Train The AI To Recommend You – with Andy Crestodina

We talk with Andy Crestodina about why AI is turning search into a recommendation engine and forcing marketers to rethink what “visibility” even means. We share how to train the internet to trust your brand by combining SEO with conversion rate optimization, real buyer insight, and bottom-line accountability.

• AI search shifting from rankings to recommendations
• Visitor psychology and empathy as the missing SEO skill
• Multi-parameter prompts revealing the buyer’s real story
• Websites as controllable training data for AI citations
• AI traffic converting higher because consideration happens earlier
• Traffic volume becoming a vanity metric as AI filters the unqualified
• Answering real sales questions beyond generic FAQ blocks
• Thought leadership and original research as durable content moats
• Moving from one-off prompts to reusable prompts, assistants, and tools
• Revenue accountable CMO skills tied to CRM and business outcomes
• Working backwards from the end of the buyer journey to prioritise impact

Looking for the structured conversation and key takeaways for CMOs and AI marketing leaders? Read the cleaned and structured reference version here:
How to Get AI to Recommend Your Brand: Andy Crestodina on AI Search, SEO, Content and the Future CMO

Stop Chasing Traffic And Start Chasing Qualified Leads

Generative AI is changing search behavior faster than most marketing dashboards can keep up with. The shift is not only about ranking in Google or getting “visibility” in a chatbot result. AI now sits on top of the internet as a recommendation layer, helping buyers move from problem aware to solution aware and even brand aware before they ever click. That means AI search optimization is less about chasing keywords and more about earning trust in a multi-parameter prompt where the user’s real story shows up: constraints, preferences, context, timing, and risk. If you do not understand the visitor, you cannot predict the prompts that shape your shortlist.

A practical way to think about “being recommended” is to treat your website as training data for the internet. Your brand site is the only place you fully control the message, depth, proof, and clarity. Research shared in the conversation highlights that a large share of AI citations can come directly from brand websites, and that AI agents often land on homepages far more than any other page. If the homepage and key service pages do not answer basic sales questions, you are effectively training the model to hesitate. Modern SEO needs conversion rate optimization and visitor psychology built into the content architecture, not bolted on later.

AI referred visitors often convert better because the consideration stage increasingly happens inside the model. Instead of bouncing between search results and ten tabs, buyers ask follow-up questions in the conversation, refine requirements, and arrive “warmer” because they already feel partly qualified. This is why traffic volume becomes a weaker success metric. For years, web analytics was inflated by unqualified visits from people who needed to browse to self-qualify. AI filtering removes that noise. The KPI focus shifts toward qualified leads, pipeline quality, click-through rate on calls to action, and the CRM outcomes that prove revenue impact.

To avoid being silently excluded, brands need to publish the information buyers actually ask for, even when it is not a keyword. A common mistake is treating an FAQ block as a checklist item. The value is not the accordion UI, it is the substance: real pre-sale objections, requirements, and decision criteria gathered from sales calls, demos, customer support logs, and win-loss notes. If prospects repeatedly ask “Will this work in my situation?” or “What will break?” or “What is required on our side?”, those answers belong on the page in plain language with evidence. That is how you build recommendability and reduce friction at the moment of action.

When AI makes average content effortless, differentiation comes from what AI cannot genuinely produce: thought leadership and original research. Thought leadership requires a point of view, a line in the sand, and the courage to be specific. Original research requires access to proprietary data, careful methodology, and real analysis, whether that is GA4 patterns, conversion data, surveys, or benchmark studies. Marketers should also evolve from ad hoc prompt engineering to reusable prompts, custom assistants, and even lightweight tools built by subject matter experts. The most valuable leaders connect all of this to business outcomes, prioritize high-leverage improvements near the end of the buyer journey, and keep building a brand people seek out directly.