The 7-step technical SEO checklist for AI search.

Technical SEO for AI search isn't a different discipline from technical SEO for Google search — but the priorities shift. Some things that used to be "nice to have" became mandatory; some things that were table stakes are now nearly irrelevant. Here's the 7-step checklist we run for every new SEONova customer in their first week.

1. Ship Article + Organization JSON-LD on every URL

LLMs prefer to cite content that's structurally legible. Schema.org isn't a ranking factor in AI search the way it is in traditional SEO, but it materially improves citation rates because the model can extract clean metadata (author, publishDate, topic) without parsing your HTML.

2. Allow AI crawlers in robots.txt (selectively)

GPTBot, ClaudeBot, PerplexityBot, Google-Extended — each one separately. Blocking them via wildcard is the single most common reason brands disappear from AI recommendations. If you want to opt out of training corpora but be included in retrieval-augmented search, the right config is more nuanced than "User-agent: * Disallow: /".

# Allow retrieval-augmented search but opt out of training:
User-agent: GPTBot
Disallow: /

User-agent: ChatGPT-User
Allow: /

User-agent: Google-Extended
Disallow: /

User-agent: Googlebot
Allow: /

User-agent: PerplexityBot
Allow: /

3. Fix your canonicals (yes, still)

LLMs are surprisingly sensitive to canonical confusion. If you serve the same content at /blog/post and /post and /p/post, you'll see your citations fragmented across all three URLs in source lists, which dilutes your aggregate authority.

4. Get your dates right

Models heavily prefer recent content. If your articles don't have a clearly-marked publish date (in schema, in meta tags, and visually), they get downranked in retrieval-augmented citation pools. "Last updated" stamps help too.

5–7: Sitemap, internal linking, performance

Maintain a clean XML sitemap (we wrote a separate piece on this); cross-link your core content into topical clusters so the model can infer your topical authority; keep Core Web Vitals green because slow pages are deprioritized in retrieval the same way they are in Google's main index.