How to build a competitor SEO analysis workflow in Claude that actually ships insights.

This is the workflow that impressed us most among the DIY approaches. Install the Playwright MCP (for browser-based site crawling) and the SerpAPI MCP (for live search result data), verify both are connected by asking Claude to list its available tools, then fire a single prompt that tells Claude to crawl your competitor’s top pages AND pull their ranking keywords in one session.

The output we received in testing: a ranked table of 47 keyword gaps sorted by search volume, a content-structure breakdown of the competitor’s five top-ranking pages, and a prioritised brief for our first three new articles — all in under 25 minutes. The setup takes about 15 minutes the first time. After that, you can re-run the same prompt weekly. For more on connecting Claude to live data sources, see our guide on how to connect Claude to Google and Meta Ads via MCP.

Where Playwright requires a local browser environment, Firecrawl handles page scraping via API, making this combination easier to run in cloud environments or hand off to a less technical team member. Perplexity’s search API gives Claude access to real-time web results so it is not working from stale training data when it assesses what is currently ranking.

In our tests, this combination produced excellent content gap analysis — identifying 38 topics a target competitor covered that our test site did not, grouped by funnel stage. Where it fell short was precision keyword ranking data: Perplexity’s search results give a directional picture, not position-by-position accuracy. For high-stakes keyword decisions, pair this with a quarterly Ahrefs export.

This approach — popularised by the AI Agent Strategy blog — starts with a simple markdown file (drafts/competitors.md) listing each rival’s domain, category (direct, adjacent, adjacent-use-case), and key pages like /pricing, /features, /blog, and /customers. Upload this into a Claude Project and every session starts with that competitive context pre-loaded. Five to eight competitors is the optimal range — more than that and the analysis gets shallow.

The limitation is that Claude is working from its training data plus whatever you paste in — not live SERP results. For a first competitive audit or for teams that update the file quarterly, this is a strong low-friction starting point. For live rankings, layer in SerpAPI or pair it with a periodic Ahrefs CSV export (approach #5 below). See our related guide on targeting competitor keywords with Claude AI for the exact prompts that work best inside Claude Projects.

If your team already runs Ahrefs or SEMrush, you are sitting on a gold mine of competitor data that most teams barely skim. Export the keyword gap report, the top pages by traffic for each competitor, and the backlink profile as CSVs. Upload them to Claude and prompt it to analyze the keyword profiles of your three main competitors against your domain, identify the top 10 untapped opportunities sorted by search volume and ranking difficulty, and produce one-paragraph content briefs for each.

In our testing, this took 8 minutes from CSV upload to a formatted brief document with 12 prioritised keyword targets. The catch is freshness: you are working with data from the last time you exported, which for most teams is weeks ago. Run this monthly at minimum, or combine it with a live-data MCP for the best of both worlds. The keyword gap analysis workflow alone typically uncovers 100–200 new keyword opportunities per competitor, with 15–25% being immediately actionable according to our test runs.

This approach — detailed in the GenAI Unplugged Content OS series — builds a Claude Code subagent that combines Perplexity search with Firecrawl scraping and runs on a structured knowledge base you create once. The agent reads your competitor watchlist (three to five names per category, direct and adjacent), your business context, and your content strategy before it starts any analysis, so every output is filtered through “is this relevant to us?” rather than dumping raw competitive data.

The output quality was the highest of any DIY method in our testing: a weekly brief that flagged two new competitor blog clusters we had not noticed, a backlink opportunity from a resource page one rival had just earned, and a content gap report grouped by funnel stage and keyword difficulty. The setup takes 45–60 minutes to build the workspace files properly. After that, the agent runs itself. For the specific prompt structures that power this, our competitor keyword targeting guide covers the keyword gap analysis prompts in detail.

Before you analyze competitors, you need to know where you stand. Export your Google Search Console performance data (queries, pages, clicks, impressions, average position) and paste it into Claude alongside a competitor’s top pages list from Ahrefs or a manual check. Ask Claude to identify which competitor keywords your site is showing up for at positions 11–30 (the “striking distance” zone) and which competitor topics you have no coverage of at all.

In 10 minutes you get a two-part output: a list of pages worth a quick on-page SEO push (because you are already close), and a list of content gaps to add to your editorial calendar. This is the lowest-friction entry point in the entire list and pairs well with any of the above approaches. The limitation is that GSC only shows your own data — you still need a separate source for competitor keyword data, which is why this works best as the “know your baseline” step rather than a standalone workflow.

The simplest entry point: a library of structured prompts you run directly in Claude.ai. Prompts like “Analyze the keyword profiles of [Competitor 1], [Competitor 2], and [Competitor 3]. Identify keywords they likely rank for that my website [Your Website] does not, categorized by search intent and ranking difficulty. Highlight the top 10 opportunities to target first.” — or “Examine the top five ranking pages from [Competitor] for [Target Keyword]. Break down their content structure, estimated word count, heading hierarchy, and content elements (images, videos, tables). What patterns make these pages successful?”

These prompts work surprisingly well for content structure analysis, messaging pattern identification, and understanding broad keyword themes. They fall down on precision: without live data, Claude is estimating based on training knowledge. Use this for directional research and ideation, then validate with a tool that pulls real SERP data before investing in content creation. The full prompt library for competitor keyword targeting on our blog covers 10 more workflow variations.

Max Mitcham’s widely shared workflow used the Trigify MCP to set Claude Code monitoring a competitor across four social platforms simultaneously — Reddit, X, LinkedIn, and YouTube — analyzing 165 posts in minutes for positioning themes, product launches, pricing signals, and audience frustrations. The agent saves a structured knowledge.md file that becomes the input for every future competitive session.

For SEO purposes, this workflow is most valuable for surfacing content angle opportunities: if a competitor’s LinkedIn posts about a particular pain point are getting 10x their normal engagement, that is a strong signal the topic has demand worth targeting with organic content. It does not replace keyword-data workflows but adds a layer of qualitative signal that spreadsheets miss entirely. Pair it with approach #2 or #3 for both the quantitative and qualitative picture.

Ad Age and Adweek both covered the wave of agencies in 2026 using Claude Code to vibe-code custom competitive intelligence apps for clients. The workflow: use Claude Code to build a bespoke application that monitors how a brand and its competitors appear in AI-generated search answers (GEO), tracks keyword rankings, and generates branded reports. Some agencies are packaging these tools and selling them as standalone SaaS products to clients.

The output potential is the highest in this list — fully custom, client-branded, and productized — but so is the investment. This is not an approach you run on a Tuesday afternoon. It suits agencies with an enterprise Claude agreement and a developer comfortable enough with Claude Code to build and deploy a web app. For most teams, the ROI calculus favors an autonomous platform like Ryze AI that delivers the same ongoing intelligence without the build cost. For a deeper look at how agencies are using Claude for paid channels alongside SEO, see our guide on connecting Claude to Google and Meta Ads via MCP.