AI Citations: Why some websites get cited and others don't?
Updated: October 2026 | ORBA Insights
Why does AI reference one website but overlook another, even when both offer similar information?
As more people use ChatGPT, Google AI Search, Perplexity and other AI-powered tools to research products, services and businesses, a new form of online visibility is becoming increasingly important: AI citations.
For years, businesses have focused on appearing in Google’s search results. Now, they also need to consider whether their content is being discovered, understood and referenced by AI systems.
But how do AI citations actually work? Do AI models favour established brands? Does ranking first on Google improve your chances? And can businesses influence which websites AI chooses to cite?
The answer is more complicated than a single ranking factor.
Research published in 2026 is beginning to reveal how AI search platforms select sources, what makes website content useful to AI systems and why being cited is not always the same as influencing an answer.
Here’s what businesses need to understand.
What Are AI Citations and Why Do They Matter?
AI citations are references or links to sources that an AI-powered system provides alongside a generated answer. They help users identify where information came from and explore the original content.
For example, someone might ask an AI assistant:
“What should a small UK hotel consider when choosing dynamic pricing software?”
Rather than displaying only a list of websites, an AI-powered search system might produce an explanation covering pricing automation, integrations, revenue management and operational costs.
The answer may include links to relevant software providers, hospitality research, industry publications or product documentation.
Those references are AI citations.
For businesses, AI citations create another potential route to discovery. A website can be introduced to a prospective customer during the research process, sometimes before that person decides which businesses to visit.
However, an AI citation is not automatically a recommendation or endorsement. A source may simply provide supporting information for one part of an answer.
This distinction matters when measuring AI visibility.
How Do AI Models Select Websites for AI Citations?
AI models do not all select sources in the same way. Their search capabilities, retrieval systems, available indexes and methods of generating answers can differ.
However, AI-powered search often involves several broad stages.
1. Understanding the user’s question
The system first interprets what the user wants to know.
A question such as “Which cybersecurity services should a financial company consider?” contains several important concepts: cybersecurity, financial services, business needs and potentially regulatory requirements.
The AI system may need information about each of these concepts to construct a useful response.
2. Searching for relevant information
When an AI product uses web search or retrieval, it can look for relevant sources rather than relying exclusively on information learned during training.
Google explains that its AI search features can use a technique called query fan-out, which involves issuing multiple related searches across different subtopics.
This means a single user question can lead to several searches behind the scenes.
3. Retrieving and evaluating potential sources
The system identifies material that may help answer the question.
Depending on the platform, this could include website pages, research articles, product documentation, industry information or other accessible sources.
The relevance of a page to the question matters, but being relevant does not guarantee selection.
4. Generating an answer
The AI system uses available information to construct its response.
Some retrieved material may influence the final explanation. Other material may be disregarded or used only indirectly.
5. Displaying AI citations
The system may then present supporting source links or references.
Importantly, not every retrieved source will necessarily appear as a citation, and a displayed citation does not prove that every statement in the answer came from that page.
This is why businesses should distinguish between retrieval, citation and actual contribution to an answer.
Five Factors That May Influence AI Citations
There is no publicly verified universal formula for earning AI citations across every AI platform.
Nevertheless, official guidance and emerging research suggest several areas businesses should pay attention to.
1. Topical Relevance Is Fundamental to AI Citations
A website needs to provide information relevant to the user’s question.
Imagine someone asks:
“What are the benefits of outsourced marketing for small manufacturing companies?”
A general marketing agency homepage may be relevant.
But a detailed page explaining outsourced marketing specifically for manufacturers could provide a closer match to the question.
The second page might explain the industry, target audience, services, practical challenges and outcomes in greater detail.
The lesson: Create content that answers real questions about your business, services and specialist areas.
This does not mean publishing hundreds of near-identical pages for different search phrases. It means providing useful, specific information.
2. Clear Business Information Supports AI Understanding
AI systems need enough context to interpret what a business offers.
Consider these two website descriptions.
Example A:
“We deliver innovative solutions that transform business potential.”
Example B:
“We provide cybersecurity consulting, penetration testing and compliance support for UK financial-services businesses.”
The first sounds polished but reveals little about the company.
The second communicates the services, audience, industry and geographic market.
Clear descriptions help make a business easier to interpret when its website is retrieved.
This is closely connected to AI website understanding: how effectively an AI system can identify a business and interpret the information presented on its website.
3. Evidence and Specific Information Can Make Content More Useful
Useful content often contains information that can support a specific answer.
Depending on the topic, this may include:
* Clear definitions and explanations.
* Original research or documented findings.
* Product specifications and service details.
* Relevant examples and comparisons.
* Transparent pricing information, where appropriate.
* Credible references supporting factual claims.
For example, a page explaining how a software product works, which integrations it supports and who it is designed for may be more useful for a product-comparison question than a page containing only promotional statements.
The objective is not to insert statistics or citations artificially. It is to make claims verifiable and information useful.
4. Website Accessibility and Technical SEO Still Matter
Even excellent content may have limited visibility if AI search systems cannot access it.
Google states that established SEO practices remain relevant to its AI search features.
Pages generally need to be accessible, indexable and eligible for relevant search features.
Similarly, OpenAI advises publishers that allowing its search crawler, OAI-SearchBot, helps make public website content eligible for discovery through ChatGPT Search.
Sources:
Technical accessibility is a foundation, not a guarantee of AI citations.
5. Content Structure Helps Explain Information
Well-organised content can make it easier for readers and automated systems to locate relevant information.
Useful practices include descriptive headings, short explanations, clearly labelled services, consistent terminology and relevant internal links.
For example, a page about cybersecurity consulting should clearly distinguish penetration testing, risk assessments and compliance support rather than presenting everything under an undefined label such as “Our Solutions.”
Structured data can also help search engines interpret certain information when it accurately reflects the visible page.
However, adding schema markup does not automatically produce AI citations.
What Does Research Reveal About AI Citations in 2026?
One of the more interesting developments in generative engine optimisation is the growing distinction between being cited and being used.
A 2026 research paper titled From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms, by Kai Zhang, Xinyue He and Jingang Yao, examined citation behaviour across ChatGPT, Google AI Overview/Gemini and Perplexity.
The researchers analysed:
* 602 controlled prompts.
* 21,143 valid search-layer citations.
* 18,151 successfully fetched pages.
* 72 extracted features.
The study distinguished two stages.
Citation selection: Whether an AI search platform chooses a website as a source.
Citation absorption: Whether information from that website meaningfully contributes to the final AI-generated answer.
The researchers observed differences between platforms. In their dataset, Perplexity and Google cited more sources on average, while ChatGPT showed greater average citation influence among the pages the researchers successfully fetched.
Pages associated with stronger answer-level influence tended to have richer structure, closer semantic alignment and more extractable evidence, such as definitions, comparisons and numerical information.
These findings are observational and should not be treated as a guaranteed optimisation formula.
Nevertheless, the study highlights an important point for businesses:
A higher number of AI citations does not necessarily mean a website has more influence on what an AI system says.
Read the research: From Citation Selection to Citation Absorption — arXiv, 2026.
Why Google Rankings Don’t Guarantee AI Citations
A common misconception is that the highest-ranking website on Google will automatically become the first source cited by an AI model.
That is not how every AI-powered search experience operates.
Traditional search rankings and AI citations are related, but they represent different outcomes.
Google’s AI Overviews and AI Mode are connected to its existing Search systems. However, Google explains that these features can perform additional searches and identify supporting pages across different subtopics.
The links displayed can therefore differ from those shown in a conventional search results page.
Other AI platforms may use different retrieval systems and source-selection methods.
This creates several possibilities:
A business might rank well for a keyword but not appear as a citation in a particular AI answer.
A specialist article might be cited for a narrow question even if it is not the highest-ranking result for a broader search term.
The same website might receive AI citations for one question but not for another.
And two AI platforms might cite entirely different sources when answering a similar question.
Ranking well remains valuable. But businesses should not assume that traditional search performance fully represents their AI visibility.
Do ChatGPT, Google AI Search and Perplexity Choose the Same AI Citations?
Not necessarily.
Although these platforms can all provide answers supported by web sources, they do not use identical systems.
ChatGPT Search can search the web and display relevant source links. OpenAI says search results are ranked using multiple factors intended to help users find relevant, reliable information, and inclusion is not guaranteed.
Google AI Overviews and AI Mode use Google’s Search infrastructure and may retrieve additional information through query fan-out. The supporting links can vary between AI features and individual searches.
Perplexity also provides source-supported answers, but its retrieval and citation behaviour can differ from other platforms.
The 2026 citation-selection research reinforces this distinction: citation frequency and the influence of cited content varied across the systems examined.
For businesses, this means AI visibility should not be assessed using a single prompt on a single platform.
A broader set of relevant questions, tested consistently over time, provides a more useful directional picture.
How Can Businesses Improve Their Chances of Receiving AI Citations?
No business can guarantee that an AI model will cite its website.
However, businesses can improve the quality, clarity and accessibility of the information they publish.
Start with the questions customers actually ask
Think beyond short keywords.
Instead of targeting only “accounting services”, consider questions such as:
“What accounting support does a new limited company need in the UK?”
A useful article could explain bookkeeping, annual accounts, tax obligations and how professional services fit together.
Make every important service understandable
Each major service page should explain what the service is, who it helps, which problems it addresses and what customers should expect.
Avoid relying on slogans to communicate essential information.
Publish information worth referencing
Original insights, detailed explanations, practical examples and well-supported factual information give readers more reasons to trust and use a page.
Where relevant, identify the author, explain their expertise and show when information was updated.
Connect related information
Use descriptive internal links between service pages, educational articles, FAQs and supporting resources.
For example, an article about AI citations can link to a guide about Google AI Search or an explanation of AI website understanding.
This helps readers navigate the subject and makes related content easier to discover.
Keep business information consistent
Business names, service descriptions, locations and other important details should not contradict each other across the website.
Consistency helps reduce ambiguity.
Maintain technical accessibility
Review indexing, crawl restrictions, page availability and other foundational SEO requirements.
A technically inaccessible page cannot reliably participate in retrieval-based AI search.
Measure outcomes rather than guessing
Track which pages are receiving AI citations where measurement is available, alongside AI referral traffic and changes in the questions for which the business appears.
Treat visibility measurements as signals, not proof that a particular change caused an AI model to cite the website.
Can Businesses Measure AI Citations?
AI citation measurement is becoming more practical, although visibility remains incomplete across platforms.
In February 2026, Microsoft introduced AI Performance in Bing Webmaster Tools.
The reporting includes metrics such as total citations, average cited pages, page-level citation activity and grounding queries across supported Microsoft AI experiences.
In June 2026, Microsoft expanded its AI visibility reporting with additional insights, including Citation Share, which shows a website’s share of displayed citations for particular grounding queries.
Microsoft emphasises that these metrics are observational. They do not represent a universal AI ranking or authority score.
Businesses can also examine relevant AI referral traffic in their analytics platforms and manually review a consistent set of questions.
Sources:
These methods have limitations, but they provide a starting point for understanding AI discovery beyond tr+aditional rankings.
How ORBA Helps Businesses Understand Their AI Visibility
Before trying to improve AI citations, businesses need to answer a more fundamental question:
Does AI clearly understand what our business does?
A website might be accessible and rank for relevant keywords while still communicating its services, audience or expertise unclearly.
ORBA is designed to help businesses identify those gaps.
By analysing a business website, ORBA provides an AI visibility assessment, highlights what the website communicates about the business and identifies areas where information could be clearer.
Its recommendations focus on practical improvements to the information businesses can control.
ORBA cannot guarantee AI citations, recommendations or rankings. Those outcomes depend on the individual AI platforms and the questions being asked.
What ORBA can help businesses do is understand how effectively their website explains who they are, what they offer and where improvements may be needed.
AI Citations Are About More Than Being Found
AI citations are introducing another dimension to online visibility.
Businesses still need strong technical SEO, useful content and accessible websites. But as AI-powered search becomes another route to discovery, they also need to think about how clearly their information can be interpreted and used.
Research suggests that the number of citations a website receives is only part of the picture. Whether its information contributes meaningfully to an answer is another important consideration.
For businesses, the practical starting point is straightforward: explain what you do, answer relevant questions, support important claims and make information easy to find.
Because the question is no longer only:
“Can customers find our website?”
It’s also:
“When AI looks for an answer, is our business clear and relevant enough to be part of it?”
That is the challenge ORBA is helping businesses understand.
Frequently Asked Questions
What are AI citations?
AI citations are source references or links displayed alongside AI-generated answers. They allow users to explore information used or identified as supporting material by an AI-powered system.
How do AI models decide which websites to cite?
AI models with search capabilities may retrieve relevant web content, evaluate potential sources and select references that support an answer. The process varies between platforms, and no single publicly verified formula determines all AI citations.
Does ranking first on Google guarantee AI citations?
No. Traditional search rankings and AI citations are different outcomes. Ranking well can support discoverability, but AI systems may retrieve and reference other sources depending on the question and platform.
Can a small business receive AI citations?
Yes. A smaller business can potentially be cited when its accessible content is relevant and useful to a particular question. Business size alone does not determine whether a website can appear as a source.
Does structured data guarantee AI citations?
No. Accurate structured data can help search engines understand certain information, but it does not guarantee that an AI system will retrieve, use or cite a page.
Are AI citations the same as AI recommendations?
No. An AI citation identifies a source, while an AI recommendation suggests a product, service or business as an option. A website may be cited without its business being recommended.
Can ORBA tell me what AI understands about my website?
ORBA analyses business website content to assess how clearly the website communicates important information about the business and identify opportunities for improvement. Its assessment is not a guarantee of how every AI model will interpret or cite the website.
