Everyone wants the agentic SEO results. Nobody asks about the daily routine that produces them.
I hit 200 clicks/day on a blog that had 5 impressions/week three months ago. Not from one viral article—from a repeatable Claude Code SEO workflow I run every morning in 10 minutes. Here's the exact routine, step by step.
The Stack (Quick Recap)
If you've read my agentic SEO stack breakdown, you know the components. Quick recap:
- Claude Code — the AI agent running in your terminal
- Supabase — database-backed CMS with REST API access
- Google Search Console — performance data per page and query
- Knowledge files — CLAUDE.md, database schema, tone rules, internal links
The stack doesn't change. What matters is how you use it daily. That's what this article covers.
The Daily Routine: 10 Minutes
Minute 1-3: GSC Health Check
I open Claude Code and ask it to pull my Google Search Console data. Not all of it—I focus on three things:
New opportunities: Queries where I appeared for the first time in the last 24-48 hours. These are Google testing my content for new keywords. If a new query shows 50+ impressions, it's worth paying attention to.
Bleeding pages: Pages where impressions are up but clicks are down compared to last week. That's a CTR problem—usually a meta title or description that doesn't match search intent.
Position movers: Pages that moved from position 8-15 to position 4-7. These are striking distance pages. A small optimization can push them to page 1.
The agent reads the data and flags what matters. I don't export spreadsheets. I don't open tabs. One prompt, one report.
Minute 3-5: Fix What's Broken
If the agent flagged a bleeding page, I fix it right there. The agent already has the context—it knows the page content, the search queries, and the performance data. I tell it to:
- Rewrite the meta title to match the dominant search query
- Adjust the meta description to match search intent
- Push the changes directly to Supabase
This takes 60 seconds per page. The agent reads the current title, checks what queries trigger the page, and rewrites to match. Then it pushes via the Supabase API. No CMS login, no clicking through admin panels.
Minute 5-8: Content Gap Check
I ask the agent to cross-reference my published articles against the queries GSC shows. Specifically: are there queries with 500+ impressions where I don't have a dedicated page?
If yes, the agent drafts a content brief: target keyword, search intent, suggested structure, internal links to existing content. I review the brief and either approve it for writing or park it for later.
This is where the knowledge files pay off. The agent knows my database schema, my existing articles, my tone—so the content brief is ready to execute, not a generic suggestion.
Minute 8-10: Write or Optimize
Two options depending on what the data says:
Option A: Quick optimization. The agent identified a page that needs a better FAQ section, an updated intro, or improved internal linking. It reads the current content from Supabase, makes the changes, pushes the update. Both EN and DE versions. Done.
Option B: New content. If there's a content gap worth filling today, the agent starts drafting. It follows my exact structure—hook, substance, verdict, FAQ—because the knowledge files define the format. I review the draft, tweak the take if needed, and the agent publishes as draft.
Not every day has a new article. Some days are pure optimization. The data decides, not a content calendar.
The Weekly Deep Dive
Once a week I spend 30 minutes instead of 10. The extra time goes to:
Internal linking audit: The agent checks every published article and identifies pages that should link to each other but don't. It adds the links. This is the most underrated SEO tactic—internal links compound faster than backlinks.
Content refresh: Articles older than 4 weeks get a freshness check. Are the facts still current? Has search intent shifted? The agent reads the article, checks GSC data for that page, and flags anything that needs updating.
Strategic planning: I ask the agent to analyze my topical clusters. Where am I strong? Where are gaps? This feeds the content pipeline for the next week.
Weekly insights on AI Architecture. No spam.
For the monthly comprehensive audit, I use Agent Teams—multiple AI specialists running in parallel. But for weekly work, a single Claude Code session is plenty.
The Prompts That Work
I don't use elaborate prompts. The knowledge files do the heavy lifting. Here are the three prompts I use most:
Morning check:
Check my GSC data for the last 48 hours. Show me: new queries with 50+
impressions, pages where CTR dropped vs last week, and pages that moved
into position 4-7. For each, recommend the specific action.
Quick fix:
Page [slug] has [X] impressions for "[query]" but only [Y] clicks.
Read the current content and meta. Rewrite the meta title and description
to match the search intent for that query. Push to Supabase.
Content gap:
Cross-reference my published articles in Supabase against GSC queries
with 500+ impressions. Which queries don't have a dedicated page?
For each gap, give me: target keyword, search intent, suggested
structure, and 2-3 internal links to existing content.
That's it. Three prompts cover 90% of daily SEO work. The agent has the context to execute each one because the knowledge files define the database schema, the writing rules, and the project state.
Why This Works: The Compounding Math
Daily optimization creates a compounding effect that weekly or monthly cycles can't match.
Week 1: You fix 5 meta titles. Two pages move from position 8 to position 4. Clicks on those pages triple.
Week 2: The increased clicks improve your domain signals. Other pages start ranking better. You fix 5 more titles and create 2 new pages for content gaps.
Week 3: The new pages start getting indexed. Your internal linking pushes authority to them. GSC shows new query clusters you hadn't targeted.
Week 4: The compounding kicks in. Pages you fixed in week 1 are now climbing into position 1-3. New pages from week 2 are entering the top 10. You have more data, more opportunities, more leverage.
Manual SEO can't keep up with this cycle because each iteration requires hours of spreadsheet analysis. With Claude Code, each iteration takes 10 minutes. That's the real advantage—not that the AI writes better content, but that the optimization cycle runs daily instead of monthly.
What I Don't Do
Just as important as the workflow is what I skip:
I don't use keyword research tools. GSC tells me what Google already shows me for. That's more valuable than any keyword tool's volume estimate. I optimize for real queries, not projected ones.
I don't follow a content calendar. The data decides what gets written. If GSC shows a content gap today, I fill it today. If there's no gap, I optimize existing content. Calendars are for marketers. Data is for engineers.
I don't batch-optimize. Some people save up SEO tasks and do them all on Friday. That's wasted compounding time. Every day you delay a meta title fix is a day of lost clicks.
I don't manually check rankings. The agent reads GSC. I never open GSC directly. The agent filters signal from noise and only shows me what needs action.
The Numbers
The workflow produces consistent, predictable growth:
| Week | Daily Clicks | Daily Impressions | Action Count |
|---|---|---|---|
| 1 | 30 | 2,000 | 15 fixes, 2 new pages |
| 2 | 80 | 5,000 | 12 fixes, 3 new pages |
| 3 | 140 | 9,000 | 10 fixes, 2 new pages |
| 4 | 200 | 14,000 | 8 fixes, 3 new pages |
Notice the pattern: fixes decrease as there's less to fix. New pages fill gaps the data reveals. Clicks compound because every optimization builds on the previous ones.
The Verdict
Claude Code SEO is not about having an AI write articles for you. It's about running a data-driven optimization loop every single day for 10 minutes.
The stack is simple: Claude Code + Supabase + GSC + knowledge files. The workflow is simpler: check data, fix what's broken, fill content gaps, repeat. The results compound because daily iteration beats weekly cycles every time.
Start with the morning GSC check. That single habit—10 minutes, every day—is worth more than any SEO course or tool subscription. The data tells you exactly what to do. The agent executes it in seconds. You review and approve.
That's the entire workflow. No mystery. Just engineering applied to content.