What are the AI-assisted keyword clustering methods?

Short answer: Two method families are reliable: SERP overlap clustering (queries sharing result pages belong together) and embedding-based semantic clustering (queries vectorized and grouped by similarity). The most robust results come from combining both, finished with human intent labels.

How the methods work

• SERP overlap: when two queries share a meaningful portion of top results, Google treats them as the same intent; the cluster gets one URL.
• Embedding clustering: queries are vectorized with an embedding model and grouped via cosine similarity using k-means or HDBSCAN.
• LLM labeling: ChatGPT, Gemini, Claude, or DeepSeek propose intent labels; an editor makes the final call.
• Validation: cross-check embedding clusters against SERP overlap; semantically close queries with distinct SERPs still need separate pages.

Bind the clusters to your URL plan and refresh them periodically, because intent drifts over time.