SEO for AI AgentsSEO for blogs and content sites
When the product is the content, the value concentrates in depth, structure, and authorship. The pack hunts the failures that quietly cap a content site: orphan posts, posts cannibalising each other, thin archives, and missing author signals, and formats real answers so AI engines can cite them.
The Visibility Ladder, tuned
Where sites like this actually fail
Same five rungs, same order, different emphasis. Diagnose top to bottom, fix bottom to top.
Usually fine on a content CMS, but client-rendered article bodies and infinite-scroll archives can hide posts from crawlers.
The heaviest rung. Depth, originality, and genuine expertise matter most; thin templated posts are the main failure. Strong heading outlines, real bylines, accurate dates.
Article or BlogPosting with a real Person author and truthful dates, plus breadcrumbs.
Topic clusters are the signature move: pillar pages linking to detailed posts and back. The pack hunts orphan posts and consolidates cannibalising ones.
Topical authority across the cluster, and honest freshness: genuinely updating cornerstone posts rather than bumping dates.
Content sites are prime AI-citation candidates. Question-shaped headings, self-contained answers, definitions, and comparison tables pay off most here.
Priority tilt: Weight Read and the AEO layer highest, with Connect (clusters) close behind. Quick wins: fix orphans, add real author schema, surface buried answers.
Checks specific to this kind of site
Cannibalisation
Two posts competing for one query split the signals. The pack finds them and recommends merging or re-targeting.
Author entities
Real, consistent author pages with Person schema and links to genuine profiles. Strong for ranking and for AI attribution. Never fabricated.
Archive hygiene
Thin auto-generated tag archives bloat the index. The pack noindexes low-value ones and keeps useful hubs.
Honest freshness
Cornerstone posts get genuinely updated and re-dated. Date-bumping without changes is flagged, not performed.
Skills that do the work
- seo-orchestratorApplies the content-blog profile and runs the audit lifecycle
- seo-content-auditScores every URL: keep, improve, consolidate, or prune
- seo-content-editingImproves real copy while preserving your voice; never generates filler
- cite-aeo-geoAnswer-block formatting where it pays off most
Driven by the content-blog profile in the pack.
Structured data it works with
- Article / BlogPosting
- Person (real authors)
- BreadcrumbList
- WebSite
Always honest: markup describes only what's real and visible on the page. Fabricated ratings, authors, or credentials are removed and flagged, never added.
Try it on your site
Paste this into Claude Code, Cursor, or any AI coding agent. It's read-only: the agent audits and reports, and changes nothing until you ask.
Use the seo-for-ai-agents skill pack (https://github.com/manibharij/seo-for-ai-agents) with the content-blog profile: audit my blog on the served HTML. Map topic clusters, find orphan posts and articles cannibalising the same intent, review author schema, and flag thin archives. Report only, don't change any code or files yet.
Common questions
What is content cannibalisation and do I have it?
Cannibalisation is when two or more of your own posts target the same search intent, so engines split the signals between them and neither ranks as well as one strong page would. It's one of the most common issues on mature blogs. The pack detects competing posts and recommends merging or re-targeting them.
Will it write blog posts for me?
No. The content skills assess and edit real content: they score what to keep, improve, consolidate, or prune, and tighten existing copy while preserving your voice. Where new content is needed, the output is a brief for a human, because generated filler is exactly the thin content the audit exists to catch.
Do author pages really matter for SEO?
Yes, increasingly. Real author pages with Person schema and links to genuine profiles support E-E-A-T, and they matter for AI attribution too, because answer engines prefer citing content with a clear, verifiable author. The pack marks up real authors and refuses to fabricate them.