Page patterns AI engines quote: how to write passages that get lifted and credited.

The short answer.

To get your content cited by AI Overviews and ChatGPT, first make sure the page can be crawled, indexed and shown with a snippet. Then write passages an engine can lift whole: open each section with a direct answer, phrase headings as the buyer’s question, define terms in one sentence, put comparisons in real tables, and give every number a date and a named source.

None of this is a secret AI technique. Google says no special optimisation is needed for its AI features, and we agree. These are patterns of clear writing that make a passage easier to quote accurately and to credit to you. They raise your chances; no one can promise a citation. If you are still deciding whether this work deserves budget at all, start with our guide to AEO, GEO and SEO as an investment decision.

What does an AI engine need from a passage?

An AI answer is short and stitched together from several sources. In a July 2025 analysis of the browsing of 900 US adults during March 2025, Pew Research Center found that the typical Google AI summary was 67 words long, and that 88% of summaries cited three or more sources. Your passage is competing for a sentence or two inside that answer.

Two things follow. Question-shaped searches are where AI summaries appear most, so pages that answer questions directly are the ones in contention. And because few people click through, being named and quoted accurately inside the answer matters as much as the click.

For a passage to be used well, it has to pass three tests. Can it be lifted: does it make sense on its own, without the paragraph before it? Can it be trusted: is it specific, dated and sourced? Can it be credited: is it clear who is saying it? Every pattern in this guide serves one of those three.

What Google says, and what it means.

Google’s guidance on AI features is short and direct. Its Search Central page, “AI features and your website”, says: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” It adds that there is “no special schema.org structured data that you need to add.” To be eligible, a page “must be indexed and eligible to be shown in Google Search with a snippet.”

The same page lists ordinary SEO practices: allow crawling in robots.txt and at your CDN or host, make content findable through internal links, provide a good page experience, keep important content in text, support it with good images and video, and make sure structured data matches the visible text. In a Search Central blog post on 21 May 2025, Google added the advice to focus on “unique, non-commodity content”.

How to read that honestly

“No special optimisation” does not mean nothing you write makes a difference. It means there is no hidden switch, file or markup that puts you in AI answers. What helps is what has always helped a page get quoted: being findable, being clear and being the most useful source on the question. The patterns below are that work done carefully at the level of the paragraph. Google also gives you controls the other way: nosnippet, data-nosnippet, max-snippet and noindex limit what can be shown from a page, so check that none of them is set on pages you want quoted.

Other engines publish less, so we keep the patterns general: they suit any system that retrieves and quotes a passage. Our AI search visibility work applies them to the pages that carry a buying decision.

Pattern 1: open with the answer.

The most useful habit is also the simplest. Put the direct answer in the first two or three sentences of the page, and in the first sentence of each section. Context, method and caveats come after. A passage that begins with background forces an engine to dig for the claim; a passage that begins with the claim can be lifted as it stands.

Our newer Insights articles, this one included, open with a chapter called “The short answer”. Our guide to why a website isn’t generating leads starts: “If your website gets visits but few real enquiries, look at the path an enquiry travels before you buy more traffic.” That sentence answers the title on its own. Here is the same idea applied to two typical passages.

To apply this week: open your five most important pages and read only the first paragraph of each. If it does not answer the question the page title asks, rewrite it until it does.

Pattern 2: headings phrased as the buyer’s question.

A heading tells both a reader and an engine what the passage below it answers. “Our approach” tells them nothing. “How long does a website migration take?” tells them exactly when the passage is relevant. Since AI summaries appear far more often on question-shaped searches, headings that mirror those questions make the match obvious.

Not every heading needs a question mark. What matters is that it names the specific topic and that the first sentence beneath it answers it.

Where to find the questions

• Your sales team’s notes: the questions buyers ask on first calls.
• Your enquiry forms: the words buyers use when they describe their problem.
• Search Console: the queries already bringing people to the page, especially those starting with how, what, why or which.

Keep one question per section. For a technical product, our guide on SEO for complex B2B SaaS products shows how to split one dense page into sections that each answer one buyer question.

Pattern 3: definitions and comparison tables in real text.

Two kinds of passage are lifted again and again: definitions and comparisons. Both work well when they are short, plain and written as HTML text that a crawler can read, not as an image, a slide or a PDF.

Definitions

Write the definition as one sentence in the form “X is a Y that does Z”, then one sentence on why it matters. Put it where the term first appears, not in a glossary at the bottom of the page. For key terms, use a short definition box like the one below.

Comparison tables

When a buyer asks “X or Y?”, a table answers faster than paragraphs. Use a real HTML table with a caption, row headers and short cells. Keep one fact per cell, avoid “see above”, and add the date the comparison was checked. The crawler table in chapter 10 of this guide is built that way.

Pattern 4: numbers with a date and a named source.

Specific, attributed facts are what make a passage worth quoting. The most cited research on this is “GEO: Generative Engine Optimization” by Aggarwal and colleagues, first published in November 2023 and presented at the KDD 2024 conference. The authors rewrote web pages in different ways and measured how much of each page appeared in generated answers. Adding citations, quotations and statistics achieved a relative improvement of 30–40% on their main visibility measure. In their test on Perplexity, keyword stuffing performed 10% worse than the unchanged pages.

Read that with care. It was a benchmark built on 2023-era engines, including a simulated engine on GPT-3.5, and the authors note that results may change as engines change. It is good evidence that specific, sourced writing helps; it is not a formula.

A stale number travels too

Dates protect you as well as the reader. On 6 November 2026, when we asked ChatGPT about Prolinkage by name, it described us fairly as “worth shortlisting”, but it quoted an older version of one of our client figures that our own case page had since updated. Engines repeat what is written. If a number changes, change it everywhere it appears, and always write the period it covers.

• Write the number, the unit, the period and the source in the same sentence.
• Name the publisher in the text, not only in a footnote link.
• Prefer primary sources: the study, the platform’s own documentation, your own analytics.
• Keep a list of every number on your site and where it came from, so updates reach every copy.

Pattern 5: say who is speaking, and when.

A passage can only be credited if it is clear who wrote it. Every page that carries an argument or a number should state its author, either a named person or the organisation, and the date it was last updated. Link the author to a page that says who they are and why they can be trusted on the subject.

We publish in the company’s name. Our Insights articles show “By Prolinkage”, linked to our About page, and a real updated date that changes only when the content changes. Changing the date without changing the page is a habit to avoid: it tells readers and engines something that is not true.

Keep your facts the same everywhere

In the same test on 6 November 2026, ChatGPT gave two different founding years for us, because old directory listings disagreed with our current profiles. If your founding year, location, services or leadership differ between your site and your listings, an engine may repeat either version. List your core facts once, then check each profile against that list. Our article on why brands go missing from AI shortlists covers the off-site sources engines rely on.

Pattern 6: FAQ sections and structured data, without the myths.

An FAQ section is a set of passages that are already shaped as question and answer. Done well, each answer is short, direct and complete on its own, which is exactly what an engine can lift. Our newer Insights articles end with a “Questions owners ask” chapter of four to six questions, each answered in a single paragraph of 30 to 70 words.

What structured data does, and doesn’t do

Structured data is code that describes a page in a standard vocabulary. Google says it uses it to understand “the meaning of a page”. That is useful, but three limits are worth knowing.

• It is not needed for AI features. Google says there is no special schema to add for AI Overviews or AI Mode.
• FAQ rich results are now rare. Since August 2023, Google shows FAQ rich results only for “well-known, authoritative government and health websites”. It also said there is “no need to proactively remove” the markup.
• It must match the visible page. Markup that describes content a visitor cannot see is a liability, not an advantage.

Our own pages show the approach. The FAQ markup on our articles is generated from the visible questions chapter, so the two can never disagree, and each article carries Article markup naming Prolinkage as the author organisation. We use it because it describes the page accurately, not because it puts us in AI answers.

Pattern 7: let the right crawlers in.

A passage that is never fetched is never quoted. Each AI company runs different crawlers for different jobs, and blocking the wrong one can remove you from an engine’s search results without you noticing. The names below are from each company’s own documentation, checked on 7 November 2026.

The practical choice for most B2B companies: allow the search crawlers (Googlebot, OAI-SearchBot, PerplexityBot), and decide separately, as a policy question, whether to allow the training ones (GPTBot, Google-Extended). Our own robots.txt, read on 7 November 2026, allows every crawler except into the WordPress admin area.

robots.txt is not the only gate

Google’s own guidance mentions that crawling must be allowed “by any CDN or hosting infrastructure”, not only in robots.txt. Firewalls and bot-protection settings at your host can block crawlers your robots.txt welcomes. Ask your developer to check the server logs for each crawler name above, and to confirm that key content is in the HTML rather than loaded only after a script runs.

What doesn’t work, and what these patterns can’t do.

Some tactics are sold as AI optimisation and have little or no evidence behind them. Others actively hurt.

• Keyword stuffing. In the GEO study, it was the one rewrite that performed worse than doing nothing on Perplexity.
• Markup for content that is not on the page. Google asks for structured data that matches the visible text. Hidden FAQs or invented reviews in markup are a risk, not a shortcut.
• Special files and AI-only pages. Google says no special files or markup are needed for its AI features. Our AI shortlists article covers the evidence on llms.txt.
• Fake freshness. Changing the updated date without changing the content.
• Undated, unsourced statistics. They make a passage harder to trust and easier to repeat wrongly.
• Answers locked in images, PDFs or scripts. If the answer is not in the page’s text, it is harder to quote.

The honest limit

Good page patterns make a passage easy to quote; they do not make an engine choose you. Our newer pages follow the patterns in this guide, and on 6 November 2026, when we asked ChatGPT, Perplexity and Google’s AI Overview which agencies help B2B companies with AI search visibility, none of the three named us. They named eight, ten and six agencies, and only one appeared in all three. Page patterns are one part of the work; being described by independent sources is another. To measure where you stand, follow the method in our guide on checking your AI search visibility.

The passage citability checker.

Paste one paragraph from your website, such as the opening of a service page or an answer from your FAQ. The checker runs ten simple tests based on the patterns above. It is a heuristic for editing, not a prediction of what any engine will do. It runs entirely in your browser; nothing you paste is sent anywhere.

Try it on the “before” and “after” passages in chapter 04 to see the difference. A passage can score well and still be unhelpful, so read the result as a prompt to edit, not as a verdict.

Questions owners ask.

How do I get my content cited by AI Overviews and ChatGPT?

Make sure the page can be crawled, indexed and shown with a snippet, and that OpenAI’s search crawler is not blocked. Then write passages that open with a direct answer, sit under a clear question-style heading, and give every number a date and a named source. No one can promise a citation, but this makes one more likely.

Does Google require special optimisation for AI Overviews?

No. Google’s Search Central guidance says there are no additional requirements or special optimisations for AI Overviews or AI Mode, and no special structured data to add. A page must be indexed and eligible for a snippet. Ordinary SEO practice and clear, useful content are what count.

Does FAQ schema help with AI search?

Not directly. Google says no special schema is needed for its AI features, and since August 2023 it shows FAQ rich results only for well-known government and health websites. A clearly written, visible FAQ section still helps, because each answer is a passage an engine can lift.

Should I block GPTBot?

That is a policy choice about training, not search. GPTBot crawls content that may be used to train OpenAI’s models. Appearing in ChatGPT search depends on OAI-SearchBot, which OpenAI documents separately. Many businesses allow OAI-SearchBot and decide about GPTBot on its own merits.

How long should a passage be to get quoted?

There is no fixed length, but short helps. Pew Research Center found the typical Google AI summary in March 2025 was 67 words and usually drew on three or more sources. A self-contained paragraph of roughly 40 to 150 words, opening with the answer, is a sensible editing target.

Sources: Prolinkage’s own pages and robots.txt, checked 7 November 2026; our AI answer tests in ChatGPT, Perplexity and Google AI Overview, 6 November 2026. Client result: FullTilt Team Development, Search Console Generative AI features impressions, last 3 months vs previous 3, captured 4 Sep 2026. Google Search Central: AI features and your website; Succeeding in AI search (21 May 2025); FAQ rich result changes (8 Aug 2023); Introduction to structured data; Google’s common crawlers. OpenAI, Overview of OpenAI crawlers. Perplexity, Perplexity crawlers. Pew Research Center, 22 July 2025. Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. All checked 7 November 2026.