My 15+ clients’ SEO expert friend uses this way to automate their biweekly keyword audits (and it saves him 675 clicks and 150 minutes of screen work every two weeks)
It is from a client (now a very good friend) who runs an SEO agency. I don’t provide SEO services. I suggested it to him this way, and I’m sharing it with his permission.
How to automate keyword search with Claude?
Connect Semrush or Ahrefs MCP to Claude/ChatGPT and ask:
Find every organic keyword my [domain] ranks for in [country/state/city]. Include the keyword, position, KD, monthly volume and ranking page. Group them by positions 2–15, 16–50 and 51–100. Sort each group by volume.
Now prioritize them
- Positions 2 to 15 are the low-hanging opportunities. These pages already rank well, so just improve content, title, internal links, and backlinks to push them higher.
- Positions 16 to 50 are second priority. These keywords need a little more work, but what he has observed is that if the keywords from positions 2 to 15 are optimized, then these keywords also become optimized. More like a domino effect (because these are more or less related to the keywords shown in positions 2 to 15).
- Positions 51 to 100: These are potential opportunities. Your clients may rank for some of these accidentally by writing these keywords somewhere in the post or page. Better manually review each of them and work on only relevant keywords. For local businesses, ask Claude to filter for keywords containing the city, neighborhood, or service area.
Finally, manually check the intent of each keyword. For example:
- How does a dry freezer work? It’s an informational keyword.
- Buy a dry freezer in NYC. It’s a transactional keyword.
Once you have built a workflow, create a recurring CoWork task to run it every 14 days.
Calculations:
- 675 clicks: He roughly makes 45 clicks per client to manually do this.
- 150 mins: He roughly spends 10 mins per client to manually pull this data.
How to do it Fully Automated with AI?
The issue is that you must first automate for a specific number of days, such as 14 days, but what happens if we want 7 days, 30 days, 45 days, or, let’s say, 90 days? You again have to use the MCP to pull this data.
Rather, you can have something called AI task agents that live inside your plot or ChatGPT, and you can use them to complete this very specific task with just a single prompt. No need to write every single word every time.
Sample prompt: /keyword-audit-task-agent Pull the data for Acme Corp for the last 30 days.
We can build task agents like this for you around your workflow. Each task agent is custom-built for a very specific task. If you are not sure which AI task agents you can use, you can schedule a 45-minute meeting with me, during which you explain your most painful or boring workflow. I will then suggest how you can use existing AI solutions (Claude project, artifacts, Claude code, etc.) to address it, or we will send you a quotation for a custom build.
I’m writing this post because it aligns with my mission. My mission is to free people from the boring parts of their work, so they can do more of what they enjoy.
Agents
412 organic keywords for acmecorp.com. Grouped by position, sorted by volume within each group.
Source — Ahrefs · Acme Corp · 20 Jul – 19 Aug · keyword, position, KD, volume, ranking page
You ask once. It works across your tools. You approve. It repeats.
Illustrative example. Figures are the manual effort this replaces, not a guarantee. Product names are trademarks of their owners.
No comment yet, add your voice below!