How to Get AI to Recommend Your Brand: Andy Crestodina on AI Search, SEO, Content and the Future CMO

Andy Crestodina, Co-Founder and CMO of Orbit Media, joins Joeri Billast on Web3 CMO Stories to explore how AI is changing search, buyer behavior, SEO, content marketing and B2B lead generation.

Andy argues that the next era of digital marketing is not simply about appearing in AI-generated answers. The real goal is to become a brand that AI systems such as ChatGPT, Claude, Gemini and Copilot can confidently recommend.

Drawing on Orbit Media research across 97 Google Analytics accounts and nearly 29 million website sessions, Andy explains why AI-referred visitors convert at unusually high rates, why traffic alone is becoming a weaker marketing KPI, how websites can become better training data for AI, and why original research and strong human opinions will become increasingly valuable.

The big shift: from AI visibility to AI recommendations

Joeri Billast: Andy, after roughly 25 years in digital marketing, which assumptions about visibility, content and conversion is AI forcing us to question most seriously?

Andy Crestodina: AI is a huge challenge to the SEO who, for 24 years, has been focused on appearing in search: visibility, ranking and maybe traffic.

AI changes that completely.

AI is becoming a layer on top of the internet that people use to get answers, recommendations and shortlists. It isn't just about visibility anymore.

The gap between what an SEO traditionally knows and what they need to understand next is the visitor.

If you don't understand the visitor, you will never properly understand their prompts, their moment of truth, the story in their life or the reason why they're in this buyer journey.

That's one of the problems with search optimizers claiming authority over AI search. Search optimizers aren't necessarily famous for conversion optimization, empathy, visitor psychology, personas or deeply knowing the buyer.

What makes AI recommend one brand instead of another?

Joeri: You argue that being visible in AI search isn't enough. Brands need to become recommendable. What makes an AI confidently recommend one company over another?

Andy: It is based on the user's query and how they prompt.

Imagine I'm taking a trip to northern Michigan. Instead of searching for something like:

northern Michigan vacation rental

I might tell an AI:

I've got two kids. I'm looking for a rental for one week. It needs to be near a city with a nice restaurant because it's my wife's birthday. We need some space. A beach would be nice. It doesn't have to be Lake Michigan. Maybe one of the smaller lakes.

That is a very different type of search.

The average traditional search query might be four words. An AI prompt might be 30, 40 or 50 words.

What's contained in those extra 35 words?

Context.

It's a multi-parameter search based on someone's actual situation.

The adaptation marketers need to make is relatively simple: apply conversion optimization thinking.

Answer real sales questions. Go deeper into the details. Cover things that aren't necessarily keywords. Understand your buyer.

Include evidence. Include proof. Understand what persuades people and what might prevent them from taking action.

My thesis is that AI search optimization is the overlap between search optimization and conversion rate optimization.

If those elements aren't in your content, AI will have less information available to confidently recommend your product or service.

SEOs aren't necessarily famous for knowing the buyer extremely well. Conversion optimizers obsess over visitor psychology.

The marketers who understand both sides will have an advantage.

How should websites change when the buyer journey happens inside AI?

Joeri: As more of the buyer journey happens inside an AI conversation, how should marketers rethink the role of their websites in building trust and closing the gap between recommendation and action?

Andy: There is definitely an overlap between classic search optimization and AI search optimization.

There are on-site signals. You can train AI by writing content and publishing information about your offers and your brand.

There are also off-site signals.

Are you in the directories?

Are you present in the places where people research your category?

Are you mentioned in the sources AI systems refer to when answering questions about your category?

That's similar to traditional search. There are on-page and off-site factors.

But your website is the place where the marketer has total control. You have unlimited time and space to go deeply into the details.

Something like half of AI citations come from brand websites.

We also have research showing that AI visits homepages around 15 times more often than other page types.

If your homepage doesn't answer basic sales questions, you haven't even really tried to train AI to recommend your brand.

And that's now the goal:

Train the AI to recommend your brand. That's the game.

Why do visitors from AI convert at higher rates?

Joeri: Your research found that visitors coming from AI sources are around three times more likely to become leads. What does that tell us about buyer intent and decision-making?

Andy: It confirms a hypothesis I've had.

Anyone making a buying decision typically moves through stages.

First they're problem-aware.

Then they're solution-aware.

Then they're brand-aware.

The consideration stage sits in the middle. That's where people explore options.

Increasingly, that thinking can happen inside the AI model.

Someone might say:

That place looks good, but I was looking for a rental with these other features. What about one with a pool?

They're talking through their decision with the AI.

If the consideration stage happens there, by the time they land on your website, they're much warmer.

They're ready to go.

They've already shortlisted you.

They're not bouncing repeatedly between Google search results and different websites.

The brand has already been partly qualified in their mind by the time they click.

That unusually high conversion rate from AI sources is another reason I believe AI search optimization is really an overlap between search and conversion.

Should marketers stop using website traffic as their main KPI?

Joeri: AI may send brands less traffic, but better-qualified traffic. I see it with clients too: fewer people might visit, but they stay longer or have stronger intent. Should marketers stop treating traffic growth as a primary measure of digital success?

Andy: One hundred percent.

Think about it this way.

In the pre-AI era, analytics contained huge numbers of unqualified visitors.

People who weren't a fit for your brand still had to visit the site, look around and figure out whether you were a good option.

AI can filter some of those people out earlier.

So the internet becomes more efficient.

Marketers can target audiences more specifically, and audiences can qualify brands before clicking.

Traffic may be lower now.

But another way to describe that is that traffic was artificially high before.

What information does AI need before it can shortlist your company?

Joeri: If AI is effectively shortlisting companies before buyers ever visit their websites, what information must brands make visible so they aren't silently excluded?

Andy: This is where I see a deficiency in the traditional SEO mindset.

Take FAQs.

The idea makes sense: answer people's questions.

That's why you have a page. That's why you have a visitor.

But sometimes SEOs think that simply adding an FAQ section completes the job.

Add it at the bottom of the page and you're done.

No.

FAQs need to answer real sales questions that people actually have before they're willing to take action.

Think about the vacation rental example.

One question I might have is:

Is the internet fast?

I may have to give a webinar during the trip. I won't book unless the internet connection is fast enough.

An SEO might completely miss that detail because they're looking at keyword tools, Google's People Also Ask results or competitor websites to determine which questions to answer.

Instead, marketers should listen directly to prospects.

A lot of businesses have recorded sales calls.

If prospects regularly ask a question during sales conversations, put the answer on your website.

Other prospects are probably wondering the same thing.

And without that information, AI may not have enough evidence to recommend your company when someone includes that requirement in a prompt.

That's why FAQs can be useful.

But there's nothing magical about the letters F-A-Q.

The important word is frequently.

What do people frequently ask?

Then answer those questions naturally in the flow of the page.

Build out the page and explain things properly.

That's how you win in AI search.

Does every webpage need an FAQ section for GEO?

Joeri: Someone once told me you need an FAQ on every page of your website. Is that exaggerated?

Andy: The goal isn't to have an FAQ on every page.

The goal is to make sure your pages answer the questions people genuinely ask.

It doesn't need to be a page block filled with expandable accordion sections containing irrelevant questions.

The job of the website is to train the internet to recommend your brand.

Answer the questions wherever and however it makes sense.

Why would you necessarily put an FAQ on an About page?

That page might be about your story, values, people and why you exist.

Answer those questions in a format that flows naturally for the human visitor.

There is no magic in adding an FAQ section.

But there is deep and real power in making sure your pages answer the actual questions people ask.

Can brands manipulate AI systems into recommending them?

Joeri: You've said marketers need to train AI to become a sales representative for their brands. Where is the line between influencing an accurate recommendation and manipulating the system?

Andy: If you're manipulating the system, you may attract visitors who aren't qualified.

Is that what you want?

You might drive traffic that never converts.

You might generate citations nobody clicks.

What good does that do?

What is the actual goal?

If you're a full-funnel marketer, you don't want unqualified leads in the pipeline.

You don't want traffic that doesn't convert on your sales pages.

Why would we try to trick a robot into recommending us to people who don't need what we do?

That doesn't make sense.

What kind of content will still stand out when AI can create unlimited content?

Joeri: AI makes average content almost effortless to produce. What will distinguish content that genuinely earns attention, trust and authority?

Andy: I've been thinking about this a lot because I'm a content marketer.

I believe there are two types of content AI cannot really create.

The first is thought leadership.

Most of the things we've discussed in this conversation are based on a point of view.

I'm taking a stand.

I'm drawing a line in the sand.

I'm saying this is right and that is wrong.

AI doesn't really do that.

It's an averager.

You could almost say AI stands for average information.

It can beautifully summarize the internet and try to be correct, but it doesn't have a worldview, perspective or lived experience.

What I've said here is based on my own views and roughly 25 years in digital marketing.

The second format AI can't independently create is new, original research.

You mentioned one of our studies.

To determine conversion rates from AI traffic sources, I had to extract data from 97 Google Analytics accounts.

AI helped me write some of the code, but I had to access the data and manually categorize conversion types as high or low intent.

It took almost 20 hours of research.

AI couldn't simply create that study because it didn't have access to those analytics accounts.

Businesses are sitting on huge amounts of proprietary data that they rarely use in their content programs.

Original research, surveys and internal business data can create genuinely differentiated content.

So the two formats I believe remain strongly differentiated are:

Strong opinions and original research.

They've always been effective, but AI makes them even more valuable.

Why podcasts, books and expert conversations become more valuable in an AI world

Joeri: That resonates with me because this podcast gives me conversations and strong opinions from people like you. That's also how I wrote my book, The Future CMO. I listened to many experts and tried to make sense of what they were saying.

Andy: That's human-based marketing.

These are relationships.

AI has zero friends.

These are people with personal brands. AI doesn't really have one.

Everything we're doing and discussing here includes things that are impossible for a model to genuinely experience.

You can't truly quote an AI as a person with lived experience.

Marketers keep asking:

How can I differentiate my content from AI?

If AI can completely write what you're publishing, maybe you should question whether that content needed to be published in the first place.

From ad hoc prompts to reusable AI assistants and tools

Joeri: You've also described an evolution from ad hoc prompts to reusable prompts and AI assistants. How should marketers determine which parts of their expertise to automate and which should remain deeply human?

Andy: The people who write the best prompts, make the best assistants and build the best AI tools are often subject-matter experts.

We should be using AI for the things we're already very good at.

If you're telling yourself:

I don't need AI for that because I'm already a professional in that topic or channel,

you may be missing opportunities.

You may actually be the best person to write a high-performing prompt, build an extremely useful assistant or create a powerful tool because you understand what good looks like.

The progression is something like this:

Ad hoc prompts → reusable tested prompts → AI assistants → dedicated AI tools

First, instead of repeatedly writing one-off prompts, create reusable prompts that have been tested and improved.

Then combine those prompts into assistants such as Custom GPTs, Claude Projects, Gemini Gems or Copilot agents.

Eventually, you can build actual tools that run in a browser, have a user experience, execute multi-step processes, save data or retrieve information.

People assume building those tools is difficult.

Often, they simply haven't tried yet.

Anyone who has experimented with tools such as OpenAI Codex or Claude Code can see how empowering this has become.

I'm not deeply technical, but I've built tools with tremendous utility.

My advantage is that some of those tools operate in areas where I have deep expertise.

What will separate great CMOs from marketers who simply manage AI tools?

Joeri: In The Future CMO, I explore how marketing leaders are becoming orchestrators across technology, data, creativity and trust. What capabilities will separate the most valuable CMOs from those who are simply managing more AI tools?

Andy: The difference between good and great CMOs will be bottom-line performance, accountability and strategy.

There are still many marketers who focus mostly on a channel:

We need to be more active on LinkedIn.

We need YouTube.

Or they focus on a format:

We need a podcast.

We need webinars.

I've gone through that era myself. I understand that level of thinking.

But the marketers who will perform best through this transformation are those who can connect all the dots.

The topic.

The channel.

The frequency.

The format.

The funnel.

And eventually, the impact on the business.

They focus on the CRM.

Are we getting qualified leads?

Which types of leads convert into deals?

Which deals create customers with the highest lifetime value?

Those questions aren't always central to how CMOs think today.

I believe the revenue-accountable CMO is the future of our category.

If you want to be a marketing leader, your goals need to align with business outcomes.

What should busy marketers prioritize first?

Joeri: A lot of marketers are extremely busy. How should resource-constrained marketers decide what deserves human attention when AI can accelerate so much work?

Andy: I have a strong point of view on this.

Performance-based marketers should work backwards from the end of the buyer journey.

Imagine you have two things you could do today.

You could make another LinkedIn video.

Or you could analyze the performance of your calls to action.

The LinkedIn video may create some brand awareness.

But analyzing your call-to-action click-through rate might reveal a gap at the exact point where visitors decide whether to become a lead.

If you can improve that conversion point, it may be enormously more valuable to your company than producing another social media post.

Ask yourself:

If I had only one hour or one day to do marketing this month, and my income depended entirely on generating qualified leads, what would I do?

That question forces prioritization.

You'd probably start near the end of the buyer journey and work backwards.

Contact page.

Calls to action.

Service pages.

Homepage.

Content creation might come much later.

AI can also help identify gaps on those high-value pages.

For conversion optimization, I'm particularly interested in developing and testing personas using sales-call transcripts.

Those transcripts contain real buyer language and are extremely useful for identifying unanswered questions and weaknesses on key pages.

What should marketers build now to remain visible, trusted and chosen?

Joeri: Let's look five years ahead. What should marketers and entrepreneurs start building today if they want to remain visible, trusted and chosen in an AI-mediated world?

Andy: First, continue building your brand.

A top-of-mind brand can circumvent digital channels entirely.

I recently shopped for a gift and went directly to a website I already knew. I started at the homepage and found what I wanted.

Strong brands do not become less powerful in the AI era.

If anything, brand recognition remains extremely valuable.

For everyone else, analyze where AI systems get information about your category.

I have a more sophisticated approach where I run synthetic prompts and analyze which sources AI systems use when answering questions.

Those are places where you may need to be present.

We should also prioritize formats AI cannot genuinely create:

• Webinars
• Videos
• Podcasts
• Live events
• Offline communities
• Conference presentations
• Original research
• Strong expert perspectives

There are many things AI cannot genuinely do because they're inherently human.

Digging deeper into the connection between marketing and sales will also make you more resilient.

Deeply understanding your buyer will make you safer.

Staying close to your CRM, data and business metrics will make you safer.

Keep training AI to recommend your brand.

But stay focused on the bottom line.

Build your brand.

And prioritize formats, experiences and ideas that AI could never independently create.

Where can marketers follow Andy Crestodina?

Joeri: If people want to follow your work or learn more about you, where should they go?

Andy: Businesses with underperforming lead-generation programs, websites, visibility or recommendations can find us at Orbit Media. We build and optimize high-performing B2B lead-generation websites.

Most of my current content is on LinkedIn.

If you find me there after listening to this episode, don't necessarily stop at the Follow button. Find the Connect option and mention that Joeri and this conversation connected us.

I also publish a LinkedIn newsletter. I've been a content marketer for a very long time, and I still love teaching, sharing and having conversations like this.

Key takeaways from Andy Crestodina

AI search optimization is no longer only about rankings or citations. The deeper opportunity is to give AI enough accurate, persuasive and specific information to recommend your brand to the right buyer.

1. Optimize for recommendations, not just visibility.
Appearing in an AI answer matters less than being confidently shortlisted for a buyer's specific situation.

2. AI search and conversion optimization are converging.
AI prompts contain rich buyer context. Brands that understand visitor psychology, objections, sales questions and decision criteria have an advantage.

3. Your website is training data for AI.
Your homepage and key commercial pages should clearly explain who you serve, what you do, how you do it, why buyers choose you and what evidence supports your claims.

4. Answer real sales questions.
Don't add FAQ blocks because someone told you they're good for GEO. Use sales conversations, CRM data and customer research to identify the questions people actually ask.

5. AI traffic may be small but valuable.
Orbit Media's research across 97 GA4 accounts and almost 29 million sessions found AI-referred visitors significantly more likely to convert into leads.

6. Stop worshipping traffic.
AI may remove unqualified visitors before they reach your website. Less traffic can coexist with better marketing performance.

7. Publish what AI cannot originate.
Strong perspectives, proprietary research, expert conversations, original data, live events and human relationships become more valuable as generic content becomes easier to generate.

8. Use AI where you already have expertise.
Experts are best positioned to build useful prompts, assistants and tools because they understand the quality standard.

9. Work backwards from revenue.
When resources are limited, optimize the moments closest to conversion before automatically producing more top-of-funnel content.

10. Build brand and trust for the long term.
Brands that people already know and actively seek out become less dependent on any individual discovery channel.

The central idea

The future of AI marketing is not about finding another GEO hack.

It is about becoming the clearest, most relevant and most trustworthy answer to a real buyer's question.

Or, as Andy Crestodina puts it:

Train the AI to recommend your brand.