Keyword research

A complete keyword research workflow captures four data points per query:

• Search volume: how often the query is run per month. Google Search Console gives you yours; third-party tools (Ahrefs, Semrush, Indexly) estimate competitor and category volumes.

• Search intent: informational, navigational, commercial, or transactional — which determines what content format will rank.

• Difficulty: a composite signal of how competitive the SERP is. High difficulty + high volume = expensive to win.

• AI prompt volume (new in 2026): how often the same intent gets asked of ChatGPT, Claude, or Perplexity. Some queries have collapsed on Google and exploded on AI assistants — only AI-aware keyword research surfaces this shift.

Sources include Search Console, Google Keyword Planner, third-party tools, AI prompt logs, sales and support transcripts, and Reddit/community mining.

Traditional keyword research targets typed queries on Google. Prompt research targets conversational queries on AI assistants.

The two diverge in three ways:

1. Length and specificity. Typed queries are 2–4 words. Prompts are full sentences with qualifiers and follow-ups.

2. Volume sources. Typed query volume comes from Google data. Prompt volume comes from prompt logs, sales conversations, and direct sampling of AI assistants.

3. Intent shape. Typed queries often start research; prompts often skip ahead to evaluation ("compare X and Y") or decision ("which is best for [scenario]").

In 2026, mature programs run both — typed-query research for SEO and prompt research for GEO/AEO — then deduplicate the overlap.

Keyword research is what prevents publishing into the void. Three concrete payoffs:

1. Content matches demand. Researched topics have measurable audience volume; un-researched topics often turn out to have none.

2. Intent shapes format. Knowing whether a query is informational or transactional decides whether to write a definition page, comparison page, or product page.

3. AI search gaps surface early. Some queries have low Google volume and high AI prompt volume — the highest-leverage opportunity for brands moving into GEO/AEO. Only AI-aware keyword research finds them.

Six-step workflow:

1. Mine first-party sources. Search Console, sales call transcripts, support tickets, community forums. The richest signal is usually internal, not bought.

2. Expand with third-party tools. Ahrefs, Semrush, or Indexly pull keyword universes around seed terms.

3. Add prompt research. Sample AI assistants directly with seed prompts and capture follow-up questions. Indexly aggregates prompt-volume data across ChatGPT, Claude, Perplexity, and AI Mode.

4. Cluster by intent. Group queries into informational, comparison, and decision intent. Different formats win different intents.

5. Score by difficulty + volume + AI demand. Build a composite priority score. Top of the list is what you fund first.

6. Re-run quarterly. Search behavior shifts — especially as AI assistants capture more informational queries. Quarterly refreshes catch the drift.