Getting Cited by ChatGPT and Perplexity: The Part Everyone Skips

Most advice on getting cited by AI answer engines is about writing better content: answer-shaped paragraphs, concrete claims, FAQ schema. All true. None of it matters if the crawler never sees the page — and that's the part almost nobody checks.

The content side, briefly

Three things actually move the needle on extractability:

• Answer-shaped content. Lead with a 40-80 word direct answer a model can lift verbatim, not three paragraphs of throat-clearing before the point.
• Concrete claims. Numbers, named entities, dates. Vague prose doesn't get quoted because there's nothing precise to quote.
• Machine-readable structure. Clean heading hierarchy, FAQ schema, fast static pages, an llms.txt at the domain root.

This is well covered elsewhere and it's real. But we shipped 85 pages built exactly this way and got zero search impressions — not zero citations, zero impressions — until the sitemap was actually submitted to Search Console. The content was never the bottleneck. Discovery was.

The check nobody runs: which engine is actually finding you

Here's the one that surprised us. We assumed Google was the crawler that mattered, because it's the crawler every SEO guide is written for. Our own server logs said otherwise: 91% of this site's real human traffic arrives via Bing and DuckDuckGo. Google sends 2% — 61 impressions and zero clicks in the most recent week we measured.

That number changes what "optimizing for AI citation" should mean in practice, because Bing doesn't just power Bing search — it's the index behind Microsoft Copilot. If your citation strategy is Google-shaped by default, you're optimizing for the crawler sending you the least traffic and possibly feeding none of the AI surfaces that matter to you.

The practical fix costs nothing: IndexNow is a one-line ping that notifies Bing (and everyone else on the protocol) the moment you publish or update a page, instead of waiting for a crawl cycle to notice. Google doesn't participate in IndexNow, so if you only measure "did Google see it," you'll never notice this lever exists.

A three-item checklist that catches what content-quality advice misses

1. Confirm your sitemap is actually submitted, not just present at /sitemap.xml. A file existing and a crawler knowing about it are different states, and only one of them gets you indexed.
2. Check your referrer logs for which engine sends real traffic, not which one you assumed would. If Bing or DuckDuckGo outweighs Google, ping IndexNow on every deploy — it's a fetch call in your CI, not a product.
3. Verify your robots.txt explicitly allows the AI crawlers — GPTBot, ChatGPT-User, anthropic-ai, PerplexityBot, Google-Extended. A blocked crawler cannot cite you no matter how well-structured the page is underneath.

None of this replaces writing good answer-shaped content — it's what makes the good content reachable in the first place. If you're already doing the content work and generating your own llms.txt and schema by hand gets old fast, SEO Autopilot automates the structural side; the discovery-side checks above are free and worth doing regardless of what generates your markup.

Full answer: https://agentkitworks.com/answers/how-to-get-cited-by-chatgpt