Lesson 5 cleared out the myths, no llms.txt, no artificial chunking, no AI-only content fork. What’s left is a smaller, more grounded set of practical techniques: writing so a clear answer is genuinely easy to find and extract, using structured data the way it’s actually meant to be used, and making sure the crawlers behind AI systems can actually reach your content in the first place. None of this is exotic, it’s mostly disciplined technical writing and standard technical SEO, applied with an AI reader in mind alongside a human one.
Lesson 5’s framing: this lesson’s techniques sit on top of, not instead of, information gain and E-E-A-T. A page with nothing new to say won’t get cited more often just because it’s well-structured.
Step 1: Write the direct answer first, then support it
An AI Overview or assistant answering a question directly needs to lift a clear, self-contained statement it can quote or paraphrase accurately. Content that builds up to its answer through several paragraphs of preamble makes that harder, not because the preamble is bad writing, but because there’s no single, clean sentence early on that answers the actual question.
Writing a direct answer early doesn’t mean cutting the depth that gives a piece real information gain, it means reordering so the depth comes after a clear, quotable answer rather than before it. A page that’s both immediately answerable and genuinely deep tends to serve both a human skimming and an AI system extracting a specific fact.
Step 2: Use structured data the way it’s actually meant to be used
Schema.org markup (Article, FAQPage, HowTo, and similar vocabularies, added as JSON-LD) gives search and AI systems an explicit, machine-readable description of your content’s structure: what’s a question, what’s an answer, what’s a step in a sequence. This is a long-standing, well-documented technical SEO practice, not something invented for AI systems specifically, and it remains genuinely useful because it removes ambiguity about how your content is organized.
Google’s own structured data policies are explicit that markup must accurately reflect the content actually visible on the page, adding structured data that doesn’t match what a reader sees is a policy violation, not a shortcut. Use it honestly:
As Lesson 5 covered, Google’s own guidance doesn’t describe a separate, AI-specific markup standard you need to adopt. Standard schema.org structured data, used correctly for content that’s honestly structured that way, is the same well-established practice it’s always been, and it helps both traditional rich results and machine parsing generally.
Step 3: Make sure AI crawlers can actually reach your content
An AI system can only cite what it can access. Several AI companies publish the user agent their
crawler identifies as, so you can make a deliberate choice in robots.txt about which ones may
access your public content, rather than blocking or allowing them by accident.
| Crawler | Associated with |
|---|---|
GPTBot | OpenAI, used for training and retrieval-related crawling |
ClaudeBot | Anthropic, used for crawling in support of Claude |
PerplexityBot | Perplexity, used for its search and answer features |
Googlebot | Google, the same crawler already indexing your site for ordinary Search and used as the basis for AI features on Search |
Some site owners deliberately block AI-training crawlers over content-use concerns, and that’s a
legitimate business decision, not something this course takes a side on. What matters is that
it’s a deliberate robots.txt choice, and you understand the trade-off: blocking a crawler
means you cannot then expect to be cited by whatever system depends on it. Check each provider’s
own current documentation for their exact, current user agent name and published crawling
policy before editing robots.txt, these details change over time.
Step 4: Citation-worthiness still comes back to substance
None of Steps 1 through 3 matter if there’s nothing worth citing underneath them. An AI system choosing what to quote or cite is, in effect, doing a compressed version of the same judgment a human editor or a Google quality rater would make: is this specific, is this accurate, is this attributable to a credible source. That’s exactly Lessons 2 through 4 again: real information gain gives an AI system something worth quoting that competing pages don’t offer, and real E-E-A-T, genuine expertise, a real named author, real cited sources, gives it a reason to trust your page enough to cite it accurately rather than paraphrase around it.
Recap
Getting cited by AI Overviews and AI assistants comes down to a small set of grounded techniques: writing a clear, direct, specific answer early in each section before the supporting depth, using schema.org structured data honestly for content that’s genuinely structured that way, and making a deliberate, informed choice about which documented AI crawlers can reach your public content. None of it substitutes for the substance from Lessons 2 through 4, information gain and real E-E-A-T are still what makes a page worth citing in the first place.
Resources & further reading
- Google’s guide to optimizing for generative AI features on Google Search, Google Search Central
- Structured data general guidelines, Google Search Central
- Schema.org vocabulary, Schema.org
- AI features and your website, Google Search Central