/blog
How to get cited by AI search
Practical setup for AI Overviews and answer engines: crawl access, llms.txt, extractable structure, and citations you can verify.
- ai-search
- llms.txt
- technical-seo
- citations
AI search does not replace classic SEO — it rewards pages that are easy to fetch, easy to quote, and hard to confuse. You cannot buy a citation slot. You can remove the reasons models skip you.
Let the bots that matter in
Check robots.txt. Do not blanket-block Google-Extended, GPTBot, ClaudeBot, or PerplexityBot unless you have a deliberate paywall strategy. Blocking training crawlers is a product choice; blocking answer crawlers often means you vanish from summaries.
Keep important URLs in the sitemap, return 200 with real HTML (not empty shells waiting on client JS for the thesis), and avoid soft-404 marketing pages that say nothing.
Ship an llms.txt (and keep it honest)
llms.txt is a simple markdown map at your origin that points models at canonical docs, pricing, and definitions. Treat it like a sitemap for language models: short descriptions, absolute URLs, no hype paragraphs.
Update it when URLs move. A stale llms.txt that advertises deleted docs is worse than none.
Write for extraction
Lead with the answer in the first paragraph under a clear H2. Use definition-shaped sentences (“X is…”, “To do Y, …”). Prefer lists and tables for steps and comparisons. Avoid burying the claim three metaphors deep.
One idea per section. Stable headings. Explicit product names and versions. Models quote what they can lift cleanly — mushy brand poetry rarely gets cited.
Earn the mention off-site
Citations still follow authority and consensus. Get named in docs, comparisons, and niche communities that already get crawled. A tracked listing board like SEOheist will not force AI citation — but a public mine page with clear niche copy and a real /go path can become one more crawlable mention.
Monitor brand + product queries in Search Console and referral logs. When a summary cites a competitor, compare structure: do they answer in 40 words under a matching heading?
What will not work
Keyword stuffing “ChatGPT” into titles. Fake FAQ schema with no visible FAQ. Paying for “AI citation packages.” Blocking all bots then wondering why you are invisible.
Be useful, crawlable, and quotable. Then hustle distribution like any other channel.