improving Click-Through Rates: Optimizing likely meta description candidates increases relevance and can boost clicks.
• Enhancing Content Relevance: Aligning content more closely with target keywords improves your page’s ranking potential.
• Identifying Content Opportunities: Reveals insights into Google’s understanding of your content, guiding further optimization.

Below, you’ll see an example illustrating our findings.  Note: At Go Fish, we prefer moving cosine similarity to a 10 point scale for easier reading.

Examples of Meta Description Predictions:

Keyword: Best Digital PR Agency
Closest Sentence: Our award-winning digital PR agency is the best at creating end-to-end campaigns that drive coverage from major media sites.
Cosine Similarity: 8.860
Meta Description Used by Google: Our award-winning digital PR agency is the best at creating end-to-end campaigns that drive coverage from major media sites.

Keyword: What are Digital PR Services
Closest Sentence: Digital PR is the marketing practice of building backlinks to a site by creating content that’s covered by media outlets.
Cosine Similarity: 8.667
Meta Description Used by Google: Digital PR is the marketing practice of building backlinks to a site by creating content that’s covered by media outlets.

Keyword: Award winning digital PR agency
Closest Sentence: Our award-winning digital PR agency is the best at creating end-to-end campaigns that drive coverage from major media sites.
Cosine Similarity: 9.041
Meta Description Used by Google: Our award-winning digital PR agency is the best at creating end-to-end campaigns that drive coverage from major media sites.

How to Identify the Meta Descriptions Google Likely Uses

Based on our observation and demonstrated above, there is a correlation between the sentence on your page with the highest cosine similarity to your target keyword. It’s likely that Google’s system uses vector embeddings for passages on your page and measures those against the keywords used in a query to find what it thinks is the most relevant text to use as a meta description for your page.

At Go Fish Digital, we developed a streamlined process to predict meta descriptions based on this cosine similarity calculation. Here’s how we do it:

Step 1: Export Keywords from Google Search Console

Select a specific page and view queries with the highest impressions.  You’ll want to take these keywords to predict what Google is using as meta descriptions for them.

Step 2: List Sentences in Google Sheets

Copy each sentence from the selected page into individual rows in a Google Sheet.

Step 3: Input Target Keywords

Enter your chosen target keywords from Google Search Console across the top as columns.

Step 4: Compute Cosine Similarity

At Go Fish, we use our proprietary Google App Script that integrates Google text embedding engine into Google Sheets to help with calculating cosine similarity. We create embeddings for each sentence and keyword, calculate the cosine similarity, and convert it into a clear 10-point scale.

Step 5: Apply Conditional Formatting

Use conditional formatting to quickly identify which sentences closely match each keyword. The sentence with the highest similarity score is likely what Google uses, unless it’s within a header or a unique HTML tag.

Optimizing Your Predicted Meta Descriptions

Once you’ve identified these critical sentences on your page, you can strategically optimize them to better align with your SEO objectives. Enhancing these sentences not only increases relevance but also potentially improves click-through rates and overall user engagement.

By leveraging cosine similarity, you can strategically refine your content, anticipate Google’s meta description selections, and achieve more effective SEO outcomes.

Need Help Doing This On Your Site? Head over to our contact us page and let us know that you’re looking for help with using cosine similarity to optimize your page content. 

About Dan Hinckley

Dan specializes in innovative digital marketing solutions, brand strategy, and technical web development and support. He works in step with clients to manage SEO, ORM, Conversion Rate Optimization, and website design and development projects, from start to finish. He is Google Analytics Certified and leads the Google AdWords programs. Dan began his career in brand strategy management before transitioning into IT consulting for the US Government. These experiences uniquely position him to develop comprehensive strategies for SEO that span across branding and IT domains. Nearly twenty years of SEO experience have made Dan a seasoned industry veteran in tune with the latest capabilities. As a consultant, his clients have included GEICO, AAG, NY Times, Business Insider, Marriott, US Department of Defense, Airbnb, US State Department, George Mason University, and Penn State University among others.