7 Cursor skills I actually use for SEO (and where to find them)
I got tired of Cursor giving me generic SEO advice. It would suggest "adding keywords" to pages already blocked by robots.txt. I started collecting SKILL.md files that teach the agent to look at my files and follow a real workflow.
I use these as instruction sets that change how the agent thinks about my repo, not just plugins you install and forget. I keep these seven in my .cursor/skills folder to handle technical audits and AI visibility. For the editor and Composer workflow we've already talked about, see our Cursor review.
| Skill | Source | Best for |
|---|---|---|
| seo-audit | Corey Haines | Technical and on-page audits |
| serp-analysis | Aaron He | Researching what ranks before writing |
| ai-seo | Corey Haines | Optimizing for AI citations |
| programmatic-seo | Corey Haines | Building template-based page sets |
| pseo-quality-guard | lisbeth718 | Blocking thin content at scale |
| seo-auditing | Spencer Pauly | Technical checklists for small sites |
| seo-analysis | nowork-studio | 30-day SEO plans |
1. seo-audit (Corey Haines)
I use this first for new projects or traffic drops. It checks robots.txt, sitemaps, and indexation before suggesting content changes.
It warns about schema detection. AI tools often miss JavaScript-injected JSON-LD, causing fake "missing schema" errors. This skill pushes the agent to use a browser pass or the Rich Results Test instead.
The catch: The file is 500 lines long. It eats up your context window. I tell the agent to "focus on the crawlability section" to keep it on track.
2. serp-analysis (Aaron He)
I open this before writing a post. It stops the agent from guessing what a "good" article looks like. Instead, it looks at what is on page one: AI Overviews, featured snippets, and top result formats.
It helps me choose between a long guide or a quick table. It also hands off to a writer skill once research is done so the outline matches searcher intent.
The catch: You need real evidence. If you don't have a search MCP or paste in top URLs, the agent will hallucinate what is ranking.
3. ai-seo (Corey Haines)
I use this to get cited in AI assistants like Perplexity or ChatGPT. It focuses on "answer-first" structures and clear entity definitions that help LLMs understand your expertise.
I run this on a page after it is indexed. There is no point optimizing for AI citations on a page Google hasn't found.
The catch: AI search changes weekly. This skill can push for a rigid structure that feels robotic if you follow every instruction. I use it for principles, not as a template.
4. programmatic-seo (Corey Haines)
I use this for directories or comparison pages at scale. It keeps the project organized with a "hub-and-spoke" model so you don't end up with orphan pages. It also has a data hierarchy—it reminds you that using public data everyone else has is the weakest way to build pages.
The catch: It is a strategy skill. It helps you plan 500 pages but won't tell you if they are worth the effort. You still make the quality call.
5. pseo-quality-guard (lisbeth718)
I use this "gatekeeper" skill to check for thin or duplicate content before deploying pages. It looks for red flags like 80% similarity or just swapping city names in a template.
It suggests building this check into your code to block deployments if quality drops. This is essential for AI content at scale.
The catch: It is a heavy check. Comparing thousands of pages for similarity is hard for an agent in one go. I run it on a small sample first.
6. seo-auditing (Spencer Pauly)
I use this for small, static sites where the Corey Haines audit feels like overkill. It is a fast checklist for meta tags, Open Graph images, and JSON-LD. It works well for patching a layout.tsx file in Next.js without fuss.
The catch: It is shallow. It doesn't cover international SEO or complex redirects. If your site is bigger than a few dozen pages, you will outgrow it.
7. seo-analysis (nowork-studio)
This is like having an SEO consultant in the IDE. It looks at Search Console data to give you a 30-day plan. It works best with an MCP server connected so the agent can see your actual clicks and impressions.
The catch: It is useless without data. If you haven't wired an MCP or won't paste GSC exports, the agent will give generic advice.
How I use these together
I don't keep all of these active at once; it confuses the agent. I drop seo-audit and serp-analysis into my project to start. I use them to fix the technical foundation and plan content.
Once I have a draft, I run ai-seo to prepare for AI search. If I am building many pages, I bring in the quality-guard to avoid shipping junk.
Skills make Cursor a better editor, but they don't replace checking your own data. Use them to build the right workflow, then monitor your actual rankings.