At Macrocosm Ultra Digital, we spend a great deal of time testing search. Not only because SEO is one of the services we provide, but because the way people discover businesses is changing quickly. Search engines are no longer the only systems deciding what gets seen. AI platforms are increasingly researching, comparing and recommending businesses on behalf of the user, which raises an important question: what does AI actually consider to be authority?

A recent conversation we had with Gemini gave us an interesting answer. We asked it to identify the top SEO companies in Cape Town and Macrocosm was not included in its initial list. That caught our attention because we compete organically at the top of an exceptionally competitive search market, against agencies whose own business is getting websites to rank. Our positions are earned organically, not through paid Google placements, which makes our own search performance one of the clearest demonstrations of the service we sell.

So we challenged the answer.

Gemini explained that its original recommendations had been influenced by regional agency assessments, industry directories, professional review platforms and other third-party sources. When we questioned whether those signals should carry more weight than demonstrated organic search performance, its reasoning shifted. Gemini acknowledged that ranking at the top of a highly competitive SEO market is a significant proof of capability and recognised that directory prominence does not necessarily provide the same evidence of technical execution.

For us, that conversation was far more interesting than whether Macrocosm appeared in a particular top-three list. It exposed a much bigger issue that we believe businesses, marketers and AI optimisation specialists are going to have to address.

Traditional search has always involved multiple authority signals, but there is something fundamentally different about an AI recommendation. When Google presents a page of search results, users can investigate the options themselves. They can compare websites, look at rankings, read reviews, consider paid advertisements separately from organic results and decide which business deserves their attention.

AI compresses much of that process into an answer.

When someone asks for the “best SEO company in Cape Town”, they are not necessarily asking for ten links to investigate. They are increasingly expecting the system to investigate those options for them and return a conclusion. That means the AI is no longer simply helping the user find information. It is interpreting the available evidence and deciding which businesses deserve to be mentioned.

Recent research reinforces just how different this environment can be. A 2026 study comparing Google Search, Google AI Overviews and Gemini across 11,500 queries found substantial differences between the sources retrieved by traditional search and generative search systems. The researchers recorded an average source overlap of less than 0.2 between the different search experiences, demonstrating that strong visibility in one environment does not automatically mean identical visibility in another.

That changes the optimisation conversation significantly.

The Gemini test led us to something we have started thinking about as Structured Authority Bias.

This is not a claim that every AI system always behaves in the same way, nor do we believe that directories and review platforms have no value. Instead, it is a working idea around a pattern we are actively exploring: AI systems may sometimes find structured representations of authority easier to interpret than more complex evidence of real-world performance.

A directory is particularly easy for a machine to understand. Companies have been categorised, services have been labelled, reviews have been gathered, locations have been assigned and businesses may already have been arranged into rankings or comparison lists. Much of the difficult work of turning a market into structured information has already been done.

Organic performance tells a different story. If an SEO agency consistently outranks other SEO agencies for commercially valuable searches, that performance is meaningful because every competitor is effectively demonstrating its ability in the same public arena. Understanding the value of that achievement, however, requires more than reading a list. The system needs to understand the competitiveness of the query, the distinction between paid and organic visibility, search intent, consistency across related searches and what the ranking itself says about the capability being evaluated.

Our question is whether AI is always making that distinction well enough.

This becomes particularly important when commercial directory models enter the picture.

Clutch, for example, provides its own ranking methodology based on factors including reviews, market presence and service specialisation. It also openly offers sponsored placement. According to Clutch’s current methodology, companies can pay for higher placement within specific directory pages, with sponsored businesses appearing above organic directory results by default. Clutch is also clear that sponsorship does not change a company’s underlying organic Clutch Rank.

That distinction is important, and it is why simply describing these platforms as “fake” would miss the real issue.

The problem we are interested in sits one level further down the chain. If an AI system uses a directory as evidence when deciding which company to recommend, how confidently can it distinguish commercial prominence on that source from broader market authority? More importantly, how does it balance that information against direct evidence that another business is outperforming competitors organically in the very service being evaluated?

A sponsored position can legitimately make a business more visible on a commercial directory. An organic search position is earned through an entirely different competitive process. Both are signals, but they do not mean the same thing.

AI needs to understand the difference.

There is a temptation whenever a new search discipline emerges to declare that the previous one is dead. We do not subscribe to that thinking.

SEO remains fundamental to digital visibility and Google continues to say exactly that. In its 2026 guidance for generative AI features, Google states that established SEO best practices remain relevant because its generative search experiences are rooted in its core Search ranking and quality systems. Google specifically points to technical accessibility, useful original content, clear site structure and other established SEO fundamentals as important foundations for generative search visibility.

What is changing is the environment surrounding those rankings.

A company can have exceptional organic visibility while an AI platform constructs its recommendation from a different collection of evidence. The question therefore moves from “Do we rank?” to “Does the machine understand why we rank, what that says about our expertise and whether that expertise should influence its recommendation?”

That is where AI Optimisation starts to become strategically important.

We do not believe AI Optimisation should become an exercise in producing content specifically for machines or finding shortcuts to manipulate recommendation systems. That would simply recreate many of the worst practices from earlier eras of SEO under a different acronym.

The real opportunity is to make genuine authority easier to discover, interpret and verify.

A brand’s website remains central, but the wider evidence surrounding that brand is becoming increasingly important. Search performance, original research, expert content, client experiences, independent mentions, accurate business information, structured data, topical authority and consistency across the digital ecosystem all contribute to the picture an AI system may eventually construct.

This is also why our approach to visibility extends beyond traditional SEO. At Macrocosm, our Visibility Quintet brings together SEO, GEO, AEO, AI SEO and LLMO because we see these areas as connected parts of the same problem. Search engines need to find the business, generative systems need to understand it, answer engines need to extract useful information from it and language models need enough consistent evidence to represent it accurately.

The objective has moved beyond appearing somewhere online. It is ensuring that when a machine is asked to interpret your market, it has sufficient evidence to reach the right conclusion.

One of the lessons from our Gemini conversation is that businesses should begin testing AI very differently.

Asking an AI system what it knows about your company is useful, but it is not enough. The more revealing questions are the ones your potential customers are likely to ask before they have decided which company they want.

Who are the best providers in the market? Which company would you recommend? Who specialises in this particular service? Which businesses appear to have the strongest expertise? Why would you choose one provider over another?

Then the important part begins: examine why the AI reached its conclusion.

Which sources influenced it? What evidence did it consider important? Which competitors appeared repeatedly? Did it understand your specialisms correctly? Did it overlook information that should have materially changed the answer?

Those questions begin to expose the difference between being visible on the internet and being understood by the systems increasingly responsible for navigating it.

For an industry built around visibility, that difference is enormous.

For years, businesses have thought carefully about the reputation they build with customers and the authority they build with search engines. We are now adding a third layer: the reputation that AI constructs from the information it can retrieve.

That reputation may not always match reality.

A business can be excellent at what it does but poorly represented across the sources an AI system relies on. Another can be extremely visible within machine-friendly sources without necessarily demonstrating the same level of real-world performance. As AI becomes more involved in purchasing decisions, the gap between those two things will become commercially significant.

This is the area we believe deserves far more attention.

The future of search will not simply belong to businesses that publish the most content or appear on the most platforms. It will favour businesses capable of building a coherent body of evidence around who they are, what they do and why they deserve authority within their market.

That evidence needs to work for humans, search engines and AI simultaneously.

We are not presenting one Gemini conversation as definitive proof of how every AI recommendation system works. AI search is too dynamic for claims like that, and current research shows that generative search results can vary considerably between systems and even between similar queries.

What the conversation did provide was a very clear indication of where we should be looking next.

Macrocosm had already done what an SEO company is expected to do: compete organically and prove our capability in search. Yet when Gemini was initially asked to identify leading SEO companies, other forms of online authority influenced the answer more heavily. Only when the reasoning was challenged did the significance of that organic performance become part of the assessment.

That is not something we intend to ignore. It is something we intend to understand.

AI Optimisation is still developing, and much of the industry is currently defining terminology around systems that themselves continue to change. We believe the companies doing meaningful work in this space will be the ones testing those systems, questioning their conclusions, measuring what influences them and applying what they learn to real digital strategies.

That is exactly what we are doing.

At Macrocosm, The Art of Visibility has always been about more than simply appearing somewhere on a screen. It is about being understood, remembered and chosen.

AI has not changed that philosophy.

It has simply made the challenge much bigger.

Because the next stage of search will not only be about whether your business deserves to rank.

It will also be about whether AI understands why it does.