How search engines and AI decide which sources to trust

More of your customers are starting their research without ever touching a list of blue links. They ask ChatGPT which tool to buy, ask Gemini who leads a category, or read an AI overview at the top of Google and never scroll. In every one of those moments, a machine is deciding which sources to trust and which to ignore. Understanding how that decision is made is now central to whether your brand is visible at all.

The encouraging part is that the signals search engines and AI answer engines rely on are not mysterious, and they are not new. Both systems are trying to answer the same question a careful researcher would ask: who is credible here, and can I rely on what they say? This article walks through how search engines judge trust, how that logic has carried into AI answers, and what the evidence shows about becoming a source these systems actually cite.

How search engines learned to judge trust

Search engines have never been able to read a page and simply know whether it is true. So from the beginning they looked for proxies for trust, and the most durable proxy has been the link. When one site links to another, it is casting a vote of confidence, and a page that earns many votes from credible sources looks more authoritative than one that earns none. This is the original logic behind Google’s ranking system, and it still holds. Backlinko’s analysis of 11.8 million search results found that the number one result has an average of 3.8 times more backlinks than positions two through ten. Links remain one of the clearest signals of trust a page can accumulate. If you want the fuller picture, our why backlinks matter explains the mechanics in depth.

Links are not the whole story. Google evaluates relevance, content quality and a framework it calls E-E-A-T, which stands for experience, expertise, authoritativeness and trustworthiness. The three pillars beyond links exist because a vote is only as good as the voter. A link from a recognized authority in a field carries more weight than a link from an anonymous, low-quality site, which is why judging link quality matters far more than raw volume. Scarcity sharpens the point further: Ahrefs found that 66.31% of pages have no backlinks at all, so even a handful of credible links can separate you from most of the web.

What Google actually means by E-E-A-T

It is worth quoting Google directly, because E-E-A-T is widely misunderstood. In its official guidance on creating helpful content, Google explains that its systems identify a mix of factors that can help determine which content demonstrates aspects of experience, expertise, authoritativeness, and trustworthiness. Crucially, it adds that of these aspects, trust is most important, and the others contribute to trust. Google is explicit that E-E-A-T itself is not a single ranking factor, but rather a way of describing the qualities its automated systems are built to reward.

The same guidance tells creators to ask Who, How, and Why about their content: who created it, how it was produced, and why it exists. It recommends clear authorship, bylines that link to author backgrounds, and content that demonstrates first-hand experience, such as expertise that comes from having actually used a product or visited a place. Google also notes that it gives even more weight to strong E-E-A-T for topics that affect health, financial stability or safety, which it calls Your Money or Your Life topics. The throughline is consistency: a trustworthy source identifies who stands behind its claims and shows the work, and that reputation is built well beyond your own website.

How that logic carried into AI answer engines

AI answer engines did not invent a new theory of trust. When ChatGPT, Gemini or an AI overview composes an answer, it typically retrieves live web sources, selects the ones it judges most reliable, and synthesises a response with citations. The retrieval step leans on much of the same infrastructure as search: the systems favor sources that are already established, authoritative and frequently referenced elsewhere. In other words, the question of who gets cited rhymes closely with the question of who ranks.

There is an important difference in stakes. A traditional results page shows ten links and lets the reader choose. An AI answer often names only a few sources, or summarizes without sending a click at all. Research on news citation patterns found that a small set of top sources captured a majority of all citations on the leading AI platforms, which means these systems concentrate trust on a relatively small set of names. Being one of those names is a far narrower target than ranking on page one, and it raises the premium on the credibility signals that help you get cited in AI answers in the first place. Visibility is shifting from can a user find you to will a machine vouch for you, and that is a higher bar.

The evidence: earned media dominates AI citations

The clearest data on what AI engines actually cite comes from Muck Rack’s Generative Pulse study, which analyzed more than 25 million links from ChatGPT, Claude and Gemini responses across 17 industries. The headline finding is striking. Earned media accounts for 84% of all AI citations, journalism alone makes up 27% of cited sources, and paid and advertorial content accounts for just 0.3%.

Two details make the case stronger. First, the numbers have held steady. Across three editions of the study going back to mid-2025, earned media has ranged from 82% to 89%, which suggests this is structural rather than a quirk of one model update. Second, the pattern is corroborated elsewhere: Nieman Lab reported on the same research line, noting that journalistic content was cited around 27% of the time overall and jumped to roughly 49% for queries that implied a level of recency. The takeaway for any brand is blunt. If you are not present in the coverage AI is reading, you are not present in the answers AI is giving.

Why links and brand mentions are the connective tissue

Look closely at both systems and the same mechanism keeps appearing: links and brand mentions in credible places. A search engine treats a link from a trusted publication as a vote. An AI engine, retrieving sources to answer a question, gravitates toward the names that appear most often in trustworthy, well-referenced coverage. The currency is the same. Coverage and citations in respected outlets tell both systems that other credible parties take you seriously, and that is precisely the signal each is built to detect.

This is why earned coverage outperforms content you publish only on your own domain. A page on your site asserts that you are an expert; a feature in a respected publication, with a link or a clear brand mention, is independent confirmation. That distinction is the heart of earning links through PR and the reason links given editorially earned through genuine coverage carry more weight than links you place yourself. It also explains a frequent frustration: brands that publish constantly but earn little outside validation often plateau, because they keep telling search engines and AI models how good they are without anyone else confirming it.

What makes a brand a citeable source

If you reverse-engineer what gets cited, three qualities recur.

Original information

The first is original information. Google’s own guidance asks whether content provides original information, reporting, research, or analysis, and AI engines favor passages with specific, verifiable facts. Proprietary data, surveys and benchmarks give both systems something concrete to anchor to, which is far more citeable than a rephrasing of what everyone already says.

Demonstrable expertise

The second is demonstrable expertise. Named authors with real credentials, first-hand experience and transparent sourcing are the human signals of trust that Google describes and that AI models learn to recognize.

Coverage in trusted outlets

The third, and the one most brands underuse, is coverage in outlets the systems already trust. Muck Rack found that AI citations skew heavily toward journalism and established publishers, and the Nieman Lab reporting showed citations concentrating on major outlets. You cannot manufacture that standing on your own site. You earn it through coverage, which is the work of our digital PR service and disciplined pitching bloggers. Original data, genuine expertise and third-party validation are the combination that turns a brand into a name worth citing.

The practical playbook to become a cited source

Turning these principles into visibility follows a repeatable sequence. It begins with something genuinely worth citing and ends with measurement, and the middle is the unglamorous work of earning trust the same way a credible source always has. The steps below map the path most directly to how both search and AI systems decide whom to rely on.

Each stage compounds on the last. Original assets give journalists a reason to cover you; that coverage produces the links and mentions both systems read as votes of confidence; and consistent expertise across your own pages makes you safer to cite once you have been discovered. For the full method, our digital PR outreach guide and the broader earning links go deeper on execution, while how to get backlinks covers the tactical groundwork. The point is to be deliberate: aim coverage at the question types and outlets your audience and the AI models actually rely on, rather than chasing volume for its own sake.

How to measure your presence in search and AI answers

You cannot manage what you do not measure, and AI answers have made measurement harder because there is no single ranking report. Start with the classic foundations, because they still drive AI retrieval:

• Referring domains. Track the referring domains and the quality of links you earn.
• Organic visibility. Track organic visibility for your priority topics.
• Branded coverage. Track the volume and prominence of branded coverage.

These are the inputs that feed both systems, and our guide to the metrics that matter sets out a sensible framework.

Then add an AI layer. Periodically prompt ChatGPT, Gemini and other engines with the questions your customers ask about your category, and record:

• Mentions. Whether you are mentioned.
• Citations. Whether you are cited as a source.
• Competitors. Which competitors and publications appear instead.

Pay attention to query type, since Muck Rack found that industry trend questions drew different sources than how-to questions, and press releases appeared far more often in trend responses. Watching how your share of mentions changes as your coverage grows is the closest thing to a rank-tracker for AI answers, and pairing it with the wider context in our the latest link data keeps your targets grounded in evidence rather than guesswork.