What I use Claude Code for Marketing (and as a marketer):

• Built custom agents to track Firecrawl mentions and content ideas from Reddit/HN/Twitter
• Auto-refresh decaying blog posts (Ahrefs → Slack → Claude fixes it)
• Internal linking agent that replaced a $30/month SaaS tool
• Scraping competitor data and research with Firecrawl integration
• File conversions, image optimization, FAQs, site speed audits
• Tracking AI visibility through Vercel bot crawl data
• Managing a chaotic research file that feeds into all my content

Note: You need to learn tech stack basics, get engineers to review your work (especially for customer-facing changes), and keep yourself in the loop. It's not faster than a developer, but it removes dependencies. I can ship without waiting on anyone.

I lead SEO at Firecrawl, and I'm not exaggerating when I say my entire workday runs on Claude Code for marketing now.

Back when I started in marketing, getting a single landing page out took weeks. Brief a designer. Wait. Review mockups. Wait. Brief a developer. Wait. Review implementation. More fixes. More waiting.

Now? If I know exactly what needs to go on a page, I can just... ship it. No waiting on anyone. No dependencies. Just done.

And it's not just landing pages. We're in this era where everyone's building their own personal apps and agents, right? For me, Claude Code is at the center of all of it. Let me walk you through how I actually use it day-to-day as a marketer at Firecrawl.

Claude Code is basically Claude in your terminal. It's a CLI tool that can read your codebase, make changes across multiple files, run commands, and basically act like a dev on your team.

Unlike the web version where you're copying and pasting code back and forth, Claude Code actually lives in your project. It can see your entire repo, understand context, and make changes directly. You can even give it access to your local files and make changes to your system - like deleting unwanted screenshots from your desktop.

Getting started with Claude Code:

1. Install it:

curl -fsSL https://claude.ai/install.sh | bash

2. Run it in your project:

claude

3. Start chatting with it like you would with a developer

Pro tip: If terminals intimidate you, use Claude Code from the desktop app instead (that's what I do). You can drag and drop screenshots, paste images, and interact with it like a normal chat app.

This post went viral recently and got me thinking - maybe I should actually write about how I use Claude Code for marketing:

So here's the breakdown of what I actually do with it:

I've completely moved away from "I wish there was a tool for [xyz] problem" and instead just build my own solution in usually less than an hour now.

This one's pretty straightforward but so useful. It's just a simple agent that uses the Firecrawl Agent under the hood. I even had Claude Code help me refine the prompt because my first version needed work.

Every Monday, I get this nice little report of interesting Firecrawl mentions from the past week. We're talking Reddit threads, Hacker News discussions, Twitter shoutouts, random niche forums - basically anywhere people are talking about us.

Why do I care? When I find these great mentions - like someone sharing how they used Firecrawl to solve a real problem or praising a specific feature - I work them into my content that week. It's way more credible to reference actual users than to just make claims.

I also have a similar setup tracking broader discussions around web scraping, web search APIs, data extraction, etc. If I see developers asking specific questions on Reddit this week, boom - that's my content roadmap for next week. Write what people are actually asking about.

Okay this one's a bit more involved but it saves me hours every week. I have Ahrefs hooked up to Claude, and there's a cron job that checks my top 10 decaying blog posts and pings me on Slack.

From Slack, I just pick like 3 posts that feel high-leverage to optimize. I tell Claude exactly what needs refreshing, and it handles the grunt work - updating old stats, refreshing code examples, adding sections for new developments, fixing internal links.

Now, I'm not just letting it run wild and auto-publishing everything. I review every single change before it goes live. Fully autonomous content is usually terrible and definitely doesn't match our brand voice. Human-in-the-loop is key here.

I used to pay $30/month for this SaaS tool that was supposed to find interlinking opportunities. It was... fine? But honestly the suggestions were kinda meh. I kept paying for it because I didn't have a better option.

Until I realized I could just build my own.

Since Claude Code has access to our entire marketing website repo, it can actually suggest smart internal links based on what makes sense semantically, not just keyword matching. It knows which anchor text sounds natural, which pages we want to prioritize, and how to avoid making it look spammy.

It's honestly so much better than the generic tool, and it's tuned exactly to how we think about our content strategy. Plus, you know, free.

Obviously we have Firecrawl integrated with Claude Code at Firecrawl (would be weird if we didn't), and it's kind of ridiculous how useful this is. We're actually an official Claude plugin now, which makes the integration even smoother.

People assume Firecrawl is just a scraping tool - but it's so much more. It can act as the complete web data layer for your AI workflows.

Using Firecrawl within Claude Code, I can research and pull data from websites and PDFs easily now. I'll tell Claude to grab screenshots from research papers or competitor sites and add them to blog posts. Or scrape real-time data for competitor analysis.

Or fact-check claims by pulling the actual source material. You can also search for emerging trends in your industry, track what competitors are talking about, or analyze what topics are getting traction across different platforms.

Example: When I'm writing a comparison article, I literally just say "Use Firecrawl to scrape the pricing pages of these 5 competitors and make a comparison table." It handles all the annoying scraping and data extraction stuff, and I can just focus on actually analyzing and writing about it.

I also use Firecrawl /agent to find real FAQs for every blog post. The prompt is pretty simple:

Use Firecrawl Agent to find common questions [target audience] are asking about [topic] on [wherever your ideal customers hang out] in the last 3 months.

Then I just answer those questions in my content. Super helpful because you're writing stuff people actually want to know, not just what you think they might search for.

File conversions used to drive me insane. Getting the right image format, compressing things properly, making sure dimensions were right - just tedious busy work.

Now it's all automated. Everything gets optimized with the right formats without me having to think about it. And this actually matters for SEO:

• Images load faster (page speed++)
• WebP format cuts file sizes by like 30-40%
• Right dimensions = no weird layout shifts
• Alt text gets optimized in the same go

Just last week I did a site health audit and Claude threw like 20+ suggestions at me. Lazy loading images, optimizing JS bundles, cutting down third-party scripts - all the good stuff.

I pinged an engineer on the team real quick to verify I wasn't about to break anything (always do this!!), and then just went down the list fixing things.

Claude's actually pretty good at looking at data and finding interesting patterns.

Here's something I've been playing with: I pull our Vercel observability data to see what pages common bots like ChatGPT-User are crawling (I check for common bots at bots.fyi). Then I feed that to Claude and ask it to figure out what questions users might actually be asking that would surface this content.

Since Claude Code has access to the entire website repository, it can go over these articles and give suggestions for optimization.

I use those questions in Profound to track our AI visibility.

Is this foolproof? No. It's pretty speculative. But honestly, the entire AEO (Answer Engine Optimization) space right now is just educated guessing. At least this is speculation with some guardrails based on actual crawl data, so... ¯_(ツ)_/¯

Whenever I come across something interesting - stats, research papers, whitepapers, podcast clips, cool projects, random good X posts, whatever - I just dump it in a local file with a quick note. This is my personal repository of interesting things I read/see online.

Before I publish any content piece like a blog article, I tell Claude: "Go through this file and see if anything's relevant to this article. If yes, work it in naturally."

It's basically like having a research assistant with perfect memory who never loses track of that one stat you read three months ago and now desperately need.

There's so much I still want to experiment with Claude Code:

• Auto-documenting my experiments in Notion - right now I manually write up all my SEO experiment hypotheses and results and it's tedious
• Dynamic content personalization - adjusting content based on traffic source or user behavior
• Automated competitive intelligence - weekly reports on what competitors are publishing, their backlinks, ranking changes, etc.
• Creating blog cover images easily - generating custom cover images with a simple prompt instead of spending time designing them

The possibilities are kinda wild to be honest.

Seriously, every time you do something and think "I'll probably need to do this again," make it a skill. And keep updating that skill each time you use it.

This is especially clutch for maintaining your writing style. I literally have a hiba-writing-style skill file.

And you don't have to write this yourself. Once you've completed a task, ask Claude to document the process and expected output and create a skill. And every time you use it, ask Claude to update it for you.

Skills are like building your own custom tools that work exactly how you need them to.

The people who tell you you don't need to know code at all to vibe code are lying. You don't need to become a developer, but you should understand your tech stack basics. Otherwise you will make a mess.

AI always finds the easiest path to solve a problem - even if it's hacky. But the easiest path isn't always the best path.

What I did: I asked Claude "what do I absolutely need to know to build a solid, high-performing Next.js website?" Then I went and learned those concepts. Now, I have a CS degree so this was easier for me, but the point is - you need to understand what Claude actually did. Ask it to ELI5 literally everything.

Learn at least:

• Basic web stuff (HTML, CSS, how JavaScript works)
• Your tech stack (Next.js, React, whatever you're using)
• SEO fundamentals (meta tags, sitemaps, structured data)
• How your deployment works

Not everything Claude suggests is actually the best solution. If it's changes to your product or website or anything customer-facing, just shoot a quick message to an engineer on your team before you ship.

I learned this one the hard way. I pushed what I thought was a "performance optimization" that completely broke mobile rendering. Luckily an engineer caught it in staging, but if that had hit production? Yikes.

The review doesn't need to be formal - literally just "hey does this approach make sense?" can save you from a disaster.

Think about what you do every week that just feels like soul-crushing busy work. That's what you should automate first.

Make a list:

• Compiling reports
• Updating spreadsheets
• Checking competitor rankings
• Reformatting content for different platforms

If you're doing it every week, automate it. You have better things to do with your time.

At least right now, human-in-the-loop is way better than trying to make everything fully autonomous. Your marketing needs personality, and AI doesn't really have that.

Let AI handle the grunt work - drafting, suggesting, automating the boring stuff. But you're the one adding:

• Your brand voice
• Strategic decisions
• Editorial judgment
• Creative direction

The goal isn't "AI does everything." It's "AI does the tedious stuff so I can focus on the things that actually matter."

Okay let me be honest about the downsides because it's definitely not perfect:

It's not always faster. A good developer will still be faster than me for most tasks. The win here isn't speed - it's independence. I can ship without waiting on anyone.

You will make mistakes. AI-generated doesn't mean correct. Test everything, review everything, and get engineers to check your technical stuff.

There is a learning curve. My first week was frustrating as hell. I broke things in dev environments. Made dumb mistakes. But that's just part of learning something new.

Note: Claude Code doesn't replace strategy. Claude can't tell you what landing page to build or what content angle to take. It executes your vision. The actual strategy, positioning, creative stuff? Still 100% on you.

I really do spend my entire workday using Claude Code for marketing now. And this is just scratching the surface - there's so much more potential here that I haven't even explored yet.

If you're in marketing and think Claude Code for marketers isn't relevant to you because it's just for engineers, you're leaving money on the table. Like, a lot of money.

We're at this weird inflection point where the gap between "I have an idea" and "I shipped it" is basically nothing now. AI isn't going to change marketing work - it already has. The only question is whether you're going to figure this out now or wait until your competitors force you to catch up.

I'm experimenting now while there's still a learning curve advantage to be had.

You should too.