Sign up for Winter Keynote, Feb 2, 16:30 CET

Save the date →

Beyond Rankings: How AI Is Changing Discovery, and Why Trust Now Starts Before the Click

AI Discoverability Series, Part 1 of 3

Cantarus
Written by Cantarus

AI answers now shape the shortlist before a buyer ever visits a website. In part one of our series with Cantarus, we look at how discovery is changing and why trust now starts before the click.

Imagine a potential buyer asking an AI assistant for recommendations.

The platform compares the information it can find and produces a shortlist of two or three organizations. By the time that buyer visits a website or contacts a partner, part of the evaluation may already have taken place.

The organizations included in that answer have entered the buyer’s consideration. Those left out may never know that the opportunity existed.

The same pattern can apply whether someone is selecting a supplier, joining an association, choosing a service, evaluating a platform or looking for authoritative guidance.

This is the emerging reality of AI-mediated discovery. It does not mean that websites or search engines have stopped mattering. It means that an increasingly important part of the journey can now happen before the website visit, and sometimes without one.

This is the first in a three-part series from Umbraco and Cantarus exploring how AI is changing online discovery and what organizations can do about it.

In this opening article, we examine the shift itself: how AI-generated answers are changing the relationship between visibility, website traffic and trust. Part two will consider whether organizations are ready to be understood and referenced by AI. Part three will explore how to make that readiness repeatable through content, technology, authority and governance.

Discovery no longer begins with a list of links

For years, digital visibility followed a relatively familiar path.

Someone entered a search query, reviewed a list of results and clicked through to one or more websites. Search Engine Optimization helped organizations improve their chances of appearing, and appearing prominently, within that journey.

That journey still exists. But it increasingly sits alongside another model.

Search engines and AI assistants can now interpret the question, retrieve information from multiple sources and assemble an answer around the user’s need. The user might still follow a link, but they can also learn, compare options and begin forming an opinion without leaving the interface.

The website remains essential. It is where an organization publishes its knowledge, explains its services, demonstrates its experience and supports deeper journeys. But it is no longer guaranteed to be the audience’s first encounter with that information.

A customer, member, supporter or stakeholder might first encounter an organization through:

• a Google AI Overview or AI Mode response;

• an answer from ChatGPT, Gemini, Claude or Perplexity;

• a comparison assembled from several sources;

• an external article or case study cited within an answer; or

• a recommendation shaped by what others have published about the organization.

AI is creating a new interpretive layer between organizations and their audiences.

Google search results page with query “Best coffee in Copenhagen” and heading “AI Overview”

Why this matters commercially

This is not simply a change to the search-results page.

AI-generated answers can affect which organizations enter someone’s consideration, how their expertise is represented and whether their claims appear credible.

In a traditional journey, an organization could observe search traffic, website behavior, inquiries and conversions. In an AI-mediated journey, some of the evaluation may happen before the person reaches any channel the organization owns.

They might ask:

• Which platforms can support a secure enterprise website?

• Which CMS is best suited to a particular sector or technical environment?

• Does this platform support the integrations we require?

• Which implementation partners have relevant experience?

• What evidence shows that they can deliver it?

• What are the limitations or risks of each option?

An AI system may answer those questions by combining information from a platform provider, implementation partners, client stories, technical documentation, media coverage and other third-party sources.

By the time someone contacts a supplier or partner, they may already have a shortlist, and an emerging view of who they trust.

AI discoverability is therefore not only a question of marketing reach. It can influence reputation, consideration and demand before a conventional lead is visible.

“AI is increasingly shaping the shortlist before an organization knows a buyer is looking. The opportunity is to make sure your expertise can be found, understood and trusted wherever that decision begins.”

Dani Barker, Marketing Services Manager, Cantarus

The evidence has moved beyond speculation

Research into AI discovery is developing quickly, and individual statistics should be treated with care. Platforms, interfaces and user behavior are still changing. However, the overall direction is increasingly clear.

Pew Research Center found that Google users were less likely to click a traditional search result when an AI summary appeared. This provided important evidence that appearing within the search environment would not necessarily generate the website traffic organizations had historically expected.

Research published by Ahrefs in February 2026 examined 300,000 keywords. It found that the presence of an AI Overview correlated with a 58% lower average click-through rate for the top-ranking result.

The figure should not be treated as a universal outcome for every query or sector, but it reinforces the need to separate visibility from traffic.

At the same time, AI platforms are creating new referral journeys. Similarweb estimates that they produced an average of 770.7 million referral visits per month worldwide between June 2025 and May 2026, an increase of 117.4% over the preceding year.

The volume, growth rate and importance of those visits varied significantly by sector.

These findings are not contradictory.

AI-generated answers can reduce clicks from some traditional search journeys while producing new referral traffic of their own. They can also influence awareness and consideration without generating a measurable visit at all.

Traffic remains important. It is simply no longer a complete measure of discovery. 

SEO, AEO and GEO: related disciplines, not competing camps

The growth of AI search has brought a new collection of terms into marketing and digital conversations.

Search Engine Optimization, or SEO, improves the ability of search engines to discover, understand and rank content. Its foundations include technical accessibility, relevance, helpful content, internal linking, authority and a strong user experience.

Answer Engine Optimization, or AEO, puts additional emphasis on whether content provides useful, direct and verifiable answers that can be understood and surfaced through answer-based experiences.

Generative Engine Optimization, or GEO, is commonly used to describe efforts to improve how an organization and its content are represented or referenced in AI-generated responses.

In practice, AEO and GEO overlap considerably. Niels Christian Laursen uses AEO as the umbrella term in his practical Umbraco audit while recognizing GEO as another common label for much the same area.

The terminology is less important than the underlying principle: AI discoverability builds on good SEO, good content and good publishing. It does not replace them.

Purple magnifying glass icon with star and “cantarus” label on a dark blue background.

“An AEO audit improves the parts of AI-search visibility you can control, but it cannot guarantee that an answer engine will cite the page.”

Niels Christian Laursen, Umbraco

Google’s guidance confirms that established SEO practices remain relevant to its generative AI features. There are no separate technical requirements or special forms of markup that guarantee inclusion in AI Overviews or AI Mode.

That distinction matters. Organizations should be cautious of anyone presenting AEO or GEO as a collection of guaranteed shortcuts.

Rankings still matter, but they no longer tell the whole story

Search rankings remain useful. Organic traffic remains valuable. Neither should be abandoned because AI-generated answers are growing.

But a ranking tells an organization where a page appeared in one type of result. It does not fully reveal whether an AI system:

  • retrieved information from the page;

  • cited or linked to it;

  • mentioned the organization using information found elsewhere;

  • represented the organization accurately;

  • selected a competitor or third-party source instead;

  • contributed to a shortlist or recommendation; or

  • produced a different answer when the question was rephrased.

Exploratory industry research published by SparkToro in collaboration with AI-visibility platform Gumshoe illustrates why AI results should not be treated like conventional rankings.

In a crowdsourced study involving 600 volunteers, 12 prompts and 2,961 responses across ChatGPT, Claude and Google’s AI experiences, the researchers found considerable variation in the brands recommended, their order and the number included.

The study was not peer-reviewed, and Gumshoe has a commercial interest in AI measurement, so its findings should be treated as directional rather than definitive. However, its transparent methodology reinforces a sensible measurement principle: repeated visibility across a broad set of relevant prompts is more meaningful than the result of one prompt on one occasion.

AI recommendations can vary between platforms, prompts, users and points in time. That makes a single universal “AI ranking” or visibility score attractive, but potentially misleading.

The goal is not to secure one permanent position in an AI-generated answer. No responsible platform, CMS or optimization provider can guarantee that.

The goal is to improve the parts of discovery that the organization can control and measure them with appropriate caution.

A leadership framework for AI discoverability

For this series, we use three connected questions to help leaders organize the challenge.

They are not a formula for how an AI system ranks or selects sources. They are a practical way to identify what an organization can improve.

Can you be found?

Relevant information must exist in a form that search and AI systems can access.

That includes established technical foundations: crawlable and indexable pages, coherent published HTML, working internal links, intentional crawler controls and content that is not trapped inside inaccessible interfaces.

Both central brand content and relevant partner content matter. A global platform website may answer the principal question, while a partner provides specialist, sector-specific or local evidence.

Can you be selected?

Being available does not make a source useful.

Content needs to address a genuine audience question and provide enough relevant detail to cover the scenarios in which the organization should be considered.

This does not mean producing pages filled with every conceivable phrase. It means answering real questions precisely and contributing something more valuable than a generic summary.

Useful content may include:

  • a clear answer;

  • distinctive expertise;

  • a practical explanation;

  • relevant technical detail;

  • original data;

  • evidence from implementation; and

  • an honest account of limitations or conditions.

The more useful detail an organization publishes, the more legitimate questions it may be equipped to answer. But volume alone is not the objective. Detail needs structure, consistency and purpose.

Can you be trusted?

AI systems do not rely solely on what an organization says about itself.

Large-scale industry research from Muck Rack’s Generative Pulse team examined more than 25 million links cited in responses from ChatGPT, Claude and Gemini. Using a broad definition that includes journalism, academic research, government and nonprofit sources, encyclopedic sites, social platforms and third-party corporate content, Muck Rack classified approximately 84% of citations as earned rather than owned or paid. Journalistic sources alone accounted for around 27%.

Muck Rack provides commercial media-intelligence and AI-visibility services, so the findings should be treated as industry evidence rather than independent academic research. Nevertheless, the scale of the dataset supports an important principle: what authoritative sources say about an organization can contribute to how it is represented in AI-generated answers.

Trust can be supported by:

  • evidence and clearly attributed sources;

  • visible authorship and subject-matter expertise;

  • accurate and consistent information;

  • relevant external coverage;

  • authoritative partner contributions;

  • detailed case studies;

  • named client outcomes; and

  • content that is reviewed and maintained.

This is where Umbraco, its partners and their clients can create something stronger together.

Umbraco can publish authoritative answers about the platform and its capabilities. Partners can contribute specialist, sector, regional and implementation expertise. Case studies can demonstrate how that expertise works in practice.

The core answer, specialist contribution and proof do different jobs. Together, they create a more credible content ecosystem.

Question card reading “Can you be… Found? Selected? Trusted?” with icons and blue progress bars

AI search needs proof, not more marketing

“AI search needs proof, not marketing” is a useful discipline rather than a literal rule.

Brand-owned content still matters, but unsupported claims become less persuasive when an answer can be assembled from product documentation, independent coverage, expert sources and proven implementation experience.

AI systems already have access to enormous volumes of polished but generic content.

Publishing another article that repeats widely available definitions is unlikely to create lasting value for the reader or provide a compelling reason for an AI platform to select it.

Organizations have a stronger contribution to make when they publish what they genuinely know:

• how a particular integration was delivered;

• which decisions affected the outcome;

• what constraints had to be managed;

• what changed for the customer or user;

• what the results demonstrate;

• where an approach is appropriate; and

• where its limitations lie.

A claim such as “our platform is suitable for complex organizations” is marketing.

A detailed explanation of the architecture, governance, integrations, accessibility requirements and measured outcomes behind a complex implementation is evidence.

A stronger approach is not to write mechanically for an algorithm. It is to publish material that helps a person make a better-informed decision and ensure that machines can access and interpret it.

Avoiding duplication: one ecosystem, different contributions

If platform providers, partners and clients all publish lightly rewritten versions of the same article, they add volume without adding much knowledge.

A stronger model gives each participant a distinct role.

Umbraco can provide authoritative platform answers and connect related expertise. A partner can contribute a named expert perspective, explain how something works in practice or answer a question specific to a market or sector. Client evidence can show that the approach has produced a real outcome.

This means Umbraco and its partners do not need to compete to own identical content.

They can create a connected path:

• The central answer: clear platform information published by Umbraco.

• The specialist perspective: expertise contributed by a named partner practitioner.

• The proof: relevant client work, implementation detail and measurable outcomes.

• The next step: a route to the most appropriate partner, resource or conversation.

This is not simply a content-distribution model. It is a trust model.

It also explains why the principal articles in this series live on Umbraco, while Cantarus contributes expertise, evidence and complementary resources through its own channels.

Is an AEO-ready CMS the answer?

It is tempting to reduce this challenge to a technology-purchasing question:

Is our CMS AEO or GEO compliant?

That is not the most useful question.

A capable CMS can provide the foundations for content to be structured, governed, maintained and published effectively.

Umbraco can support reusable content patterns, consistent fields, visible ownership, review dates, technical controls and well-implemented public output.

But a CMS cannot manufacture expertise, establish third-party consensus or guarantee that an AI platform will select a particular source.

The better question is:

Does our content ecosystem give people and AI systems a reason to trust our answer?

That ecosystem includes:

• the content management platform;

• the quality of the published implementation;

• the organization’s expertise and evidence;

• its content operations and governance;

• partner and client contributions;

• external authority; and

• its ability to measure and improve.

AEO is not a CMS feature. AI discoverability is the combined result of what an organization knows, how it publishes, who supports its claims and how well the information is maintained.

What should digital leaders ask now?

Organizations do not need to respond by rebuilding their entire digital estate around AI search.

They should begin with a more disciplined set of questions:

  1. Which audience questions influence consideration and decisions?

  2. What information do we publish that genuinely answers them?

  3. What expertise, evidence or experience do we contribute that another source cannot?

  4. Can search and AI systems access and interpret the published result?

  5. Is our information consistent across our website, partners and authoritative external sources?

  6. Who owns its accuracy and decides when it should be reviewed?

  7. What evidence will tell us whether discovery is changing?

These questions move the conversation beyond isolated optimization tactics.

They also reveal why responsibility cannot sit with one content editor, developer or SEO specialist. AI discoverability connects marketing, subject-matter expertise, technology, communications, client evidence and governance.

Start with one page, but be prepared to look wider

The scale of the challenge can feel daunting. The sensible starting point is not every page.

Niels Christian Laursen’s practical AEO content audit for Umbraco provides a 10-minute first pass for one important page.

It examines whether the page answers a genuine question, provides original and verifiable value, can be discovered through its published output, and has clear ownership and a measurement plan.

It also draws an important boundary: the audit can improve the elements within an organization’s control, but it cannot guarantee that a search engine or AI platform will cite the page.

One page is a practical place to begin. But its weaknesses may reveal larger questions:

  • Is information structured consistently across the site?

  • Do different teams use conflicting terminology?

  • Are important claims supported by evidence?

  • Are authorship, ownership and review responsibilities clear?

  • Does the technical implementation produce accessible public content?

  • Is valuable expertise held by individuals but absent from the website?

  • Do external sources support, or contradict, the organization’s own claims?

  • Can the organization measure anything beyond traffic?

At that point, the challenge is no longer simply page optimization. It is organizational readiness.

Lowres-Cantarus-Lifestyle-36.jpg (1)

Beyond rankings

AI has not made websites, SEO or search rankings obsolete.

It has made the path between a question and a decision less visible, less linear and more dependent on an ecosystem of sources.

Organizations must still earn search visibility. Increasingly, they must also ensure that their knowledge can be found, their contribution is useful enough to be selected, and their evidence is strong enough to be trusted.

The organizations best placed to adapt will not be those chasing every new acronym or supposed algorithmic shortcut.

They will be those that can turn their knowledge into clear, useful and well-supported content; publish it through strong technical foundations; establish authority beyond their own website; and maintain it through clear ownership and governance.

That is the shift from optimizing individual pages to building an AI-discoverable organization, and it is where this series goes next.

Next in the series

Part 2: Is Your Organization Ready to Be Understood and Referenced by AI?

The next article moves from the changing discovery landscape to organizational readiness. It will connect Niels’ page-level audit with the wider content, platform, governance, authority and measurement capabilities required across an organization’s digital estate.