Cited, or left out
The skill starts from one difference with classic search. An AI answer has no ranked list for you to climb. Your page is either cited in the answer or it is not, and a citation in third place still counts.
So the guiding test is plain. If an assistant wanted the single best answer to a question in your field, would your page be the obvious source? Everything the skill changes aims at that test.
Edits on the page itself
- The answer to a question goes within the first two or three sentences of its section.
- Headings use the questions people ask, taken from sources such as AlsoAsked, Reddit threads, or your Search Console data.
- Comparisons, steps, and criteria move from paragraphs into tables and numbered lists.
- Key terms in your field get precise definitions that can stand on their own.
- Each section follows an inverted pyramid: the answer, then the evidence, then the context.
Coverage and reputation off the page
One strong article rarely carries a whole topic. The skill plans clusters: a pillar page with subtopic pages around it, and it values 10 well researched pages over 40 thin ones. It also pushes for mentions in industry publications, podcasts, and roundups, and for original data that other sources want to quote.
It asks you to keep one canonical set of brand facts, such as your founding year, team, location, and product description, and to use them everywhere. Conflicting descriptions make it harder for an assistant to say who you are.
A monthly citation check
Pick the 20 to 30 questions that matter most to your buyers. Each month, the skill has you ask them in ChatGPT with browsing, Perplexity, Claude, Gemini, and Google’s AI Overviews, then record whether your brand or page is cited. It also has you note how each engine describes your company, and it sets up an analytics segment for visits from AI referrers.
The method in the ai-seo file
The page structures, the cluster plan, and the month by month roadmap are in the ai-seo skill file on GitHub. It is released under the MIT License, “Copyright (c) 2026 ScayverGraphix.” The skills follow the Agent Skills spec published by vercel-labs/skills.
Try it on one page first.
Pick a guide that should answer a known question, and ask your agent to optimize it for AI search. Phrases like “get cited by AI” or “answer engine optimization” bring the skill in on their own. Because structured data supports the same work, add it with its schema partner using npx skills add scayver/marketing-skills --skill ai-seo schema.
Markup cannot promise a citation.
The skill recommends FAQ and HowTo markup and more statistics in your content. Treat those as ways to make a page clearer, not as switches. Google’s guidance on AI features says the same search fundamentals apply, with no special markup required and no promised placement. Add questions only where readers ask them, and keep every answer visible on the page.
Pages that answer the visit
Our On-page SEO service studies the search question, the reader’s decision, the visible answer, and the page’s next step before it changes a title or adds a term. That is the same answer first thinking the skill applies, done with your real pages and data.