Garrett Sussman never started with keywords. He starts with cognitive biases.

As Director of Marketing at iPullRank, Garrett sits at the intersection of psychology and AI search. He studied psychology and English at university. He reads Kahneman. And he applies that thinking to how brands show up in AI search results.

This matters because the way someone phrases a query changes what they get back. Ask "is coffee good for me?" and you get different results than "is coffee bad for me?" Confirmation bias shapes the question. The model's editorial bias shapes the answer. And somewhere in the middle, your brand either shows up or it doesn't.

That framing is the foundation of what iPullRank calls relevance engineering: five pillars covering artificial intelligence, information retrieval, user experience, digital PR and content strategy. It's the framework behind their AI Search Manual, a 20-chapter resource written by six people, updated with four new vertical chapters in February 2026 covering e-commerce, local, YouTube and automation.

The Manual is also proof that this stuff works.

Garrett confirmed that iPullRank is generating real leads through ChatGPT. Prospects are filling out their contact form and saying they found the agency via ChatGPT after reading the Manual. Not through ads. Not through cold outreach. Through a piece of content that AI search decided was authoritative enough to cite.

That's the shift. Content doesn't just rank any more. It gets cited. And citations can drive revenue.

One of the biggest debates in the industry right now is whether AI search requires a fundamentally different approach or whether it's just solid SEO repackaged with a new name.

Garrett's view is nuanced. Some of it is semantics. Meta descriptions never mattered for rankings, only click-through rate. Now they can influence how your brand is described in AI answers. That's technically still SEO. But the paradigm has shifted.

Then there are the genuinely new things. Query fan-out means your content needs to satisfy not just one keyword but the 10 to 30 synthetic related searches that ChatGPT or Gemini generate behind the scenes. Product feeds for OpenAI's vendor marketplace require SEOs to think like ad ops. And 499 status codes, where your server is too slow for an AI crawler, mean page speed is no longer just a Core Web Vitals checkbox.

Garrett put it well: SEOs are the janitors of the web. They handle structured data, alt text, Open Graph tags, sitemaps and now AI search feeds. If the industry wants investment from the C-suite, it needs to reframe the conversation. Calling it GEO or relevance engineering isn't just branding. It's a strategic move to get a seat at the table where budgets are allocated.

The old reporting model of rankings, traffic and conversions doesn't hold up in AI search. Garrett laid out a three-tier framework that makes sense for boardroom conversations.

The key principle Garrett kept coming back to: accuracy over precision. You won't get exact attribution in AI search. The data is messy. Personalised results mean no two people see the same answer. But you can see trends, and trends are enough to make informed decisions.

Christian Ward from Yext said something similar on this podcast: you're not going to get all the data and that's going to be uncomfortable. Garrett agrees. The brands that accept that and move forward are the ones that will win.

Should brands be creating markdown versions of their content for AI crawlers? Garrett is cautious.

He pointed to research showing markdown doesn't help right now. But he also noted that Google has a history of pushing standards that benefit Google (HTTPS, AMP, Core Web Vitals) while positioning them as user experience improvements. A markdown directive, similar to llms.txt, could follow the same pattern.

His advice: experiment, but understand the risks. Don't go all in on something unproven. And don't ignore it either.

Google still holds around 90% of the organic search market. That number keeps coming up in conversations like this and it matters because it shapes strategy.

Garrett believes conversational search will eventually become the default. AI mode or whatever it evolves into will replace the ten blue links. But Google's rollout has been slower than expected because they're still figuring out how ads work in that format.

On Perplexity, Garrett was blunt: he thinks they'll get acquired. They had good PR but couldn't keep up. It's Google and everyone else, with OpenAI as the only real challenger.

On the DOJ antitrust case, he was equally direct. The focus on default search engine contracts was probably the easiest win for prosecutors but also the least consequential. Google's real moat is its ecosystem: Chrome, Workspace, Analytics, Ads, Gemini. End to end, across everything.

Garrett's shortlist for anyone leading AI search strategy:

Finally, Garrett gave a rundown of upcoming The SEO week conference. Four days in New York, 27 to 30 April, with 450 attendees and 40 speakers across themed tracks: the science of AI search, the psychology, the ecosystem (e-commerce, local, YouTube, enterprise) and the future.

If you're in the AI search space and can get to New York, it's worth the trip.

Watch the full conversation with Garrett Sussmann on the Franc Talking channel.

The best articles this week on AI Search and how it affects business, marketing and tech.