Semantic AI-powered SEO Keyword Research
- Getting your data – data sources run-through
- How to move from traditional to semantic keyword universe – Checklist, based on course tasks
- Search Query Sequences and Query Path
NEW: AI Search Optimisation and Agentic SEO Course by Beatrice Gamba -> Immediate start, on-demand learning for how to optimise for discovery in agentic search ✨ ✨
Search & SEO Concepts
In shortGoogle's SERP feature showing a group of related questions; expanding one reveals a short answer from a web page and often loads more questions.
People Also Ask (PAA) is a Google SERP feature that shows a group of questions related to the search. When you expand a question, Google shows a short answer taken from a web page, with a link to the source, and often adds more questions to the list. Google’s own documentation calls it a “related questions group”.
MLforSEO explains PAA as one of the features Google builds from aggregated search behaviour, alongside Related Searches and People Also Search For. The questions reflect common transitions between queries within and across sessions, which is why PAA is so useful for studying query sequences and query paths. Google states that site owners cannot control which questions appear; the answer shown when a question is expanded is a featured snippet, so the way to influence it is the same as for featured snippets.
The box is dynamic. Each expansion loads further questions, and the list can drift away from the original topic the more you click.
In MLforSEO’s SERP analysis guide, PAA questions come free with the full SERP pull from the DataForSEO API. Asking the API to expand the box one level costs $0.00015 per click and returns about six questions per SERP instead of four (the API allows a click depth of 1 to 4). The workflow then counts each question across keywords and markets, and asks an LLM with structured outputs to group them into themes. In the test, one theme was clearly off-topic: expanding a question led Google to “Can ChatGPT do SEO?”. Filtering that out kept the FAQ section focused. For thousands of questions, the guide recommends embeddings and clustering, with the LLM only naming the clusters.