Showing Up in AI Search

A joint edition by Tobias Zwingmann and
Mike Stratta, Founder & CEO Arcalea

When people think about “AI impact on their business” they typically think about how they can use AI to boost efficiency, drive revenue, or develop new products.

But AI isn’t only something that you do.

It’s also something that happens to you.

One example is how customers find you online and whether they decide to consider you.

In the early days of online search, companies learned how much it mattered to appear high up on Google’s front page for the right keywords.

We believe AI creates another important moment for business discovery.

The goal of this article is to explain what to look for and give you a practical way to get started.

Let’s dive in.

People are increasingly asking ChatGPT, Gemini, Perplexity and other tools for things they previously might have typed into Google:

That raises two central questions:

You’ll hear terms such as answer engine optimization (AEO) and generative engine optimization (GEO). People use these labels for work intended to improve how businesses and information appear in AI-generated answers.

There is already plenty of tactical advice: add technical markup, publish FAQs, create comparison pages, participate on Reddit.

But starting with tactics is the wrong way around.

First, you have to actually understand what the AI is doing, what your buyers are asking and which sources actually matter in your category.

Let's unpack that.

Imagine asking a friend for a restaurant recommendation. They might answer from what they already know, look something up, or send you an article.

The analogy is imperfect, but it helps explain four things worth separating:

Those are related, but they are different outcomes.

Your website might be cited without your company being recommended. Your company might be recommended while someone else’s article is cited. Your business might be described accurately when asked about by name, yet be absent when someone asks which providers to consider.

Even appearing first in a list does not necessarily mean the assistant prefers you.

Testing conditions matter, too.

In a separate Arcalea study through OpenAI’s API rather than the consumer ChatGPT interface, Stanford had the highest observed MBA mention rate with search optional; Wharton led a later required-search run. Timing and request settings also differed, so the comparison does not isolate search as the cause. These were mentions, not measures of recommendation strength or school quality.

That is why “Where do we rank?” is too narrow a question.

A better one is:

“Are we being considered, for which buyers, and on what basis?”

Here are three steps to investigate.

Start with questions prospective customers actually ask. Sales conversations, customer interviews, and support requests are useful starting points.

For example:

Try those questions across at least two AI assistants relevant to your buyers.

Save the complete answers and record:

Also record the date, product or model where shown, search settings, and relevant location or personalization settings. Where the interface indicates that search was used, record that too.

Always use fresh conversations to reduce the influence of earlier messages, while recognizing that this does not remove every contextual difference.

Keep questions that name your business separate from questions that do not.

“Is Company X a good choice?” tests how AI describes a business the buyer already knows.

“Which companies should I consider?” tests whether it appears during discovery.

Those are two completely different business problems.

Neither question tells you everything. A business missing from the tested answers is not proof that the model knows nothing about it.

Repeat the questions and look for patterns. One answer is an observation, not a reliable measurement program.

And keep the business objective in view: more mentions are not, by themselves, proof of more qualified sales opportunities.

Some advice assumes there is a universal set of websites that matters for AI visibility: Reddit, YouTube, Wikipedia, and so on.

Our research suggests that investigating your own category is worthwhile.

In an August 2026 analysis, Arcalea examined 9,046 AI answers across four industries, recording 45,150 citations and domain references across 4,384 distinct web hosts. Approximately 98% of those hosts appeared in only one of the four industries tested. About 22% of the recorded references named a domain without linking to a specific page. Questions and testing conditions also matter when interpreting the findings.

Reddit illustrates that variation.

It ranked first among cited domains for business schools and tenth for commercial mechanical contracting in that analysis. That does not make Reddit irrelevant to contractors. It means a generic instruction to “invest heavily in Reddit” needs more investigation.

Look at the actual pages appearing around your buyers’ questions. What information do they contain? Is it relevant, accurate, and useful? Citations are clues, not a complete record of everything that influenced an answer. Also note that a link does not guarantee the source actually supports the claim beside it.

Use this evidence to identify opportunities rather than copy another company’s list of tactics.

Before creating more content, check that relevant search systems can access your important pages. A useful first step is to ask whoever manages your website to review crawling and indexing settings.

Compare what buyers ask with what your website actually explains.

Who is the product or service for? What is included? Where is it available? What does the buying process involve? What are the limitations or trade-offs?

“We offer industry-leading solutions tailored to your needs” gives a buyer little to work with. Instead, provide evidence for claims where you can: documented capabilities, clear product specifications, or case studies that explain their context.

There is no guaranteed technical shortcut. Google, for example, explicitly says its AI search features do not require special AI-specific markup beyond the normal Search foundations. That does not mean ordinary structured data is useless.

But don’t bet on it to boost your AI visibility.

Look at external sources appearing around relevant buyer questions: publications, associations, review platforms, communities, creators, and directories.

Then consider where you can make a legitimate contribution.

Maybe your industry association has an incomplete listing. Maybe an independent publication would value your subject-matter expertise. A public discussion may benefit from an informed answer.

Focus on accurate information and useful participation.

Fake reviews and promotional posts scattered across communities are not a sound foundation for visibility.

They’re spam hurting you in the long run.

Another useful place to start is making your business easy to identify and understand.

Your company name, services, locations, and important offer details should agree across relevant sources. Where differences are legitimate, explain them.

Imagine one page describing a course as entirely online while another says attendance is required. A prospective student needs to know which statement applies.

Get your own pages in order and request factual corrections elsewhere where you can. That sounds boring, but these are pracitcal improvements you can make once you‘ve confirmed the facts.

No one else will do it for you.

At this point, we can agree on a few anti-patterns.

❌ Don’t change your strategy because of one answer.

❌ Don’t assume another industry’s source relevance applies to you.

❌ Don’t publish volumes of thin pages or manufactured praise.

❌ Don’t treat citations, mentions, and recommendations as the same outcome.

❌ Don’t turn every observed gap into an urgent project.

Prioritize gaps that recur, affect an important buying decision, and are within your ability to address.

You don't need an enterprise AEO platform to get started.

Try this instead:

This measures performance on your test questions, not your share of all customer searches.

A different answer after an edit does not prove the edit caused the change. Look for sustained patterns, and track relevant inquiries and business outcomes alongside visibility.

The exercise can still tell you where to investigate first.

For a more detailed walkthrough, Arcalea’s six questions an AEO diagnostic should answer can help you assess the answers and prioritize what to improve.

We like the comparison with the early days of online search as a reminder that discovery habits can change.

That is not to say that AI will follow the same path, or that someone has found a universal formula for being recommended.

The opportunity right now is to understand how your buyers use these tools and learn where your business is accurately represented, overlooked, or misunderstood.

That creates room to learn and observe faster than others.

Start with one buying decision. Examine the answers. Improve the information buyers need.

Then repeat.

See you next Saturday,
Mike & Tobias