Citation & Source Influence

How to Write Content That Gets Cited by AI Search (2026 Craft Guide)

Content gets cited by AI search when it is the most relevant, specific, and checkable answer to the question, not because of how it is formatted. A 252,000-trial controlled study found topical relevance and a source's position in the list the model was given drove which source was cited first, explicit prices and recent dates helped, and formatting-only edits had little impact. The 9 patterns in this guide (definition blocks, bulleted lists, comparison tables, sourced statistics, direct-answer paragraphs, visible FAQ blocks, numbered steps, glossary entries, and methodology notes) are ways to package that substance so each passage names its subject and carries its evidence. They serve readers first and passage-quoting engines such as ChatGPT search, Perplexity, and Claude second; for Google's AI features, Google says the lever is still SEO.

Published 2026-05-22 · Updated 2026-09-26

Questions this guide answers

How do I write content that AI engines cite?
What content patterns does ChatGPT cite?
How do I make my blog posts AI-friendly?
Does article length matter for AI citations?
Should I add FAQ schema to my articles?

Direct answer

Why some content gets cited and most doesn't

Two things decide whether a page gets cited: whether the engine retrieves it at all, and whether, once retrieved, it is the best available evidence for the claim the answer is making. Writing affects the second far more than the first. Retrieval depends on the engine's index and on how much it already trusts your domain for the topic, so a well-written page on a domain the engine never pulls for the query won't be cited. Our reading of the What Gets Cited study covers that gap in detail.

Once a page is in the candidate set, the controlled evidence is fairly consistent. In "What Gets Cited" (arXiv 2605.25517, May 2026), researchers ran about 252,000 paired trials across six LLMs, changing one of 18 content factors at a time. Topical relevance and list position were the biggest drivers of being cited first; explicit price information and a recent timestamp helped consistently; completeness and trust cues added smaller gains; and formatting-only edits had little impact. The test anonymized brands and used consumer product reviews, so it measures what wins between two equally trusted sources, not source trust or site size.

The engines also assemble answers differently. ChatGPT search, Perplexity, and Claude's web search fetch pages at answer time and quote passages, so a passage that names its own subject and carries its own evidence survives being lifted out of the page. Google's AI Overviews and AI Mode are built on Google Search's index and core ranking systems, and Google's guide to generative AI search says there's "no requirement to break your content into tiny pieces" and "you don't need to write in a specific way just for generative AI search." For Google, the craft below is simply good writing on top of SEO foundations, not a separate AI format.

So treat the patterns as editorial habits, not hacks. They are what careful reference writers (encyclopedia editors, product reviewers, academic authors) have always done: answer first, name the subject, show the evidence. A July 2026 survey of 45 GEO studies (Martinez, arXiv 2607.14035) found that "no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability," so anyone promising citation lift from formatting alone is overselling.

9 content patterns with before/after examples

Each pattern below shows a "before" (commodity prose, the version most blog posts default to) and an "after" (the same substance in a shape a reader can scan and an engine can quote). The shape helps; the substance (a direct answer to the actual question, specifics, dates, and evidence) is what the controlled research rewards. Use the patterns where they fit the content. Don't force all nine onto every page.

1. Definition blocks: "X is Y"

A one-sentence definition that names the term and its scope can be lifted into an answer without rewriting.

Before: "When people talk about Answer Engine Optimization, they usually mean a kind of evolution of SEO that takes into account the fact that AI tools are now answering questions for users, which means you have to think about how those tools work."

After: "Answer Engine Optimization (AEO) is the practice of optimizing content, structured data, and source-layer presence so that AI answer engines (ChatGPT, Perplexity, Google AI Overviews, Claude, and Gemini) cite your brand when they answer buyer questions."

The "after" version is about 35 words, leads with the term, follows with a precise definition, and names the engines. It can be quoted with zero rewriting. The "before" version cannot.

2. Bulleted lists: scannable, parallel, finite

Lists have natural boundaries, parallel grammar, and a finite count, which helps a reader scan and lets a passage-quoting engine lift the set whole.

Before: "There are several reasons your content might not be getting cited, including issues with your page structure, your authority on the topic, how recent the page is, whether it's easy for crawlers to access, and a few other technical factors that all interact in different ways."

After: Five common reasons content doesn't get cited by AI:

The page blocks the engine's search crawler (OAI-SearchBot for ChatGPT search, PerplexityBot, or Googlebot).
The page doesn't answer the specific question buyers ask.
The key claim is buried mid-page instead of in a direct-answer paragraph.
The page has no specifics or sources: no prices, dates, named methods, or citations.
The page shows no update date, or hasn't been updated in 18+ months.

3. Comparison tables: the whole comparison in one unit

A table compresses a comparison into one unit a reader can scan and an engine can lift whole. Review publishers lead with spec and comparison tables; TechRadar, one of them, was among the three most-cited domains in our 36,268-citation study of the AEO category (5 engines, 30 days ending June 11, 2026).

Before: "ChatGPT tends to prefer different kinds of content compared to Perplexity, and Google AI Overviews has its own preferences too, while Claude often favors editorial sources..."

After: the table in the "Engine-by-engine notes" section below, which puts the same comparison in five rows, one per engine, with the same columns for each.

The table is liftable as a single unit. The prose version makes the reader, and the engine, assemble the comparison themselves.

4. Sourced statistics: a number with its source and date

A number without provenance is a claim. A number with a named, dated source is evidence a reader can check and an engine can attribute.

Before: "AI answers often get facts wrong, and nobody really knows how often."

After: "In a 40-day study of 55,393 trending queries (March 13 to April 21, 2026), Xu, Iqbal, and Montgomery found that 11.0% of 98,020 claims in Google AI Overviews were not supported by the pages cited for them, with omission the most common failure." (arXiv 2605.14021)

The "after" version names the researchers, the sample, the window, and the metric. A reader can check it, an engine can attribute it, and it stays true because it is dated.

5. Direct-answer paragraphs: 40–80 words at the top of sections

A direct-answer paragraph is a self-contained 40–80 word summary placed immediately under the heading (or, for the page-level version, at the very top). For engines that quote passages, such as ChatGPT search and Perplexity, it is often the passage that gets lifted.

Before: A 600-word section that opens with three paragraphs of context before stating the answer.

After: A 60-word paragraph that states the answer in the first sentence, qualifies it in the second, and lists the components in the third, followed by the long-form discussion.

The "after" shape is what every section in this article uses. It is also how Wikipedia opens every article.

A second, less obvious rule decides whether the passage survives. When an engine lifts a passage, it usually drops the heading above it. A direct-answer paragraph that leans on its heading to supply the subject reads as an orphan once it's pulled out of context.

Before (under a heading "Do you need a separate AEO tool?"): "It depends on team size and how many engines you track."

After: "Whether you need a separate AEO tool depends on team size and how many engines you track."

The rewrite costs a few words and makes the passage self-contained: it names its own subject, so it stays correct and quotable after the heading is stripped. Test every direct-answer paragraph that way. Read its first sentence with the heading removed and confirm it still says what it is about.

6. FAQ blocks: real questions, not schema-only

A real FAQ block (visible questions and answers written for people) gives a reader quick answers and gives passage-quoting engines self-contained question-and-answer pairs. FAQPage markup without visible questions on the page breaks Google's structured-data guidelines. Google also stopped showing FAQ rich results on May 7, 2026, although the markup itself remains valid.

Before: A FAQPage JSON-LD block listing five questions, with no visible FAQ on the page.

After: A visible FAQ section with each question as a subheading (or in bold) and each answer as a clean 40–100 word paragraph. Add FAQPage markup only if it mirrors the visible content exactly.

7. Step-by-step lists: numbered procedures

Numbered steps suit procedural questions ("how do I…"). They have an explicit order, parallel grammar, and a finite count.

Before: "First, you'd want to audit your existing pages to see what's working, and then it makes sense to think about what content shapes you're missing, after which you can start adding structure where it's most needed and so on."

After: How to retrofit an existing page for AI citation, in order:

Check that the page answers the specific question buyers ask; topic match comes first.
Add a direct-answer paragraph under the H1 (40–80 words).
Add the specifics a buyer needs: prices, limits, integrations, and dates.
Convert one prose comparison into a table or list.
Cite a named, dated source for every number.
Change the visible date only if the content changed, then rerun the same tracked prompts after 14 days.

8. Glossary entries: short, term-then-definition

A glossary entry is a definition block with an even tighter shape: the term in bold, a one-sentence definition, and an optional one-sentence example. It is the shape of an encyclopedia lede, which is part of why reference pages are so easy to quote.

Before: A paragraph defining "retrieval-augmented generation" inside a longer discussion of AI systems.

After:

Retrieval-Augmented Generation (RAG): a system design in which a language model retrieves relevant documents at query time and grounds its answer in them. ChatGPT search, Perplexity, and Google AI Overviews all retrieve before they answer, each from a different index.

9. Methodology disclosures: "we measured X by Y"

A methodology note turns a claim into something a reader can check and an engine can attribute to you.

Before: "Our research shows that comparison tables get cited more often than other content shapes."

After (illustrative scenario only; the brand and numbers are invented to show the shape): "We ran 40 buyer prompts for a B2B observability brand through ChatGPT, Perplexity, and Google AI Overviews once a day for 30 days (window ending June 30, 2026), 3,600 responses in all, and counted a brand as mentioned only when it was named in the answer text. The brand appeared in 17% of responses; the two category leaders appeared in 38% and 41%."

The "after" version tells a reader what was measured, how, over what window, and under what counting rule: enough to reproduce or dispute it. The "before" version is a marketing claim.

What the research does and doesn't show

The research points one way: invest in the substance an engine can use (a direct answer to the actual question, specifics, dates, and evidence) and let structure do its job of making that substance easy to find. Reformatting thin content, or publishing more of it, isn't supported by any of this evidence. Google's own guide puts it plainly: "a high quantity of pages doesn't make a website higher quality or more relevant to users."

Five findings behind that conclusion, each with its limits:

Relevance beats formatting. "What Gets Cited" (about 252,000 paired trials, six LLMs, 18 factors) found topical relevance and list position were the biggest drivers of being cited first; explicit prices and recent timestamps helped; formatting-only edits had little impact. Limit: two anonymized documents per trial, drawn from consumer product reviews. It doesn't test source trust, domain authority, or how many pages a site has.
No technique has shown durable cross-engine lift. Martinez's survey of 45 GEO studies (arXiv 2607.14035, July 2026) found topical relevance and context position the most reproducible effects, poor transfer of generic heuristics across platforms, and "no reviewed technique" with "a stable, longitudinal, cross-platform causal effect on organic discoverability."
Being cited is not being recommended. In Lily Ray's June 2026 analysis of 100 B2B "best [category]" queries in Google AI Overviews, brands' own self-promotional listicles were cited while the brand was left out of the recommendation 224 of 323 times (69%).
Cited pages don't always support the answer. Xu, Iqbal, and Montgomery found 11.0% of 98,020 AI Overview claims unsupported by the cited pages, mostly through omission. State a claim and its qualifier in the same sentence so a summary can't drop one.
For Google, it's still SEO. Google's guide (published May 2026, updated July 10, 2026) answers "Is SEO still relevant for generative AI search?" with "In short, yes!" and says optimizing for generative AI search "is optimizing for the search experience, and thus still SEO": no llms.txt, no special schema, no chunking requirement, and no need to rewrite content just for AI. Its scaled-content-abuse policy covers separate pages made for every query variation.

What consistently cited reference publishers do

Wikipedia, TechRadar, and arXiv were all among the six most-cited domains in our 36,268-citation study of the AEO category (5 engines, 30 days ending June 11, 2026). Their craft choices are worth copying, not because the shapes are magic, but because they make substance easy to find and verify.

Wikipedia: every article opens with a one-sentence definition (pattern 1) and a short lede (pattern 5), backed by an infobox of facts and inline citations (patterns 3 and 4). Sections are short and headings are explicit.

TechRadar: reviews lead with a verdict (pattern 5), put specs and prices in tables (pattern 3), and list pros and cons (pattern 2). The prices and test dates are exactly the specifics the controlled research rewards.

arXiv papers: an abstract that states the finding first (pattern 5), a methods section (pattern 9), and numbers tied to a named sample (pattern 4).

The common thread: structure that respects the reader and the retriever, carrying real substance. None of it depends on AI-specific files or markup.

Engine-by-engine notes

Each engine retrieves differently, so the same page pays off differently. The short version: passage-quoting engines (ChatGPT search, Perplexity, and Claude) reward self-contained, specific passages; Google's AI surfaces and Gemini reward pages that already do well in Google Search. The substance that helps one helps all five, which is why writing for one engine in isolation is a losing approach.

A summary as of September 2026:

Engine	How it retrieves	What to prioritize	Watch out for
ChatGPT search	Third-party search providers, OAI-SearchBot, and on-demand fetches by ChatGPT-User; quotes passages	Self-contained direct answers, specifics, and accurate docs and official pages; since August 2026 it cites far fewer forum threads	Blocking OAI-SearchBot (GPTBot is the training crawler); answers buried mid-page
Perplexity	Real-time web retrieval; cites the pages it used	Recent, dated, sourced pages	Stale dates; unsourced claims
Google AI Overviews / AI Mode	Google's Search index, core ranking and quality systems, and query fan-out	Helpful pages that already rank, with clear headings written for people	Chunking, llms.txt, or special schema as a Google tactic (Google says none is needed); one page per query variation
Claude (web search)	Web search (provider not named by Anthropic; reported as Brave in 2025); has never cited Reddit	Authoritative editorial sources and in-depth analysis	Thin pages; relying on community threads for visibility
Gemini	Grounded in Google Search	The same foundations as Google's AI surfaces	Same as Google AI Overviews

Anti-patterns (what to stop doing)

The AI-search content market has produced a long list of "tricks" that don't work, don't survive policy review, or have been explicitly refuted by the engines themselves. Seven to retire:

Question-shaped heading quotas. Google's guide asks for clear headings written for people; it doesn't ask for question headings. Use a question heading where the section genuinely answers a question, and don't force the rest.
FAQ markup on pages without visible FAQs. It breaks Google's structured-data guidelines, and FAQ rich results stopped appearing in Google Search on May 7, 2026. Markup must mirror visible content.
llms.txt as a Google lever. Google says its Search doesn't use llms.txt or other special AI files; creating one neither helps nor hurts there. Ship it only if it doesn't delay higher-leverage work. See Why llms.txt Is Not a Strategy.
"AI-only" schema. Google: "there's no special schema.org markup you need to add." Keep structured data that matches visible content, and stop chasing vendor-invented AI schema.
Chunking or rewriting pages for Google's AI features. Google says there's no requirement to break content into tiny pieces and no need to write a special way for generative AI search. Self-contained passages are good writing for every reader; they are not a Google-specific format.
Mass programmatic pages. Google's scaled-content-abuse policy covers creating separate content for every query variation, including fan-out queries, "primarily to manipulate rankings or generative AI responses."
LLM-generated content without human editing. Google's guide warns against content that just recycles what others have said or "could easily be produced by a generative AI model." Use AI as a drafting layer, not the published layer.

Run a free baseline, then apply the patterns

The fastest way to apply this guide is to start with the buyer prompts where ChatGPT already names competitors instead of you. Start Free: the free workspace runs 10 buyer-intent prompts through ChatGPT on your domain (email verification, no credit card) and shows where you're mentioned or skipped, which sources shaped each answer, and one free content audit of a page.

Use it as the input to a rewrite plan: pick the 5 highest-value missing prompts, check each target page for topic match and specifics first, apply the patterns that fit, republish, and rerun the same prompts after 14 days. That's the loop: Measure → Diagnose → Execute → Verify. Treat any before/after change as an observation, not proof of cause; engines change their retrieval without notice.

For the strategic version of this idea (how to choose which pages get the rewrite investment), read How to Earn LLM Citations: the Content-Source Match Framework and How AI Engines Choose Sources. For platform-specific notes, see Optimize for ChatGPT Search and Google Endorses AEO Fundamentals.

Sources

FAQ

How long until I see citations after publishing?

In our experience, often 2–8 weeks for ChatGPT search and Perplexity to begin citing a new page, and it depends on whether the engine already retrieves your domain for the topic. Google AI Overviews draws on Google's Search index and ranking, so pages that already rank can appear quickly, and pages with no SEO foundation will rarely be cited there regardless of shape.

Does article length matter?

Less than writers assume. Google's guide says "there's no ideal page length." What matters is whether the page answers the specific question with specifics a reader can use and an engine can quote. A 600-word page with a strong direct answer, a table, and a visible FAQ is easier to use than a 4,000-word page with the same claim buried in paragraph 18.

What about question-shaped headings?

Use them when the content under the heading is a real answer to a real question. Don't force every heading into question shape: Google's guide asks for clear headings written for people, not question formats, and forced questions make pages harder to skim.

Should I use schema?

Yes, but only markup that mirrors visible content (Article, Product, Organization, BreadcrumbList, and FAQPage when there's a visible FAQ). Google's May 2026 guide says structured data isn't required for its generative AI features and that there's no special AI schema to add, though it still helps with rich-result eligibility. Honest structured data is a comprehension aid; mismatched markup breaks Google's guidelines.

Do I need a real FAQ block?

A visible FAQ block (questions and answers on the page) is one of the most useful shapes for readers and for engines that quote passages. A schema-only FAQ with no visible questions breaks Google's structured-data guidelines and gives readers nothing. Write the FAQ for humans first; add markup only if it matches.

Will AI search reward me for publishing more pages?

Not by itself. An engine cites the page that best answers the question among the few it retrieves, so one strong page beats ten thin ones. Google's guide says "a high quantity of pages doesn't make a website higher quality or more relevant to users," and making separate pages for every query variation can violate its scaled-content-abuse policy. Better pages, not more pages.

Are AI citations guaranteed if I follow these patterns?

No, and any vendor claiming "guaranteed AI citation lift" is overselling. The patterns make relevant, specific content easier to quote; they don't guarantee retrieval or citation. Source trust, third-party corroboration, and engine-specific behavior still matter, and a 45-study survey found no GEO technique with a stable, cross-platform causal effect. Treat this as craft, not a hack.

Free ChatGPT visibility check

See where AI answers skip your brand — then fix it, free

Start a free workspace with your domain: 10 buyer-intent prompts through ChatGPT show where you are mentioned, cited, or skipped, and who gets recommended instead. A free content audit in the same workspace hands you the first fix to ship.

Start Free

Free · No credit card · About 5 minutes

Related guides

Citation & Source Influence

What Gets Cited in AI Answers: Reading the New 252K-Trial GEO Study

A 252,000-trial study across six LLMs ranks what makes AI answer engines cite a source. Then it anonymizes away the brand trust that decides citations in production. How to read it without over-rotating.

Strategy & Positioning

The Pages You Were About to Kill Are Your New Content Template

In the AI-answer era your informational pages weren't penalized, they were absorbed. The pages that survived are your new content brief. One question tells you which.

Citation & Source Influence

How to Earn LLM Citations Without Becoming Spam: The Content-Source Match Framework

AI engines preferentially cite 5 specific content patterns. This guide breaks down the content-source match framework, the 12 examples of cited content, and the slow-burn strategy that compounds without manipulation.

Citation & Source Influence

How AI Answer Engines Choose Sources: The 7 Signals We've Mapped

AI engines like ChatGPT, Perplexity, Google AI Overviews, and Claude choose sources using overlapping but distinct signals. This guide maps the 7 signals that drive citation eligibility and the engine-specific weighting differences.

Strategy & Positioning

Why llms.txt Is Not a Strategy

llms.txt is a proposed standard for AI-friendly content delivery, but it is neither widely adopted by major AI engines nor a substitute for AEO fundamentals. This essay explains what llms.txt does, what it does not, and why brands should focus on the unsexy basics.

AI Engine Optimization

How to Optimize for ChatGPT Search: The 2026 Practitioner Guide

ChatGPT Search uses Bing's index, OpenAI's crawlers, and on-demand fetches. This guide breaks down the five ranking signals, the crawler access checklist, and the content patterns that get cited in ChatGPT answers.

Strategy & Positioning

Google Just Endorsed the Anti-Hack Side of AEO. Here's the CMO Read.

Google's May 2026 generative AI optimization guide explicitly rejects the AEO/GEO hacks that have dominated vendor pitches for the last 18 months — llms.txt, AI-only schema, forced chunking, FAQ-schema citation 'lifts,' page-per-variation production. For a CMO evaluating an AEO program or vendor in 2026, this is the most consequential piece of category-level guidance to land yet. Here's the practical translation.

Citation & Source Influence

The New E-E-A-T: Where Experience, Expertise, Authority & Trust Actually Live in AI Search

E-E-A-T isn't dead in AI search, it moved. The four qualities still decide whether AI recommends you, but they're no longer signaled on your own page, they're established across the sources the model trusts. How each of experience, expertise, authoritativeness, and trust relocates for AEO.