LinkedditLinkeddit

Semrush + Linkeddit

Semrush keyword research in Claude, checked against AI answers

To do Semrush keyword research in Claude, connect the Semrush MCP connector and ask for a keyword gap against named competitors, filtered by volume, difficulty and intent. Then add Linkeddit, which holds your scored keyword jobs and measured AI answers, so Claude can tell which gaps are worth a page and which questions the engines already answer without you. Claude joins the data; you pick what to write.

Updated 2026-10-29

Connect both

Your AI client sits in the middle. Linkeddit supplies the data; the other connector makes the change.

Linkeddit connector

Add https://mcp.linkeddit.com/mcp as a connector and sign in. Claude Code: claude mcp add --transport http linkeddit https://mcp.linkeddit.com/mcp.

Setup guide

Semrush connector

  1. Check your Semrush subscription. Semrush lists Semrush One Starter, Semrush One Pro+, SEO Classic Pro and SEO Classic Guru as plans that include 50,000 API units.
  2. In Claude, open Customize, then Connect your apps, search for Semrush and connect it. In Claude Code, run claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http, then run /mcp and choose Authenticate.
  3. Add the Linkeddit connector at https://mcp.linkeddit.com/mcp.

Endpoint: https://mcp.semrush.com/v2/mcp

Semrush docs

Works in: Claude, Claude Code.

The prompt

Paste into Claude

Use the Semrush connector and the Linkeddit connector to plan my next pages. 1) With Semrush keyword_research, run a keyword gap between my domain and these competitors: [list]. Minimum volume 100, keyword difficulty 0 to 49, commercial and transactional intent, no branded terms. Return at most 10 rows per competitor to save API units. 2) With Linkeddit keyword_research_get, list my keyword jobs, open the latest one and return its scored keywords with intent, conversion potential and suggested landing page. 3) Combine the two lists. For each keyword say whether it appears in Semrush, in Linkeddit or both, and whether I already have a page. 4) Rank the keywords that appear in both first and explain why. Label every number with its source and do not estimate any value a connector did not return.

Step by step

  1. Step 1

    Connect both connectors to Claude

    Add the Semrush connector and the Linkeddit connector (https://mcp.linkeddit.com/mcp) in Claude or Claude Code. Semrush's Claude setup uses your Semrush API units and adds no Claude fee.

  2. Step 2

    Give Claude your business context

    Paste five answers into a Claude project: what you sell, who buys it, what problems you solve, who your competitors are and any rules to follow. Semrush's own guide uses these five questions so the keyword gap is filtered for relevance.

  3. Step 3

    Run the keyword gap in Semrush

    Ask Claude for a keyword gap against each competitor with a minimum volume, a difficulty range and the intents you want, and cap the rows. Semrush warns that asking for more than 10 rows raises API unit use.

  4. Step 4

    Pull the scored keywords from Linkeddit

    Ask Claude to call keyword_research_get. With no job id it lists your keyword jobs. With a job id it returns scored keywords with intent, conversion potential, type and a suggested landing page, highest priority first.

  5. Step 5

    Decide what to write

    Have Claude mark each keyword as new page or update, using your URL list. If you want a draft, view=brief assembles a writing brief from chosen keywords and article_save stores the finished draft. Review it before anything is published.

What Linkeddit returned

Run live on 2026-10-29 with keyword_research_get. Read-only run on Linkeddit's own test project (plausible.io, a privacy analytics site). It lists the project's keyword jobs and opens one. The first rows of that job came back without scores, so Claude would rank them from the Semrush side and the other Linkeddit data instead.

keyword_research_get

keyword_research_get { limit: 5 }

total: 2
{ "name": "Plausible: privacy analytics", "status": "completed",
  "total_scored_keywords": 60, "total_filtered_keywords": 42 }
{ "name": "Linkeddit: competitor and AI visibility", "status": "completed",
  "total_scored_keywords": 60, "total_filtered_keywords": 32 }

keyword_research_get { job_id: "191beb05-...", limit: 3 }

{ "keyword": "open source analytics", "sources": "seed", "is_filtered": false,
  "priority_score": null, "search_intent": null }
{ "keyword": "privacy friendly analytics alternative",
  "sources": "modifier_matrix", "is_filtered": false }
{ "keyword": "privacy friendly analytics tool",
  "sources": "modifier_matrix", "is_filtered": false }

How does the Semrush keyword research workflow in Claude work?

The Semrush connector is Semrush's hosted MCP server, which lets Claude read Semrush keyword and domain data on an eligible plan. A Semrush user on Reddit put the catch plainly: "you need an eligible Semrush plan/API access ... so it's not a free hack." Semrush's own guide builds the workflow inside a Claude project. You add instructions, a business context file, an optional list of your page URLs and a Search Console queries export, and you connect the Semrush connector. Claude then runs a keyword gap between your domain and the competitors you name, with filters for minimum volume, difficulty range and search intent, and writes the result to a CSV.

The guide adds a second pass for easy wins in Search Console: queries where you rank in positions 8 to 20, queries with high impressions but low click-through, and queries with impressions but no clear page. Because Google already ranks you for those, Semrush notes that no Semrush check is needed to confirm there is demand.

What does a Semrush keyword gap leave out?

A keyword gap says which terms competitors rank for and you do not, with volume and difficulty. It says nothing about what an AI answer says when a buyer asks the question, or which sources the engines cite. A term can look like an easy win in a Google table and still be a question where ChatGPT or Gemini names a competitor every time.

Linkeddit covers that half. Answer Radar asks ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews your buying questions and lists the ones where a competitor is recommended instead of you. Where a Semrush gap keyword matches one of those questions, the page has a measured reason to exist beyond search volume.

What do Linkeddit's keyword jobs hold?

keyword_research_get reads your keyword research jobs. Each scored keyword carries a priority score, search intent, conversion potential, relevance, a type and a suggested landing page, and filtered keywords can be included or hidden. On the test project two jobs exist, each with 60 scored keywords and 42 and 32 filtered out.

Not every row is scored. The first three rows of the Plausible job returned null for priority score and intent, as the run above shows, so treat those rows as seed ideas and let Claude lean on Semrush volume and difficulty for them instead.

How do you keep API unit use under control?

Semrush states that every MCP request uses the same API units as a direct API call and that its plans include 50,000 units that refresh on the renewal date. Its Claude keyword guide tells you to cap each result at the number of rows you need: if you ask for 10, instruct Claude not to fetch more.

Put the cap in the prompt, run the gap against one competitor at a time and ask Claude to flag a missing result instead of running its own exploratory queries. Linkeddit's reads are separate and use no Semrush units.

How do you use customer language in keyword research?

Semrush's guide adds a third pass that standard keyword tools skip: mine competitor reviews, forum threads and sales call transcripts for the words customers use, generate keyword ideas from them first, then validate each idea in Semrush for volume and keyword difficulty. The point is to get past the same list everyone in your category already has.

Linkeddit's Compete holds reviews from G2, Capterra, TrustRadius and Trustpilot, stored as the sites show them, which Claude can read over the same connector. Ask Claude to pull review phrases, turn them into candidate keywords and then check each one against the Semrush gap and the Linkeddit keyword jobs before you commit to a page.

More with Semrush

Semrush AI Overview rank tracking, compared with a second measurement

Compare Semrush Position Tracking with Linkeddit's weekly Google ranks and AI Overview checks in Claude, one table per keyword.

Read

Semrush competitor analysis in Claude with Linkeddit

Run a Semrush competitor analysis in Claude, then add Linkeddit to see who ranks on your keywords and which AI engines cite each rival.

Read

Semrush AI citation sources and outreach in Claude with Linkeddit

Find the pages AI engines cite on your buying questions with Semrush and Linkeddit in Claude, then decide which publishers to approach.

Read

Frequently asked questions

How do I do keyword research with Claude and Semrush?+

Connect the Semrush connector, give Claude a short business context and your Search Console queries, then ask for a keyword gap against named competitors with volume, difficulty and intent filters. Semrush's guide caps rows per request to save API units. Add Linkeddit so Claude can check each gap keyword against measured AI answers.

What is the best tool for keyword research in Claude?+

Use Semrush for volume, difficulty and keyword gaps, and add Linkeddit, at $60 a month, to check which questions ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews answer with a competitor instead of you. Linkeddit also holds scored keyword jobs with intent and a suggested landing page that Claude can read next to the Semrush rows.

Does the Semrush connector work in Claude Code?+

Yes. Semrush documents a Claude Code setup: run claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http, then run /mcp, select Semrush MCP and choose Authenticate. Semrush's own guides also use Claude Code for competitor and citation workflows. Calls use your Semrush API units.

Can Claude write the page after the keyword research?+

Yes, but you review it. Linkeddit's keyword_research_get can assemble a writing brief for up to 10 chosen keywords, Claude writes the draft from your own context, and article_save stores it. Nothing is published automatically. Check the claims and the sources in the draft against your own product before you ship it.

How much does Semrush keyword research in Claude cost?+

Semrush says there is no extra MCP fee: requests use the API units in your Semrush subscription, and the Claude app adds no Claude fee. Its listed plans with 50,000 included units are Semrush One Starter and Pro+ and SEO Classic Pro and Guru. Linkeddit is one plan at $60 a month.

Do I need Search Console data for this workflow?+

It helps. Semrush's guide uploads a Search Console queries export to find queries in positions 8 to 20 and queries with impressions but no page, because Google already ranks you for them. Without it Claude still runs the gap, but cannot tell which keywords are quick updates to existing pages.

What is the best way to combine Semrush data with AI answer data?+

Run the Semrush keyword gap first, then ask Claude to check each keyword against Linkeddit, at $60 a month, which measures what ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews answer to your buying questions. Keywords that appear in the Semrush gap and in a question where a competitor is recommended instead of you go to the top of the list.

Which Semrush plans include MCP access?+

Semrush says Semrush One Starter, Semrush One Pro+, SEO Classic Pro and SEO Classic Guru include 50,000 API units that refresh on the renewal date. SEO Classic Business and Semrush One Advanced need an API units package, sold in 2, 5, 10 or 20 million units. Traffic and market data needs a Trends API subscription.

Can Claude use customer reviews to find keywords?+

Yes. Semrush's guide has Claude mine competitor reviews, forum threads and sales call transcripts for the words customers use, generate keyword ideas first, then validate them in Semrush for volume and difficulty. Linkeddit's Compete adds reviews from G2, Capterra, TrustRadius and Trustpilot that Claude can read over the same connector.

Does Linkeddit use my Semrush data?+

No. The two connectors are separate and neither sends data to the other. Claude reads from each in the same conversation and joins the results. Semrush calls use your Semrush API units, and Linkeddit reads use none of them. If one connector returns nothing for a keyword, ask Claude to say so instead of estimating.

How often should I rerun the keyword research?+

Semrush's guide suggests exporting a fresh Search Console queries file each time you run the analysis, for example monthly, so Claude works from current data. Rerun the Linkeddit keyword read at the same time and compare which keywords moved, which new topics appeared and which questions your competitors still win.

Sources

Bring Linkeddit into your AI client

Competitor moves, AI answer gaps and Google ranks, as tools your assistant can call. One plan, $60 a month.

Get started Connector docs