# How We Use Claude Code to Run Our Own Marketing (No Agency) > We are an AI studio that does not hire a marketing agency. Here is the exact Claude Code workflow we use to run SEO, content, and outreach -- and what it actually takes to make it work. - **URL**: https://www.maximalstudio.in/blog/claude-code-own-marketing --- Maximal StudioApproachResourcesBlogToolsGet In Touch<- Back to BlogHow We Use Claude Code to Run Our Own Marketing (No Agency)Jun 29, 2026-Shubham RasalWe are an AI studio that does not hire a marketing agency. Here is the exact Claude Code workflow we use to run SEO, content, and outreach -- and what it actually takes to make it work.We build AI products for clients. We don't have a marketing team, a content manager, or an agency. What we have is Claude Code, a few connected tools, and a set of workflows we've built and refined over the last six months. This is a post about what we actually do -- not what's theoretically possible with AI marketing, but the specific workflows that run our blog, our outreach, and our SEO without a human owning each step. The Honest Starting Point Most "AI replaces marketing" content is about what an AI tool can do. This post is about what we actually use, including what doesn't work as well as it sounds. We use Claude Code for: Identifying what to write (GSC + DataForSEO data -> brief) Drafting blog posts Generating post images (Nano Banana, Excalidraw style) Personalizing outreach emails Monitoring what's ranking and what's slipping We do not use AI for: Deciding strategy (still human judgment) Editing drafts (we read everything before it goes out) Client relationship communication That line matters. AI writes the first draft of everything. Humans decide if it goes out and edit it until it's worth publishing. The Content Workflow Every blog post starts the same way. Claude queries our GSC data via the MCP server and finds the keyword gaps -- queries getting impressions but low clicks, or topics we should rank for but don't. 1. GSC MCP -> pull 90-day impression data 2. Identify page-2 keywords (pos 11-30, impressions >200) 3. Cross-check against existing posts 4. Generate brief for the top gap 5. Draft the post from the brief 6. Nano Banana -> generate Excalidraw-style images 7. Human review -> edit -> publish Step 1-6 takes Claude about 4 minutes. Step 7 takes us 20-40 minutes. The ratio is good. The Outreach Workflow When we identify a potential client (from inbound or from a list), we don't write cold emails from scratch. Claude researches the company from their website, identifies a specific pain point their content reveals, and drafts a 3-sentence email that references it. Input: company URL Output: 3-sentence email with: - One specific observation about their current approach - One sentence on how we'd approach it differently - One ask (call, reply, specific question) Response rates are meaningfully higher than template emails. The difference is that the specificity is real -- Claude actually read their site. The SEO Monitoring Workflow Every Monday, a Claude agent pulls our GSC data and produces a one-page report: what moved up, what slipped, what new queries appeared, and what to do about each. It takes 90 seconds and replaces a task that used to take an hour. The report gets posted to a Slack channel. If something significant moved, we respond. If it's stable, we move on. What Actually Makes This Work The workflows above sound simple. They are, once built. What makes them work in practice: Good prompts take iterations. The GSC brief prompt we use now is version 8. Earlier versions produced briefs that were too generic to be useful. Getting to useful required testing against real keyword gaps. Connected tools matter more than smart prompts. A Claude prompt that can pull live data from GSC, check DataForSEO, and write to a Google Sheet is dramatically more useful than a Claude prompt working from a copy-pasted spreadsheet. The MCP connections are the leverage. Human review is not optional. We've published posts that were 90% Claude's draft. We've also edited drafts down to 20% of what Claude wrote. The human step is where voice and judgment live -- you can't remove it without losing what makes the content worth reading. Want to Build This for Your Team? The workflows above took us months to build and refine. If you want them running for your product without building from scratch -- talk to us. We can scope what your marketing automation stack should look like and build it.Keep exploringWork with usBuild your AI product ->We ship AI integrations, dev tools, and full products for teams.Free ToolsAI Calculators & Utilities ->ROI calculator, LLM cost estimator, workflow tools.Case StudiesReal-world AI builds ->See how we've shipped AI automation for real businesses.BlogMore posts ->Practical guides on AI, automation, and building fast.Maximal StudioAI & automation for builders.PagesToolsBlogCase StudiesApproachResourcesOfficeIndiaBangaluru, Karnataka, IndiaConnectLinkedInXEmail© 2026 Maximal Studio. All rights reserved. --- Maximal Studio builds custom AI systems, automations, and growth tooling.