To get cited in AI Overviews, ChatGPT and Perplexity, make your pages easy to crawl, quick to understand and consistent enough to trust. In practice that means server-rendered pages that AI crawlers may fetch, an opening paragraph that answers the page's question in two or three sentences, question-shaped headings, a short FAQ, structured data and the same facts everywhere you appear. None of it replaces SEO; it builds on it.
This is the setup I use on mohitgoyal.in. I cannot promise any engine will quote a page, and nobody honestly can. What I can do is remove every reason for it not to.
How do AI answer engines choose which pages to cite?
They work in roughly the same order a person researching a question would: find candidate pages, read the relevant parts, and quote the clearest source. Each step has a way to fail.
Google's AI Overviews draw on Google's own index, and Google's documentation on AI features says the same SEO fundamentals apply: a page must be indexed and eligible to show with a snippet. So a page that does not rank or cannot show a snippet is unlikely to be cited. ChatGPT search and Perplexity use their own crawlers, documented by OpenAI and Perplexity, alongside search partners. In every case, the page has to be reachable and the answer has to be easy to lift out.
How do you make your site crawlable for AI?
Start with access. In robots.txt, allow the crawlers you want, such as GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot, and point to your sitemap. If your host, CDN or security plugin blocks bots by default, check that these are not caught by it. The free AI Crawler Checker reads your robots.txt and reports whether 14 named AI crawlers, from OpenAI, Anthropic, Perplexity, Google and others, may fetch your home page, and which rule decides it. It also checks for an llms.txt file and lists the sitemaps your robots.txt declares.
Google-Extended works differently, and it is often misunderstood. According to Google's crawler documentation, it is a control token in robots.txt, not a separate crawler, and it decides whether Google may use your content for training and grounding Gemini and Vertex AI. It does not control whether you appear in AI Overviews. Those follow normal Googlebot rules, so blocking Google-Extended will not remove you from them, and allowing it will not get you in.
Then make sure the content is in the HTML. Some AI crawlers do not run JavaScript: Vercel's analysis of AI crawler traffic found that major ones, including GPTBot and ClaudeBot, fetch JavaScript files but do not execute them. A page that loads its copy in the browser after the first request can look empty to them. Every page on this site is server rendered on TanStack Start, and the Next.js and React websites I build are server rendered too, so the full text arrives on the first request. Check yours with View Source: if the words are not there, fix that first.
How should you write pages so AI can quote them?
Write so that any single passage stands on its own. Answer engines retrieve passages, not whole pages.
| Practice | Why it works | How to apply it |
|---|---|---|
| Answer first | The opening is the passage most often lifted | Open every post and key page with a two or three sentence answer |
| Question-shaped headings | They match how people phrase prompts | H2s such as "When should you choose Next.js?" |
| Direct, complete sentences | A quoted sentence must make sense out of context | Name the subject instead of "it" or "this" at the start of a section |
| Short FAQ | Distinct questions with short answers are easy to match | Four to six FAQs per page, also output as FAQPage schema |
| Tables for comparisons | Structured rows are easy to extract | Comparison tables in decision posts |
| Specific facts | Vague claims give an engine nothing to cite | Real numbers and named projects, never invented ones |
The service pages on this site take the answer-first idea further. Each one opens with a direct reply to the question a client would ask, such as "Yes, I build multivendor marketplaces on WooCommerce with WCFM Marketplace", followed by what you get. That sentence works as a quote without any of the page around it.
What does an answer-first opening look like?
Most pages open with scene-setting that an engine cannot quote. Here is a typical opening for a page titled "Shopify or WooCommerce for a small store?":
In today's fast-moving digital world, choosing the right platform for your online store can feel overwhelming. In this post, we'll explore the options and help you decide.
Nothing in that passage answers the question. An engine looking for a source would skip it and keep reading, or move to another site. Here is the same opening rewritten:
Shopify suits a small store owner who wants hosting, security and updates handled for them. WooCommerce suits one who wants full control of the code and data and is happy to maintain a WordPress site.
The rewrite names both platforms, gives the deciding factor and makes sense if it is the only thing a reader sees. The rest of the page can then explain costs, apps and migration, as I do in my Shopify vs WooCommerce comparison. Apply the same test to every important page: if the first two sentences were quoted alone, would they answer the title?
What role do structured data and llms.txt play?
They hand engines your facts in a form that does not need interpreting. Give each page JSON-LD for what it is, such as BlogPosting, Service or CreativeWork, and link them all to one Person or Organization entity. Pages with an FAQ can also output FAQPage. My guide to JSON-LD structured data shows the full setup.
llms.txt is a plain markdown summary of the site at the root, written for AI tools. It is a proposed convention with uneven support, so I treat it as a small extra rather than a lever. The details are in what is llms.txt and how to add it.
Why does consistency matter so much?
Because an engine that finds two versions of a fact trusts neither. If your home page says one number, your about page another and your LinkedIn profile a third, a careful system will avoid repeating any of them.
On this site, the summaries are written once and reused. Each free tool has one answer-first summary that appears under its heading, in its JSON-LD and in llms.txt. My key facts, such as 100+ sites in 4+ years and 1,000+ installs of my WordPress plugin, are the same in the page copy, the structured data and the llms.txt file.
Audit your facts before your schema
List the five facts you most want an AI to repeat about you: what you do, where, for whom, since when and your best proof. Then check every page and profile says them the same way.
What should you do first?
- 1
Check access
Confirm robots.txt allows the AI crawlers you want, and that your firewall or CDN does not block them.
- 2
Check the HTML
View Source on your key pages. The headline and body copy must be there without JavaScript.
- 3
Rewrite the openings
Give each important page a two or three sentence answer to its main question, at the top.
- 4
Add question headings and an FAQ
Match the questions customers ask, with short, distinct answers.
- 5
Add structured data and llms.txt
JSON-LD per template, linked to one entity, plus a factual llms.txt at the root.
- 6
Spot-check the engines
Ask AI Overviews, ChatGPT and Perplexity your customers' questions every month and note who they cite.
Good titles and descriptions still decide whether people click when you do appear. The free Meta Tag Generator writes options sized for Google, and the Schema Markup Generator writes JSON-LD from a plain description. To plan a new page, the free SEO Content Brief Generator outlines the headings and questions to answer for a keyword, and the FAQ Generator drafts an FAQ with FAQPage schema from your content.
Is AEO different from SEO?
Mostly it is SEO with more attention to how passages read on their own. The technical base is the same: indexable pages, fast loading, clean metadata and structured data, as in my technical SEO checklist for Next.js. The writing changes: answers move to the top, headings become questions and facts become specific.
I build sites with all of this in place from launch, and audit existing ones through my technical SEO and Core Web Vitals service. If you want to know why AI engines are passing your site over, book a 20-minute call and I'll go through your robots rules, rendering, page openings and structured data with you.
