How AI Search Engines Choose Which Websites to Cite

Something has changed fundamentally about how people find information online, and it has implications for every South African business with a website.

When someone uses Google AI Overviews, asks ChatGPT a question, or searches on Perplexity, they do not get a list of blue links to visit. They get a direct answer, generated by an AI, often with a small number of cited sources. Being one of those cited sources is the new version of being on page one of Google - and the rules for getting there are somewhat different from traditional SEO.

This article explains how AI search systems decide which websites to use as sources, and what you can do to improve your chances of being cited.

How Do Google AI Overviews, ChatGPT, and Perplexity Actually Find Sources?

AI search engines like Google AI Overviews, ChatGPT, and Perplexity choose sources based on relevance, content clarity, and authority signals. They strongly favour pages that answer questions directly and early, use clear headings, demonstrate genuine expertise, and carry structured data like FAQ schema. For South African businesses, the practical implication is that well-structured content written in plain, factual language on a technically sound website is more likely to be cited than content that buries answers or lacks clear organisation.

Each platform works differently under the hood, but there are common principles.

Google AI Overviews sit at the top of certain Google search results and provide a synthesised answer to the search query, with small links to the sources Google used. Google’s AI systems read web pages from Google’s existing search index (the same index that powers traditional Google search), extract the most relevant passages, and synthesise a response. The sources cited are the pages Google determined were most relevant and authoritative for the query.

Perplexity performs live web searches at the time of your query, reads the top results, and synthesises an answer with numbered citations. It functions much like a research assistant that reads several articles and then summarises what they say. The sources it uses are largely determined by what ranks well in search (particularly Bing, which Perplexity draws from heavily) combined with Perplexity’s own quality assessment.

ChatGPT with Browse (when web search is enabled) searches the web in real time and reads pages to answer questions. Without Browse, ChatGPT draws on its training data, which has a knowledge cutoff date. When web search is on, it behaves similarly to Perplexity - searching, reading, and citing sources.

Bing Copilot is deeply integrated with Microsoft’s Bing search engine. It searches Bing, reads top results, and generates responses with citations. For South African businesses, Bing is often underestimated as a traffic source - but via Copilot, it increasingly matters.

What Makes a Website More Likely to Be Cited by AI Search Tools?

This is the question that matters. AI citation is not random. There are identifiable characteristics of content that AI tools prefer.

Clear, direct answers to specific questions. AI systems are looking for content that can answer a question unambiguously. A page that buries the answer in three paragraphs of preamble is less useful to an AI extractor than a page that states the answer clearly in the first sentence. Think about structuring your content the way a journalist writes a news story: the most important information first.

Well-structured content with meaningful headings. AI tools read the structure of your page. Headings (H2, H3) act as signposts that tell the AI what each section is about. A page with clear, specific headings - particularly headings phrased as questions - is much easier for AI systems to parse and extract from.

Topical authority and depth. AI tools are more likely to cite sites that have demonstrated genuine expertise on a topic through multiple pieces of content. A plumbing company that has ten well-written blog posts covering different aspects of plumbing is more likely to be cited on plumbing questions than one with a single generic page. This is why a content strategy built around your area of expertise matters.

E-E-A-T signals. Google’s E-E-A-T framework (Experience, Expertise, Authoritativeness, and Trustworthiness) is relevant not just for traditional SEO but for AI citation too. Google AI Overviews draw from the same quality assessments Google uses for traditional rankings, which means pages that demonstrate real expertise through author credentials, cited sources, and verifiable facts are favoured.

Factual accuracy and verifiability. AI systems have a strong preference for content that makes claims supported by evidence, data, or attribution. Unsubstantiated claims, vague assertions, and filler content are less likely to be used as sources. If you cite a study, name the source. If you state a fact, make it verifiable.

Schema markup. Structured data helps AI systems understand your content with less effort. FAQ schema, in particular, maps perfectly onto the question-and-answer format that AI search tools use to generate responses.

What Is llms.txt and Should Your Website Have One?

A llms.txt file is an emerging convention (not yet an official standard) that some websites are using to provide AI language models with a structured, curated summary of their content.

The concept is borrowed from robots.txt (which tells search crawlers which pages to visit or avoid). A llms.txt file, placed at yourwebsite.co.za/llms.txt, provides AI models with a clean, plain-text summary of what your site contains and what it is about - formatted specifically for LLM consumption rather than human reading or traditional crawler parsing.

For most South African small businesses, implementing llms.txt is not urgent. It is more relevant for larger sites with complex content structures. However, it is worth knowing about because it represents the direction things are moving. AI tools are increasingly a primary way people find information, and the infrastructure for helping AI tools navigate your content will become more important over time.

What is more immediately useful for most businesses is making sure your robots.txt file does not accidentally block AI crawlers. Google’s AI crawlers, Perplexity’s crawler, OpenAI’s crawler (GPTBot), and others use specific user agent names. If your robots.txt blocks all unknown bots (a common security recommendation that is now counterproductive in some cases), you may be blocking AI tools from reading your content.

Check your robots.txt at yourwebsite.co.za/robots.txt and look for any Disallow: rules that apply to AI-specific crawlers. If you want to specifically allow or disallow particular AI crawlers, you can add rules for their user agent names:

• Google: Googlebot (same as traditional Google crawling)
• OpenAI: GPTBot
• Perplexity: PerplexityBot
• Anthropic: ClaudeBot

Most businesses want to allow all of these. If your website is intentionally blocking them, consider whether that is still the right decision given how much of search behaviour is now AI-mediated.

What Are E-E-A-T Signals and How Do They Affect AI Citation?

Google’s quality rater guidelines use E-E-A-T as a framework for assessing page quality: Experience, Expertise, Authoritativeness, and Trustworthiness.

Experience refers to first-hand experience with the subject matter. A tradesman writing about a plumbing technique they use daily carries different weight than a content writer describing the same technique based on research.

Expertise means demonstrated knowledge in a field. A dentist writing about dental procedures, or an attorney writing about property law, signals expertise. On a website, this is shown through credentials, depth of content, accuracy, and how well the site covers its subject area.

Authoritativeness is about your reputation in your field - whether other credible sources cite you, link to you, or mention you. It is earned over time and cannot easily be faked.

Trustworthiness encompasses accuracy, transparency, and honest representation. Sites with clear authorship, accurate factual claims, working links, and honest business information score higher.

For AI citation specifically, E-E-A-T matters because AI systems are trying to provide accurate, reliable answers. They are more likely to cite sources that signal quality and trustworthiness than those that do not.

Practical ways to improve E-E-A-T on your website:

• Add proper author bios to blog posts and articles, with real names and credentials
• Keep your About page updated with genuine information about your team and experience
• Fix factual errors promptly if they are pointed out
• Use citations and links to authoritative sources when making factual claims
• Maintain a consistent, professional tone across all content

How Should You Optimise Content to Appear in AI Answers?

The practical content recommendations are straightforward.

Answer the question directly and early. Do not make an AI tool (or a human reader) hunt for the answer buried in paragraph five. State it clearly in the first sentence or two of each section.

Structure content around questions. Use headings phrased as questions your audience actually asks. These match the question format that AI tools use to query their sources, making it much more likely your content is pulled when someone asks a similar question.

Write in clear, specific language. Vague generalities are hard for AI systems to extract usefully. “Prices vary depending on your requirements” is unhelpful. “Our SEO maintenance service starts at R850 per month” is citable.

Maintain depth and comprehensiveness on your core topics. A single, well-researched, comprehensive page on a topic you genuinely know about is worth more for AI citation than ten thin pages.

Keep your content accurate and updated. AI tools that detect outdated or inaccurate information on your site will deprioritise it. Review your key content pages regularly.

AI search is not replacing traditional SEO - it is layering on top of it. The fundamentals are the same: good, trustworthy content on a technically sound website. The specific optimisation for AI involves writing more directly, structuring more clearly, and building the authority signals that make you a reliable source.

Our full guide to AEO (Answer Engine Optimisation) covers this topic in much more depth if you want to go further. And if you want your ongoing SEO strategy to account for both traditional and AI search, our monthly SEO maintenance keeps your site optimised across both dimensions.

Want your site cited by AI search and ranking on Google? Our SEO services build the authority and structure that both need. Request a quote to get started.

Frequently Asked Questions About AI Search and Website Citation

Can I pay to be cited by Google AI Overviews or ChatGPT?

No. Neither Google AI Overviews nor ChatGPT (or Perplexity, or any mainstream AI search tool) sells citation placement. The sources cited are determined by the AI systems’ assessment of relevance, quality, and authority - not by payment. This is actually good news: it means that a small South African business with genuinely excellent, well-structured content on its core topic can be cited alongside or instead of much larger competitors, purely on merit. The only way to earn AI citation is to earn it.

Does being cited by AI search tools increase traffic to my website?

Sometimes, but not always. Some AI tools display answers so completely that the searcher gets what they need without clicking through to any source. Google AI Overviews in particular have been associated with reduced click-through rates on some query types. However, being cited still builds brand visibility and credibility. When a potential customer sees your business cited as a source by an AI tool they trust, that is a meaningful credibility signal - even if they do not click through at that moment. AI citation is increasingly a brand awareness channel, not just a traffic source.