Keyword Research and Claude – How It Can Help (2026 Guide)

Updated October 2026 · Written by the My SEO Experts team, based on hands-on testing and the latest guides from Semrush, Anthropic and leading SEO practitioners.

Keyword research used to mean hours in spreadsheets: exporting lists, removing duplicates, guessing intent and grouping terms by hand. In 2026, Claude — Anthropic’s AI assistant — can do most of that work in minutes. Connected to the right data, it can find keyword gaps against competitors, mine your Search Console for quick wins, turn customer reviews into search queries, cluster thousands of keywords and hand you a prioritised content plan.

But there’s a catch that most “AI keyword research” articles skip: Claude does not have search volume or keyword difficulty data built in. Used on its own, it can only brainstorm. Used with the right connectors, Projects and Skills, it becomes one of the most powerful keyword research assistants available.

This guide covers everything: what Claude can and can’t do, the five ways to set it up, step-by-step workflows with copy-paste prompts, a reusable Skill template, Google Ads and local SEO use cases, costs, limitations and the mistakes to avoid.

Key Takeaways

• Claude is the analyst, not the database. It brainstorms, classifies, clusters and prioritises — the numbers must come from an SEO tool.
• MCP connectors are the game changer. Semrush, Ahrefs, DataForSEO, Keyword Tool and others let Claude pull live keyword data during a conversation.
• Context beats clever prompts. A Claude Project with your business details, competitors and existing URLs produces far more relevant keywords than any single prompt.
• Customer language finds keywords tools miss. Reviews, Reddit threads and sales calls reveal the exact phrases buyers use.
• Skills make it repeatable. A keyword research Skill gives you the same structured output every time, for every client.
• Humans still decide. Use Claude to do the heavy lifting, then review, prioritise and approve.

Table of Contents

1. Can Claude do keyword research?
2. 5 ways to use Claude for keyword research
3. Connecting keyword data: MCP connectors compared
4. Step-by-step: setting up Claude for keyword research
5. 8 keyword research workflows (with prompts)
6. How to prioritise keywords with Claude
7. Build a reusable keyword research Skill
8. Claude for Google Ads and local SEO keywords
9. Prompt library
10. Limitations, costs and mistakes to avoid
11. FAQs

Can Claude Do Keyword Research?

Yes — but it helps to be clear about which parts of keyword research Claude is good at, and which parts need outside data.

What Claude can and can’t do for keyword research in 2026.

Keyword research task	Claude on its own	Claude + SEO data (MCP or CSV)
Brainstorm seed keywords and topics	✅ Excellent	✅ Excellent
Generate long-tail and question keywords	✅ Good (unvalidated)	✅ Validated with real data
Search volume, keyword difficulty, CPC	❌ Guesses only	✅ From your connected tool
Classify search intent and funnel stage	✅ Excellent	✅ Excellent, at scale
Cluster keywords by topic	✅ Good for small lists	✅ Thousands of keywords in batches
Competitor keyword gaps	❌ No access to rankings	✅ Live gap analysis
Search Console quick wins	⚠️ Only if you upload CSVs	✅ CSV upload or third-party connector
Live SERP features (AI Overviews, PAA)	⚠️ Web search only	✅ Via SERP-enabled connectors
Turn keywords into content briefs	✅ Excellent	✅ Excellent

Rule of thumb: if Claude gives you a search volume number and you haven’t connected a data source or uploaded a file, treat that number as made up.

5 Ways to Use Claude for Keyword Research

There isn’t one “right” way to use Claude for keyword research. Think of it as five levels — each adds better data, more automation and more consistent results.

Level	How it works	Best for	Setup time	Data quality
1. Chat only	Ask Claude for ideas, questions, intents and angles	Brainstorming, new niches	None	No metrics
2. Export & paste	Export CSVs from Semrush, Ahrefs, Keyword Planner or Search Console and upload them	One-off analysis, cleaning and clustering	Minutes	As fresh as the export
3. Projects	Saved instructions, business context and files in a Claude project	Ongoing work for one site or client	15–30 min	Depends on files
4. MCP connectors	Claude calls an SEO tool directly mid-conversation	Regular research, competitor gaps	5–15 min	Live, per request
5. Skills & Claude Code	A packaged, repeatable workflow that outputs spreadsheets and briefs	Agencies, large sites, scaled research	1–3 hours	Live and consistent

Export-and-paste vs live MCP connection

	Export & paste	Live MCP connector
Where the data comes from	A file you download from your SEO tool	Claude queries the tool directly
Freshness	Stale once exported	Live on every request
Follow-up questions	Need a new export	Claude fetches more data itself
Multi-step research	Manual stitching	Claude runs the whole loop
Cost	Your existing tool subscription	Tool subscription or API credits
Best for	Occasional, one-off analysis	Ongoing research and content planning

Connecting Keyword Data: MCP Connectors Compared

MCP (Model Context Protocol) is the open standard that lets Claude use external tools. In Claude you add them from Settings → Connectors (or the “+” menu in a chat); in Claude Code you add them with the claude mcp add command. Here are the most common keyword data sources used with Claude in 2026:

Features and pricing change often — check each provider before you commit.

Data source	How it connects	What Claude gets	Pricing model	Good to know
Semrush	Official connector (Semrush plan with API/MCP access)	Volume, KD, intent, keyword gap, competitor rankings	Semrush subscription + API units	API units can run out fast — limit rows in prompts
Ahrefs	Official MCP (paid plan)	Keywords Explorer data, matching/related terms, rankings, backlinks	Ahrefs subscription	Large index; strong difficulty scores
DataForSEO	MCP server with API credentials	Volume, KD, CPC, intent, SERP results, ranked keywords	Pay-as-you-go credits	Low cost per query; popular with Claude Code users
Keyword Tool (keywordtool.io)	Connector directory or custom connector URL	Suggestions, volume, CPC, competition, Google & Bing data	Keyword Tool Pro plan, daily request limits	Quota shared between web, API and MCP
Google Search Console	CSV export (no official Google MCP)	Your real queries, clicks, impressions, CTR, position	Free	Re-export monthly; the most trustworthy data you have
Google Ads / Keyword Planner	CSV export or third-party MCP	Search terms, conversions, CPC ranges	Free with an Ads account	Best source for commercial keywords
Claude web search	Built in	Live SERPs, competitor pages, Reddit threads	Included	No volume data — use for qualitative research

Tip: Don’t connect every tool at once. Start with one keyword data source plus your Search Console export. Two consistent sources beat five that disagree.

Step-by-Step: Setting Up Claude for Keyword Research

Step 1: Create a Claude project

In Claude, create a new Project (for example “Keyword Research – myseoexperts.com”). Everything you add here — instructions, files and past chats — is available in every new conversation inside the project, so you don’t have to repeat yourself.

Step 2: Add project instructions

Instructions tell Claude your process, which data to trust and what the output should look like. Here’s a template you can adapt:

You are the keyword research assistant for My SEO Experts.

DATA SOURCES
- Live keyword data: the connected SEO MCP (Semrush / Ahrefs / DataForSEO).
- Search Console: the latest Queries.csv and Pages.csv in this project.
- Existing URLs: site-urls.csv in this project.
- Customer language: reviews, Reddit threads and call transcripts I paste in.

RULES
- Never invent search volume or difficulty. If a metric is missing, write "no data".
- Always say which data source and country each number comes from.
- Exclude branded and navigational keywords unless I ask for them.
- Return only the number of rows I ask for, to save API credits.
- If a tool call fails, tell me instead of retrying repeatedly.

OUTPUT
- Default to a table with: keyword | volume | KD | CPC | intent | funnel stage | cluster | new page or existing URL | priority | reason.
- Offer a CSV download for anything over 20 rows.

Step 3: Give Claude your business context

Generic input produces the same keyword lists your competitors get. Upload a short business context document that answers five questions:

1. What do you sell? Services, products, price points and what you don’t offer.
2. Who buys it? Industries, company size, job titles, locations.
3. What problems do you solve? In your customers’ words, not marketing language.
4. Who are your competitors? 3–5 domains that rank for what you want.
5. What are the rules? Topics to avoid, regulated claims, brand terms, target countries.

Also upload a list of your existing URLs (export your sitemap) so Claude can tell you whether a keyword needs a new page or belongs on one you already have — this prevents keyword cannibalisation.

Step 4: Connect your keyword data

• Connector: open Settings → Connectors, find your SEO tool (or add a custom connector URL), click Connect and sign in.
• Search Console: in GSC go to Performance → Search results, set the last 3 months, click Export → CSV and upload Queries.csv and Pages.csv to the project.
• Test it: ask “What is the search volume and keyword difficulty of ‘seo services’ in the United States? Tell me which tool you used.” If Claude names the tool and returns numbers, you’re connected.

8 Keyword Research Workflows With Claude (With Prompts)

Each workflow below includes a copy-paste prompt. Replace the text in [brackets] with your own details.

Workflow 1: Seed-to-cluster expansion

The classic starting point: turn a handful of seed keywords into validated, clustered opportunities.

Using the connected keyword tool (Google, United States), expand these seeds:
[seed 1], [seed 2], [seed 3]

1. Pull matching terms, related terms and questions for each seed.
2. Keep keywords with volume ≥ 50 and KD ≤ 40.
3. Classify intent (informational / commercial / transactional / navigational) and funnel stage.
4. Group them into topic clusters and name each cluster.
5. Show the 30 best opportunities in a table and tell me which cluster to write first and why.

Workflow 2: Competitor keyword gap analysis

Find keywords your competitors rank for that you don’t — the fastest way to discover proven demand in your niche.

Run a keyword gap analysis between myseoexperts.com and these competitors:
[competitor1.com], [competitor2.com], [competitor3.com]

- Find keywords where at least one competitor ranks in the top 10 and we don't rank in the top 50.
- Filters: volume ≥ 100, KD ≤ 45, exclude branded terms.
- List the top 10 per competitor, then up to 10 keywords ALL competitors rank for but we don't.
- For each keyword, suggest whether it needs a new page or fits an existing URL from site-urls.csv.

Workflow 3: Search Console quick wins

Your own Search Console data is the most accurate keyword data you have. Claude is excellent at finding the opportunities hidden inside it.

Analyse the attached Search Console Queries.csv and Pages.csv (last 3 months).
Find:
A) Queries ranking in positions 8–20 with 100+ impressions (striking distance).
B) Queries with high impressions but CTR below 1% (title/meta rewrite candidates).
C) Queries with impressions where no page clearly matches the intent (content gaps).
Output a table: query | page | position | impressions | CTR | issue type | recommended action.
Sort by estimated extra clicks if we reach position 3.

Opportunity type	What it means	Typical action
Striking distance (positions 8–20)	Google already sees the page as relevant	Improve content depth, internal links and title
High impressions, low CTR	You rank but searchers don’t click	Rewrite title tag and meta description
Impressions with no matching page	Demand you haven’t targeted directly	Create a new page or section
Two pages for one query	Keyword cannibalisation	Merge pages or differentiate their intent

Workflow 4: Mine customer language (reviews, Reddit and sales calls)

SEO tools only know about keywords that already have measurable volume. Your customers often search with phrases tools miss. Paste competitor reviews from G2, Capterra, Trustpilot or Google, Reddit threads in your niche, or anonymised sales call transcripts, and ask Claude to turn them into search queries.

Below are 40 customer reviews of competitors (pasted).
1. Group them into pain-point themes.
2. For each theme, write 5 search queries a frustrated customer would type into Google.
3. Quote the review snippet each query came from.
4. Then check each query in the connected keyword tool and add volume and KD (write "no data" if none).
Don't drop zero-volume queries — flag them as "validate manually".

Zero volume doesn’t mean zero demand. A phrase like “how to find freelance clients consistently” may show no volume in tools yet bring highly motivated buyers. Ask Claude for close variations — a slightly different wording often does show measurable volume.

Workflow 5: SERP and search intent analysis

Before writing, understand what Google is already rewarding for a keyword — including whether an AI Overview appears.

For the keyword "[keyword]":
1. Pull the current top 10 organic results and SERP features (AI Overview, featured snippet, People Also Ask, video, local pack).
2. Summarise the format and angle of each ranking page.
3. Tell me the dominant search intent and the best content format to compete.
4. Identify the angle or sub-topic none of the top 10 cover well.
5. List the People Also Ask questions we should answer on the page.

Use Claude to classify intent for every keyword in your list. This table shows the classification we ask Claude to use:

Intent	What the searcher wants	Example keyword	Best page type
Informational	To learn or solve a problem	what is local seo	Guide, blog post, video
Commercial investigation	To compare options before buying	best seo agency for small business	Comparison, list, case study
Transactional	To buy or hire now	hire local seo expert	Service page, pricing page
Navigational	To reach a specific brand or site	semrush login	Usually skip (not your brand)

Workflow 6: Keyword clustering at scale

Claude can cluster a few hundred keywords in a single pass. For larger lists, process in batches of about 200 keywords and ask Claude to merge the clusters at the end. Ask it to cluster by search intent and SERP overlap rather than just shared words — “seo cost” and “how much does seo cost” belong on the same page; “seo cost” and “seo cost calculator” may not.

Workflow 7: Turn clusters into content briefs

Turn the approved cluster "[cluster name]" into a content brief:
- Primary keyword, 5–10 secondary keywords and questions to answer
- Search intent and recommended format (guide, comparison, service page, tool)
- Suggested title tag (≤ 60 characters) and meta description (≤ 155 characters)
- H2/H3 outline based on the top-ranking pages, plus the unique angle we'll add
- Internal links: 3 pages from site-urls.csv to link from, and 3 to link to
- Schema type and any table, image or video the page needs

Workflow 8: The one-prompt content plan

Once your project is set up, you can run the whole process in a single request: “Find 30 keyword opportunities around [topic] with volume ≥ 100 and KD ≤ 35, cluster them, check our existing URLs, and give me a 3-month content calendar with a title and unique angle for each article.” Claude will call the connected tools, filter the results and return a plan you can review.

How to Prioritise Keywords With Claude

The biggest mistake in keyword research is sorting by search volume. A 50-search keyword with buying intent is often worth more than a 5,000-search informational one. Ask Claude to score keywords with transparent weights and to explain its reasoning, so you can challenge it.

Merge every validated keyword list from this chat into one table.
Score each keyword from 1–10 using:
- Business relevance (40%) — how closely it matches what we sell
- Intent (25%) — transactional > commercial > informational
- Difficulty vs our authority (20%) — lower is better
- Existing ranking (15%) — bonus if we already rank 8–20
Then place each keyword in the opportunity matrix (quick win / content cluster / fill-in / skip),
explain the top 10 scores in one line each, and export everything as CSV.

Example scoring model — adjust the weights to your goals.

Factor	Weight	What Claude checks
Business relevance	40%	Does the keyword match a service or product you actually sell?
Search intent	25%	Transactional and commercial intent score higher than informational
Difficulty vs. authority	20%	Can your domain realistically rank in the top 5?
Existing rankings	15%	Bonus for keywords already ranking in positions 8–20

Example of the final output

Here’s the kind of table Claude returns at the end of the process (figures are illustrative):

Illustrative example of Claude’s prioritised keyword output.

Keyword	Volume	KD	Intent	Cluster	Page	Priority
local seo services	2,400	38	Transactional	Local SEO	/local-seo-services/ (existing)	High
local seo checklist	880	22	Informational	Local SEO	New guide	High
how much does local seo cost	390	18	Commercial	Pricing	/pricing/ (add section)	High
google business profile optimization	1,300	35	Informational	Google Maps	/local-map-listings/ (existing)	Medium
seo for dentists	590	29	Commercial	Industries	New landing page	Medium
what is local seo	1,900	41	Informational	Local SEO	Section in guide	Low

Build a Reusable Keyword Research Skill

Claude Skills are folders containing a SKILL.md instruction file and, optionally, scripts, templates and reference files. Claude loads a Skill automatically when your request matches its description. A keyword research Skill means every team member — and every client project — gets the same process and the same output format.

---
name: keyword-research
description: Run keyword research for My SEO Experts or a client. Use when the user asks for keyword ideas, keyword gaps, clusters, search intent or a content plan.
---

# Keyword research

## Inputs to ask for (if missing)
- 3–5 seed keywords, the domain, target country and language
- Up to 3 competitor domains

## Steps
1. Pull matching terms, related terms and questions for each seed from the SEO MCP.
2. Pull the domain's ranking keywords in positions 4–20.
3. Remove branded, navigational and irrelevant terms (see exclusions.md).
4. Classify intent and funnel stage for every keyword.
5. Cluster by topic; map each cluster to an existing URL or "new page".
6. Score priority with the weights in scoring.md.
7. Export an .xlsx with tabs: All keywords, Quick wins, Clusters, Excluded.

## Rules
- Never estimate metrics yourself. Missing data = "no data".
- Process large lists in batches of 200 keywords.
- Flag ambiguous keywords for human review instead of guessing.

Add supporting files such as exclusions.md (terms to always remove, e.g. consumer or job-seeker queries for a B2B brand) and scoring.md (your priority weights). For different industries, keep separate exclusion lists — a fintech client and a dental clinic need very different filters.

Claude Code for keyword research

If you’re comfortable with a terminal, Claude Code takes this further. It can run your Skill, call the SEO MCP, write Python to merge and de-duplicate large CSV files, and save the final spreadsheet straight to your computer. It’s ideal for agencies running research for many clients or sites with thousands of pages. Pair it with a writing Skill and you can go from seed keywords to drafted briefs in one session.

Claude for Google Ads and Local SEO Keyword Research

Google Ads

Search terms reports are where Claude shines for paid search. Export the report (or use a Google Ads connector) and ask:

Attached is our Google Ads search terms report (last 60 days).
1. Find search terms with 3+ conversions and CPA below $40 that are not yet keywords — suggest them as exact match.
2. Find terms with spend above $25 and zero conversions — group them by theme as negative keyword candidates.
3. Flag any search terms triggering more than one ad group (cannibalisation).
Explain the expected impact of each change. Do not recommend pausing anything with fewer than 20 clicks.

Use 60–90 days of data, wait for enough clicks before acting on small keywords, and always apply changes yourself — Claude recommends, you decide.

Local SEO

For local businesses and multi-location brands, Claude can build keyword sets for every city and separate map-pack keywords from location-page keywords:

We are an SEO agency serving [city]. Using the keyword tool with location set to [city/country]:
- Find "service + city", "service + near me" and "best/top/affordable + service + city" variations.
- Add neighbourhood and nearby-town modifiers: [area 1], [area 2], [area 3].
- Separate keywords for the Google Business Profile (map pack intent) from keywords for location pages.
- Suggest one location page per city with its primary keyword and 5 supporting keywords.

Claude Keyword Research Prompt Library

Goal	Prompt to use
Find question keywords	“List 30 questions people ask about [topic], grouped by funnel stage, with volume from the connected tool.”
Find long-tail variations	“Give me long-tail variations of [keyword] with 3+ words, volume ≥ 20 and KD ≤ 20.”
Check keyword difficulty in bulk	“Score this list for difficulty and tell me which ones we can rank for with a DR of [x].”
Detect cannibalisation	“Using Pages.csv and Queries.csv, find queries where 2+ of our URLs get impressions.”
Find seasonal keywords	“Show the monthly trend for these keywords and tell me when to publish each one.”
Find AI Overview keywords	“Which of these keywords trigger an AI Overview, and what format do the cited pages use?”
Refresh old content	“Which of our pages lost the most clicks year-on-year, and which new keywords should we add to them?”
Find ‘vs’ and ‘alternative’ keywords	“Find ‘[competitor] vs’ and ‘[competitor] alternatives’ keywords with commercial intent.”
Translate keywords for new markets	“Find the keywords people actually use in [country/language] for [topic] — don’t just translate.”
Plan internal links	“For each cluster, suggest which existing pages should link to the new page and with what anchor text.”

Limitations, Costs and Mistakes to Avoid

Limitations

Limitation	Why it matters	How to handle it
No built-in keyword metrics	Unconnected numbers are guesses	Connect a data source or upload exports
Metrics are estimates	Volumes differ between Semrush, Ahrefs and others	Stick to one primary tool and note the source
Usage and API limits	Long chats and big pulls use up credits	Limit rows, batch large lists, start new chats for new tasks
Not a rank tracker	Claude answers on demand; it doesn’t monitor rankings	Keep your rank tracking tool
Can misjudge intent or relevance	Ambiguous keywords may be misclassified	Ask Claude to flag uncertainty; spot-check 20 keywords
No official Search Console connector	GSC data must be exported	Re-upload CSVs monthly

What it costs

You’ll need a paid Claude plan to use Projects, connectors and Skills comfortably, plus access to at least one keyword data source. At the time of writing, practitioners report DataForSEO as the cheapest way to start (pay-as-you-go credits, with each query costing a few cents), while Semrush, Ahrefs and Keyword Tool require their own subscriptions with MCP or API access. Check current pricing with each provider before you choose.

Common mistakes

• Trusting search volumes Claude gives without a connected data source.
• Skipping business context — and getting generic keywords your competitors already target.
• Sorting by volume instead of business value and intent.
• Ignoring Search Console quick wins in favour of brand-new keywords.
• Throwing out zero-volume customer phrases without testing variations.
• Creating new pages for keywords an existing page should cover (cannibalisation).
• Letting Claude make final decisions — or apply Google Ads changes — without human review.

Claude vs Traditional Keyword Research: Where the Time Goes

Task	Traditional workflow	With Claude + connected data
Pulling keyword ideas from seeds	Manual exports from one or more tools	One prompt; Claude calls the tool
Cleaning and de-duplicating	Spreadsheet formulas and manual checks	Automatic
Classifying intent	Keyword by keyword	Thousands of keywords in batches
Clustering	Manual grouping or a separate clustering tool	Automatic, with named clusters
Mapping to existing pages	Cross-checking your sitemap by hand	Automatic using your URL list
Writing briefs	Hours per brief	Minutes per brief, then human editing
Strategy and final decisions	SEO specialist	Still the SEO specialist

Need Help With Keyword Research?

At My SEO Experts, we combine Claude-powered workflows with professional SEO tools and years of hands-on experience to build keyword strategies that bring in leads, not just traffic. We’ll research your market and competitors, find quick wins in your Search Console and deliver a prioritised content plan your team can act on.

Get a Free Keyword Research Quote →

Frequently Asked Questions

Can Claude do keyword research?

Yes, with one important condition. On its own, Claude is excellent at brainstorming seed keywords, classifying search intent, clustering and writing briefs, but it has no built-in search volume or keyword difficulty data. Connect it to an SEO tool (Semrush, Ahrefs, DataForSEO, Keyword Tool and others) through an MCP connector, or upload exported CSVs, and it becomes a complete keyword research assistant.

Is Claude better than ChatGPT for keyword research?

Both can brainstorm and cluster keywords. Claude’s strengths are long context (it can analyse large keyword exports and transcripts in one go), Projects for saved business context, Skills for repeatable workflows and wide MCP connector support. The quality of your results depends far more on the data you connect and the instructions you give than on the model you choose.

Can Claude give me accurate search volume?

Not by itself. Any search volume Claude gives without a connected data source is a guess. With a connector, Claude reports the numbers from that tool — which are themselves estimates and differ between providers. Always say which data source you used.

What is an MCP connector?

MCP (Model Context Protocol) is an open standard that lets Claude call external tools. An SEO MCP connector lets Claude look up keyword metrics, competitor rankings and SERP data while you chat, instead of you exporting and pasting spreadsheets.

Does Claude connect to Google Search Console?

At the time of writing there is no official Google Search Console connector from Google. The simplest method is to export the Performance report (Queries and Pages) as CSV and upload it to your Claude project each month. Some third-party MCP servers offer Search Console access — review their permissions carefully.

What is a Claude Skill for keyword research?

A Skill is a folder with a SKILL.md instruction file (and optional scripts or templates) that teaches Claude a repeatable process — for example: pull data for these seeds, classify intent, score priority and export a spreadsheet in this exact format. Once installed, Claude uses it automatically whenever you ask for keyword research.

Can I use Claude for Google Ads keyword research?

Yes. Claude is very good at analysing search terms reports: finding new exact-match keywords from converting queries, suggesting negative keywords, grouping terms into tight ad groups and flagging cannibalisation. Keep a human in charge of applying changes to the account.

Will AI keyword research replace SEO tools or SEO experts?

No. Claude replaces a lot of manual spreadsheet work — cleaning, de-duplicating, classifying and clustering — but it still relies on SEO tools for data and on human judgement for strategy, business priorities and final decisions.

Conclusion

Claude won’t replace your SEO tools — but it can replace most of the manual work between them. Give it real data through MCP connectors or exports, give it your business context through a Project, make the process repeatable with a Skill, and keep a human in charge of the final decisions. That combination finds better keywords, faster, than either a tool or an AI could on its own.

Further reading: Semrush – How to do keyword research with Claude · Keyword Tool MCP · Claude documentation · Model Context Protocol.