Ask ChatGPT or Google’s AI Mode a question and you get one tidy answer with a handful of sources attached. Behind that answer, the AI looked at dozens of pages and dropped most of them. Knowing how AI models choose what to cite tells you exactly why some pages make the cut and others don’t.
Here’s the short version. AI tools aren’t simply rewarding the biggest brands or the highest domain authority. They pick the pages that answer each part of a question most clearly, with evidence they can trust. That’s good news if you’re willing to write that way.
At Simba Squad, tracking how ChatGPT, Gemini, Perplexity, Claude and Google’s AI features cite sources is now part of how we plan SEO for our clients. Here’s what’s happening under the hood, and what you can do about it.
How Do AI Models Choose What to Cite?
Most AI search tools follow the same four-stage process. They interpret the question, retrieve candidate pages from a search index, score those pages for relevance, clarity and trust, then use short passages from the best few as evidence in the answer. Those passages become the citations.
The key point is that a citation is proof, not a reward. The AI is looking for text it can safely repeat. Your job is to make your page the easiest, safest source to quote.
The 4 Stages Behind Every AI Citation
Stage 1: The Question Gets Broken Into Smaller Searches
When someone asks “What’s the best CRM for a small real estate team?”, Google’s AI Mode doesn’t run just that one search. It uses a technique Google calls query fan-out, splitting the question into related searches about pricing, features, integrations and reviews, then combining what it finds.
Those sub-searches also shift over time. Surfer’s analysis of 1,600 fan-out runs found only about 27% stayed consistent across repeated runs. That’s why covering a topic thoroughly beats chasing a single keyword.
Stage 2: Candidate Pages Are Retrieved
Next, the AI pulls candidate pages from a search index. Google’s AI Overviews and AI Mode draw on Google’s own index, while tools like ChatGPT and Perplexity run their own crawlers and search systems.
If a crawler can’t reach your page, or your content only appears after heavy JavaScript loads, you’re out of the running before scoring even starts.
Stage 3: Pages and Passages Are Scored
This is where most brands get a surprise. Surfer’s 2026 study of 20,000 prompts and roughly 9 million citations found that link-based authority metrics such as PageRank showed almost no correlation with whether a page got cited.
What matters more is whether a specific passage directly answers a specific sub-question. Selection happens page by page and passage by passage, so two pages on the same website can perform completely differently.
Stage 4: The Answer Is Written and Sources Are Attached
Finally, the model writes its answer and attaches the passages that support each claim. Pages that state facts plainly, with numbers, names and dates, are far easier to attach than pages that hedge or wander.
What Makes an AI Pick One Page Over Another?
Across the research available in 2026 and what we see in client work, these signals consistently separate cited pages from ignored ones:
• Direct answers. The page answers the exact sub-question in the first sentence or two under a clear heading.
• Extractable structure. Short sections with one idea each, plus tables and lists for comparisons, steps and specs.
• Verifiable specifics. Statistics with named sources, dates and expert quotes. The KDD 2024 GEO study found these additions improved visibility in AI answers by up to 40%.
• Freshness. Recently updated pages with current data and a visible update date.
• Corroboration. Your claims and brand name appear on other trusted sites, not just your own.
• Clear authorship. A named author with real, relevant experience, in line with Google’s E-E-A-T principles.
Does Ranking on Google Still Matter for AI Citations?
Yes, but less than most people assume. Google’s AI features draw from its search index, and Google’s own guidance says the usual SEO best practices apply. Being indexed and well optimised keeps you in the candidate pool.
Ranking alone won’t secure the citation, though. Moz research across nearly 40,000 queries found that 88% of links cited in AI Mode didn’t appear in the organic top 10 for that keyword. Strong SEO gets you considered. Clear, quotable content gets you picked.
Do ChatGPT, Gemini and Perplexity Cite Sources Differently?
They do. Conductor tracked citations across ChatGPT, Perplexity, Google AI Overviews, AI Mode, Gemini and Claude between September 2025 and March 2026, and found each engine leans on different types of domains. AI Overviews, for example, consistently favoured YouTube for purchase-related queries.
The takeaway is simple: don’t optimise for one platform. A brand with a well-structured website, useful YouTube content and genuine discussion on forums and review sites gives every engine something to cite.
How to Get Your Pages Picked by AI: A Practical Checklist
Use this list whenever you create or refresh a page you want AI tools to cite:
• Open with a 40 to 60 word answer to the main question
• Turn subheadings into the real questions your buyers ask
• Include at least one sourced statistic or first-hand example per major section
• Use tables for pricing, comparisons and specifications
• Add a named author bio with relevant experience
• Keep key content in the page HTML, not loaded only by JavaScript
• Check robots.txt allows Googlebot, OAI-SearchBot, PerplexityBot and Claude-SearchBot
• Refresh the page every quarter and show the update date
For the off-site side, including brand mentions, entity building and tracking, see our full guide on how to get your brand cited in AI-generated answers.
What Doesn’t Move the Needle as Much as You’d Think
• Domain authority on its own. As the Surfer data shows, a big link profile doesn’t guarantee citations.
• Schema markup on its own. A controlled experiment reported by GrowthX found no major citation uplift after 1,885 pages added JSON-LD schema.
• Keyword stuffing. The same KDD 2024 GEO study found it performed poorly compared with adding facts and sources.
• llms.txt files. No major AI search engine has confirmed it uses them to choose citations.
How Simba Squad Gets Client Pages Picked by AI
At Simba Squad, we build pages with the AI citation process in mind from day one. Our Ahmedabad team handles the research, writing, technical fixes and prompt testing, so every page is built to be found, understood and quoted.
Whether you run a business in India or an agency in Australia, Canada, the UK or the US, our SEO services are designed to win visibility in both traditional search results and AI answers.
Frequently Asked Questions
Why does ChatGPT cite some websites and not others?
ChatGPT cites pages that directly answer part of the user’s question with clear, verifiable information. If your page is hard to crawl, vague, or buries the answer deep in the text, it may be retrieved but not cited.
Do AI models prefer big brands when choosing sources?
Not necessarily. Surfer’s 2026 study found link-based authority metrics had almost no correlation with AI citations. A smaller site with a clear, specific answer can be cited ahead of a large brand with a generic page.
What is query fan-out in AI search?
Query fan-out is when an AI search tool splits one question into several related searches, runs them in parallel and combines the results into a single answer. It’s why covering a topic in depth tends to earn more citations than targeting one keyword.
Does updating old content help it get cited by AI?
Yes. Refreshing statistics, examples and dates shows that a page is maintained, which matters for any time-sensitive topic. Add a visible “last updated” date so both readers and AI tools can see it.
Can I pay to be cited in ChatGPT or Google AI answers?
No. The sources cited in AI answers are chosen organically. Paid ads may appear alongside AI answers on some platforms, but they’re separate from the cited sources.
How do I find out which of my pages AI tools are citing?
Run your key buyer questions through ChatGPT, Gemini, AI Mode, Perplexity and Claude each month and note which of your URLs appear. You can also check GA4 referral traffic from AI platforms, or use an AI visibility tracking tool for larger sites.
Want AI Tools to Start Picking Your Pages?
The pages AI cites aren’t lucky. They’re built to be easy to find, easy to understand and easy to trust. If your content isn’t showing up in AI answers yet, it usually comes down to a handful of fixable gaps.
Get a free AI citation audit from Simba Squad. We’ll test your pages against real buyer prompts and show you exactly what’s holding them back.
