How to Do AI Keyword Research with ChatGPT, Claude, and Ooty Octopus

Keyword research with traditional tools follows a predictable loop: open the tool, enter a seed keyword, scroll through hundreds of results, export to a spreadsheet, filter manually, repeat. The data retrieval and the analysis happen in completely different places.

This tutorial walks through a different approach -- using your AI assistant (ChatGPT, Claude, or any MCP-compatible client) with Ooty Octopus to do the entire process in a single conversation. Your AI pulls live Google Suggest data through Octopus, classifies intent, estimates volumes, and helps you map everything to a content plan. No CSV exports. No tab-switching.

We will use a real example throughout: keyword research for a beginner's guide to making cold brew coffee at home.

The AI Keyword Research Workflow

Eight steps from seed topic to prioritised keyword plan -- all in one conversation

1🌱

Seed Topic

Give Claude your topic and audience context

2🔍

Discover Keywords

Pull real Google Suggest data via Octopus

3❓

Question Keywords

Surface what people actually ask

4🎯

Classify Intent

Map keywords by search intent type

5🌿

Expand Variations

Find long-tail opportunities

6📊

Estimate Volume

Prioritise by relative demand

7📝

Map to Content

Build your outline from keyword data

8✅

Priority List

Final ranked keywords with rationale

What You Need

Before starting, make sure you have:

• An Ooty account with Octopus activated -- sign up at ooty.io if you haven't already
• ChatGPT, Claude Desktop, or any MCP-compatible client with Octopus connected via your MCP endpoint URL and licence key
• A topic or business area you want to research -- the more specific the better

First-time setup takes about two minutes. In ChatGPT, go to Settings > Connected apps and add Ooty Octopus using your endpoint URL and licence key. In Claude Desktop, add it via the MCP tools section. Either way, it is a one-time task -- details are in your Ooty dashboard.

What Octopus Gives Your AI Assistant

When you connect Octopus, your AI assistant gains access to several keyword research tools:

• discover_keywords -- expands a seed topic into related keywords using Google Suggest data (three modes: basic, deep, questions)
• estimate_search_volume -- estimates monthly search volumes for keyword lists
• expand_keywords -- generates keyword variations from a seed term
• expand_search_intent -- classifies keywords by intent (informational, navigational, commercial, transactional)

You never need to call these by name. Ask what you want to find, and your AI picks the right tool.

C

Prompt to Claude

“I’m writing a beginner’s guide to cold brew coffee at home. My readers own basic kitchen equipment and have never made cold brew before. Find keywords for this guide -- start by discovering what people actually search for, then expand into question keywords, classify everything by intent, and give me a prioritised list mapped to content sections.”

Tools called:discover_keywordsexpand_keywordsexpand_search_intentestimate_search_volume

Step 1: Start with Your Seed Topic

Begin by giving Claude your topic and audience context. The more specific the context, the better the output.

Prompt:

"I'm writing a guide for people who want to start making cold brew coffee at home. They're beginners -- they've never made cold brew before and probably own basic kitchen equipment. Help me find keywords for this guide. Start by discovering what people actually search for around this topic."

Your AI calls discover_keywords with "cold brew coffee at home" and returns real Google search suggestions. You will see queries like "cold brew coffee ratio", "how long to steep cold brew", "cold brew coffee maker vs pitcher" -- actual search queries people type, not interpolated database entries.

Why this matters: Google Suggest data reflects real search behaviour. It is more reliable for understanding actual demand than keyword databases that estimate from clickstream models (Ahrefs notes that third-party volume estimates can be off by 50% or more).

Step 2: Expand into Question Keywords

Questions are high-value targets for content. They map directly to headers, match voice search patterns, and often trigger featured snippets.

Prompt:

"Now find question-based keywords -- things people ask with 'how to', 'what is', 'why does', 'how long', 'can you'. I want to understand what beginners are actually confused about."

Your AI calls discover_keywords in questions mode, returning queries like:

• "how long to steep cold brew coffee"
• "why is my cold brew bitter"
• "can you make cold brew with regular coffee"
• "how much coffee for cold brew"

These are the exact questions your content should answer. According to Backlinko's analysis of 306 million keywords, question-based queries have grown 61% over the past five years and tend to have higher click-through rates than non-question queries.

Step 3: Classify by Search Intent

Not all keywords want the same content. Someone searching "cold brew coffee ratio" wants a specific answer. Someone searching "best cold brew maker" is comparing products. The same page cannot serve both well.

Prompt:

"From the keywords you've found, classify them by search intent -- informational (learning), commercial (comparing options), transactional (ready to buy), and navigational (looking for a specific brand)."

Your AI calls expand_search_intent and returns a categorised breakdown:

• Informational: how to make cold brew, cold brew coffee ratio, why cold brew is less acidic
• Commercial: best cold brew maker, cold brew maker vs pitcher, cold brew bags review
• Transactional: buy cold brew maker, Hario Mizudashi cold brew, cold brew filter bag
• Navigational: Starbucks cold brew recipe, Chameleon cold brew where to buy

This tells you which keywords belong in your beginner's guide (informational), which need a separate product comparison article (commercial/transactional), and which you should probably skip entirely.

Search Intent Classification

Not all keywords want the same content -- intent determines the right format

Informational

55%

Learning and researching. Examples: how to make cold brew, cold brew ratio

Best content type: Guides, tutorials, explainers

Commercial

22%

Comparing options before buying. Examples: best cold brew maker, french press vs cold brew

Best content type: Comparisons, reviews, roundups

Transactional

15%

Ready to purchase. Examples: buy Hario cold brew, cold brew maker deal

Best content type: Product pages, landing pages

Navigational

8%

Looking for a specific brand. Examples: Starbucks cold brew recipe, Chameleon cold brew

Best content type: Brand pages (low priority for you)

Step 4: Find Long-Tail Variations

Once you have core keywords, expanding into variations surfaces lower-competition opportunities with more specific intent.

Prompt:

"Take 'cold brew coffee ratio' and expand it into keyword variations. I want more specific, long-tail versions -- ideally ones where the searcher has a particular problem or situation."

Your AI calls expand_keywords and returns:

• "cold brew coffee ratio for concentrate"
• "cold brew coffee ratio tablespoons"
• "cold brew coffee ratio 32 oz pitcher"
• "cold brew ratio for Toddy maker"

Long-tail keywords like these are easier to rank for and indicate very specific needs -- perfect for FAQ sections, subheadings, or dedicated content clusters.

Step 5: Get Volume Estimates

Before committing to a keyword, you want a sense of relative demand. Volume estimates are not exact, but they are useful for prioritisation.

Prompt:

"Get me rough search volume estimates for the top 15 keywords from our research. I want to prioritise which ones to focus on in my guide."

Your AI calls estimate_search_volume for your list and returns approximate monthly search volumes. Use these as relative indicators: a keyword showing 1,200 searches/month is likely more popular than one at 200/month, even if the exact numbers are imprecise.

A note on accuracy: Google's own Keyword Planner groups volumes into ranges, and third-party tools use different estimation methods. Treat any volume number as directional, not absolute.

Step 6: Map Keywords to Content Structure

This is where the conversation approach shows its real advantage. You can ask Claude to build a content outline directly from the research -- no export step, no separate planning doc.

Prompt:

"Based on all the keywords we've found, help me build an outline for my beginner's guide to cold brew at home. Map the keywords to the most relevant sections so I know what to cover where."

Your AI uses the research from your conversation to suggest a structure:

• Introduction: cold brew at home, is cold brew easy to make (informational intent)
• Equipment: cold brew equipment, do you need special coffee for cold brew (informational)
• The Ratio: cold brew coffee ratio, cold brew ratio tablespoons, concentrate vs regular (specific questions)
• The Process: how to make cold brew step by step, how long to steep cold brew (how-to intent)
• Troubleshooting: why is my cold brew bitter, cold brew too weak (problem-solving)
• FAQ: can you use regular coffee, how long does cold brew last (questions intent)

Step 7: Identify Content Gaps

Understanding what a comprehensive page needs to cover -- beyond your target keyword -- helps you build topical depth.

Prompt:

"For 'how to make cold brew coffee at home', what related topics would a comprehensive page need to address to fully satisfy search intent? What are searchers likely also wondering?"

This uses your AI assistant's reasoning to identify semantic coverage gaps. Research from Clearscope shows that pages covering a topic's full semantic cluster tend to rank for significantly more keywords than pages focused narrowly on a single term.

Step 8: Build Your Priority List

With all the data collected, ask Claude to synthesise it into a final keyword plan.

Prompt:

"Based on everything we've researched, give me a prioritised list of 20 keywords for my guide. Split them into: primary keyword (main target), secondary keywords (important variations), supporting keywords (for FAQ and subheadings). Explain your prioritisation briefly."

Your AI produces a structured keyword plan with rationale -- something you could hand to a writer with clear instructions about what the content needs to achieve.

Time Comparison

Minutes per task: traditional manual tools vs Claude + Octopus

Traditional Tools

215min

Claude + Octopus

19min

Per-Task Breakdown (minutes)

Seed expansion

25m/2m

Question discovery

30m/2m

Intent classification

45m/3m

Volume estimation

15m/2m

Content mapping

40m/5m

Gap analysis

60m/5m

Tips for Better Results

Be specific about your audience. "Beginners with basic kitchen equipment" produces better keyword targeting than just "beginners." The more context you provide, the more relevant Claude's tool calls become.

Iterate in the same conversation. If the first batch of keywords misses the mark, tell your AI what is off. "These are too advanced -- I need keywords for people who've never made any specialty coffee before." It adjusts without starting over.

Do not treat volume estimates as precise. Use them for relative prioritisation only. A keyword with 800/month estimated volume and clear commercial intent is often more valuable than one with 5,000/month and vague informational intent.

Separate research by intent. Keep informational keyword research separate from commercial keyword research. They require different content formats -- mixing them leads to pages that try to do too much and satisfy nobody.

Save the conversation. Copy your keyword lists before closing. Claude's conversation history is useful for context but is not a replacement for your own notes or project management system.

Common Mistakes

Chasing high-volume keywords without considering competition. Google Suggest data shows you what people search for. It does not show you domain authority or competitive difficulty. A 10,000/month keyword dominated by industry leaders is not easier to rank for just because you found it.

Skipping the questions mode. The question expansion often surfaces the most actionable keywords for content. People searching "why is my cold brew bitter" have a specific problem you can solve -- that is more valuable than ranking for a generic head term.

Treating all keywords as equal. Intent matters more than volume. Three keywords at 300 searches each with high commercial intent are frequently more valuable than one keyword at 2,000 searches with vague informational intent.

Keyword research without content planning. Keywords are inputs to content strategy, not outputs. If you end the session without a clear content plan, the research has not delivered its value yet. The content mapping step (Step 6) is where research becomes actionable.

What You Have Built

After working through this tutorial, you will have:

• A primary keyword with clear search intent
• A set of secondary and supporting keywords mapped to content sections
• A content outline driven by real search demand
• An understanding of adjacent topics for future content
• Enough context for a writer to create content that serves both search engines and real readers

That is keyword research connected to your content strategy -- not a spreadsheet of volumes with no plan.

Get started with Ooty Octopus