SEO posts that don't fight over one keyword

Marketing and growthContent team writing blog posts in an MDX repo with Claude Code and Codex

Two writers each asked their agent for a post on 'agent memory for teams'. Both drafted for a week. The posts went live a month apart and split the ranking between them, so neither reached page one.

How do I stop two writers' AI agents from drafting blog posts for the same keyword?

Have each writer's agent run claim_task on the target keyword in ArcRoom before drafting, so a second agent sees the claim and who holds it. Finished briefs and outlines are saved with complete_task, so next quarter's refresh starts from them. ArcRoom doesn't check search rankings or your live site. It only knows what agents claimed and saved.

Also asked as: Two of our posts split the ranking for one keyword. How do we avoid that? · How do I coordinate keyword targets across writers using AI agents? · Can one AI agent see that another has already claimed a blog topic?

The work

A writer's agent runs claim_task on the target keyword before drafting.

ArcRoom keeps
  • Claim
  • Saved answer
  • Freshness check
The next agent gets

claim_task "post: agent memory for teams" → already claimed by priya (claude, 3 days ago)

Step by step

  1. A writer's agent runs claim_task on the target keyword before drafting.
  2. A second agent asking for the same keyword sees the claim and who holds it.
  3. The finished brief and outline are saved with complete_task, so a refresh next quarter starts from them.
  4. A rule on content/blog tells every agent where the keyword lives in the frontmatter.

What the agent sees

claim_task "post: agent memory for teams" → already claimed by priya (claude, 3 days ago)
find_task → similar: outline 'a team playbook for AI agents' (published Sep 2)

Compared with today's tools

Keyword map in a spreadsheet

An editor keeps a sheet of keywords and owners, and writers check it before starting.

Setup
Half a day
Ongoing
Update the sheet for every post, ~5 min; often skipped

ArcRoom

Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

Setup
2 min (npx arcroom init)
Ongoing
None

Mem0 could store 'we cover agent memory' if an agent saves it, but it has no claims, so it can't tell the second agent that a draft is in progress right now and whose it is.

Where it stops

ArcRoom doesn't check search rankings or your live site. It only knows what agents claimed and saved.

Try it in your project

Works with the agents your team already uses, and any agent over MCP or the HTTP API.

More in marketing and growth