SERP Features

SERP features are the elements on a search engine results page that go beyond standard organic listings: featured snippets, People Also Ask boxes, AI Overviews, image and video packs, local packs, Top stories, shopping results, knowledge panels and more. Which features appear tells you a lot about how a search engine has interpreted a query.

How it works

Search engines decide which features to show for each query based on their reading of search intent, the entities in the query, freshness needs and the formats people have engaged with before. Google tests and changes the layout constantly, and its patents describe systems for generating and scoring things like short answers. Some features pull directly from structured data or the Knowledge Graph; others are built from page content.

Keyword tools such as Semrush, Ahrefs or SERP APIs report the features present for each keyword, which turns the results page into data you can analyse at scale.

How it’s used in SEO and marketing

• Intent classification: a video carousel or “Things to know” block hints at informational intent; shopping results and product grids point to transactional intent.
• Format planning: the features served for a keyword cluster show which content formats and platforms (video, images, forums, maps) you need to compete.
• Opportunity sizing: a position-one organic result under an AI Overview, ads and a PAA box gets a very different share of attention than one at the top of a clean page.
• Tracking change: shifts in feature prevalence across your keyword universe are an early signal that Google has reinterpreted a topic.

Example

In MLforSEO’s semantic keyword research lab, a keyword universe exported with its SERP feature column is enriched with new labels: keywords showing video results get a “video” format flag, those showing local packs get a “local” flag, and so on. The enriched data then feeds content briefs, so each brief states the formats the page should include. No machine learning is needed for this step, only aggregation and transformation of data you already have.

Limitations and things to watch

• Features vary by location, device, language and personalisation, so one scrape is a sample. See search personalisation.
• Some features are designed to answer the question on the page, which feeds zero-click searches. Visibility may rise while clicks fall.
• Tool coverage differs. Check how your provider names and detects each feature before comparing data sets.
• Google retires and renames features often, so dated labels in old exports need cleaning.

Learn it on MLforSEO

• What Is SERP Feature Analysis? How to Use It for Smarter Keyword Research
• 10 Ways to Automate SERP Analysis with AI
• Search Intent Classification from SERP Features in Python: Rule-Based Notebook for SEO
• Identifying Desired Content Formats & Platforms from SERP Data (Lab Notebook)
• Semantic ML-enabled Keyword Research (course)