AI Doesn't Rank Your Pages. It Picks Fragments. Here's How to Get Cited.
Written by Derek Chua, digital marketing consultant and founder of Magnified Technologies. Derek has spent the past 18 months building and deploying AI marketing systems in production for SMEs across Singapore.
I had an interesting conversation with a client last week. She runs a mid-sized legal services firm. Good website, solid Google rankings, consistent organic traffic. Then she asked: "My team has been asking why we don't show up when they ask ChatGPT or Perplexity for legal advice. What's going on?"
The honest answer is that she's been optimising for the wrong game.
Key Takeaway: AI search engines don't pick the best page — they extract the best fragment from the best pages. Getting cited in AI answers requires different optimisation than ranking on Google, and the research on what actually works has matured significantly in the past year.
Traditional SEO is about getting your page to rank. Answer Engine Optimisation (AEO) — the practice of getting your content cited inside AI-generated responses from ChatGPT, Perplexity, Google AI Mode, and Microsoft Copilot — is a different discipline entirely.
The traffic stakes are still small but growing fast. AI now accounts for 1.08% of website sessions, growing roughly 1% per month. Microsoft reported that AI referrals to top websites spiked 357% year-over-year in mid-2025. One in four Google searches triggers an AI Overview. In healthcare, it's nearly one in two. These are still minority numbers — but the compounding is real, and the window to build early visibility is now.
How AI Actually Selects Content
Here's the shift that changes everything about how you should think about your content.
Microsoft's principal product manager on the Bing team described it clearly: AI assistants "break content down into smaller, structured pieces that can be evaluated for authority and relevance. Those pieces are then assembled into answers, often drawing from multiple sources to create a single, coherent response."
AI doesn't pick a page and show it. It picks fragments from many pages and weaves them into an answer. Your page might rank No. 1 on Google and still get skipped entirely if the content isn't structured in fragments that AI can cleanly extract.
This is a meaningful shift. Most business websites are written for humans reading from top to bottom. AI reads laterally, looking for extractable snippets that directly answer a question.
What the Research Actually Says
The academic research on AI citation patterns has grown substantially over the past 18 months. A few findings worth knowing.
Earned media dominates. The University of Toronto's large-scale study across ChatGPT, Perplexity, Gemini, and Claude found that AI cited third-party authoritative sources 92.1% of the time in consumer categories — compared to Google's 54.1%. In other words, coverage on industry publications, editorial press mentions, and independent review sites carries far more weight in AI answers than the content on your own domain. If you're only optimising your own website and ignoring your presence on external platforms, you're missing most of the game.
Clarity beats persuasion. The foundational GEO research paper (Princeton, IIT Delhi, Georgia Tech) tested nine content optimisation strategies and found that writing in an "authoritative or persuasive tone" did not improve AI visibility. What worked: citing credible sources (115% visibility increase for lower-ranked sites), adding statistics, and structuring content for direct answers. Marketing language doesn't impress AI systems. Facts and citations do.
Structure is the work. Carnegie Mellon's AutoGEO study found up to 51% improvement over unoptimised content when universal best practices were applied: comprehensive topic coverage, factual accuracy with citations, clear logical structure with headings and lists, and direct answers at the start of each section.
Technical factors matter as much as writing quality. The GEO-16 framework, which analysed over 1,700 real citations from Brave, Google AI Overviews, and Perplexity, found three factors that consistently predict citation likelihood: metadata and freshness, semantic HTML structure, and structured data (schema markup).
What This Means for Your Content
A few practical changes that follow from the research.
Write sections as self-contained fragments. Each H2 or H3 section should make complete sense if read without any other section. AI extracts fragments — if your best answer only makes sense in the context of the full article, it won't get selected.
Lead with the answer. Start each section with the key information, then add context and nuance. This is the opposite of how lawyers, consultants, and academics are trained to write. But it's how AI prefers to read.
Use Q&A structure deliberately. Questions as headings with direct answers below them are native to how AI formats responses. Microsoft's guidance is direct on this point: AI can often lift these pairs word for word into its responses. If you're writing an FAQ section (which every article should have), make sure the answers are self-contained — three to five sentences, readable without the rest of the article.
Don't hide important content. Content collapsed inside expandable menus or hidden in tabs may be invisible to AI crawlers. If information matters, it needs to be in the visible HTML.
Schema markup is not optional. FAQPage, HowTo, Product, and Article schema types give AI systems explicit context instead of forcing them to guess what your content is. The research consistently shows structured data as one of the top predictors of AI citation.
At Magnified, Here's What We've Observed
At Magnified, we've been applying these principles to client content for the past eight months. The pattern holds: service pages that were restructured with descriptive H2 headings, Q&A sections, and schema markup began appearing in AI Overviews and Copilot answers within six to eight weeks of the changes. Pages with the same underlying quality but flat structure (long paragraphs, vague headings, no schema) were consistently skipped.
The more interesting finding has been on earned media. Clients who were quoted in industry-specific publications — even small ones — saw those third-party citations surface in AI answers far more reliably than their own well-optimised pages. Building presence beyond your own domain is not a "nice to have" in an AI search world. It's the primary lever.
One Action for This Week
Check your Bing Webmaster Tools account. If you don't have one, create it — it's free and takes ten minutes to verify. Microsoft now publishes an AI Performance Report inside Webmaster Tools that shows how your content is performing in Copilot citations. It's the only free, first-party tool that gives you direct visibility into how AI is reading your site.
Once you're in, check which of your pages (if any) are being cited in AI answers. Then look at the ones that aren't, and run them against the checklist: self-contained sections, direct answers up front, descriptive headings, schema markup. Fix the structure first. The writing can come later.
The businesses that build this infrastructure now — while most of their competitors are still focused on traditional SEO — will have a significant compound advantage as AI search share grows.
Frequently Asked Questions
What is Answer Engine Optimisation (AEO)? AEO is the practice of structuring your content so that AI search engines — ChatGPT, Perplexity, Google AI Mode, Microsoft Copilot — can extract and cite it in their responses. Unlike traditional SEO, which focuses on ranking whole pages, AEO focuses on making specific fragments of your content easy for AI to identify, extract, and include in generated answers.
Does AEO replace SEO? No. The research shows that 76% of AI Overview citations come from pages already ranking in Google's top 10. Traditional SEO still builds the foundation. AEO is a layer on top — it determines which high-ranking pages actually get cited inside AI responses. The two disciplines reinforce each other.
How do I know if my content is being cited by AI? Start with Bing Webmaster Tools, which has a free AI Performance Report for Copilot citations. For ChatGPT specifically, monitor utm_source=chatgpt.com in your analytics — OpenAI automatically appends this tag to referral URLs. Dedicated tools like Profound and Peec.ai track citations across multiple AI platforms starting at around $95-99 per month.
Should I block AI crawlers from my website? It depends on your goals, and you have more control than most businesses realise. You can allow AI search crawlers (OAI-SearchBot for ChatGPT search, Bingbot for Copilot, PerplexityBot) while blocking training crawlers (GPTBot, Google-Extended) if you prefer not to contribute to model training. These are separate bots with separate robots.txt tokens. Allowing search crawlers is generally recommended if you want to appear in AI answers.
Is optimising for AI search worth the effort for a small business? AI traffic is 1% of sessions today, growing fast. The effort required to restructure content for AI citation — descriptive headings, direct answers, schema markup — largely overlaps with making content clearer for human readers too. There's no trade-off. Businesses that adapt their content now are building compound advantage while the AI search share is still small enough that the competition is thin.