Episode 34: How Brands Optimize for AI Shopping Agents

Key Takeaways

• Most websites are failing AI agent usability tests
• AI shopping agents prioritize page speed, clean UI, and functional forms
• Challenger brands can use the agentic web to leapfrog offline market leaders
• Brands should track Share of Voice across both traditional and AI surfaces
• Vanity metrics like impressions and sessions are being replaced

AI shopping agents are already browsing, comparing, and converting on behalf of consumers. But most websites aren’t ready for them; they’re too slow, too complex, and too focused on looking good rather than working well.

In this episode of the Local Marketing Beat podcast, host Christian Hustle sits down with Jes Scholz, Growth Marketing Consultant and SEO Futurist, to unpack what happened when she tested 100 AI shopping conversations and what brands need to do to win in an agent-first world.

‍

‍

Timestamps

00:00 Introduction and the rise of AI shopping agents

00:58 What happened when 100 AI shopping tests failed

02:24 Why pretty design is hurting agent conversions

04:07 Page speed, JavaScript, and the cost of rounded corners

05:26 New protocols: Web MCP and the agentic web

07:01 Will AI agents kill brand loyalty?

09:46 How challenger brands can leapfrog market leaders

12:31 AI SEO vs GEO: what to call it

14:12 Tracking momentum without traditional metrics

17:00 The restaurant that saw 40% higher tickets from AI-informed diners

19:13 Why vanity metrics are blinding digital marketers

21:31 The one priority for CMOs in the next 12 months

Most Websites Are Failing AI Agent Tests

Jes Scholz: “We saw very clearly we’re not very good at designing for humans even at this point, and if it’s not good at design for humans, then these agents have even more trouble. They couldn’t understand how to use the forms. They couldn’t load up the pages. It loaded too slowly.”

Jes ran 100 shopping conversations through ChatGPT’s agent mode, tasking it to browse, compare, and purchase products across well-known retail websites. The results were sobering: Most sites failed. Agents couldn’t navigate forms, pages loaded too slowly, and even market-leading brands were tripping up their own potential conversions.

The takeaway is blunt: If your site isn’t usable for a human, it’s not usable for an agent. Brands need to invest in functional UI and page speed over visual polish. As Jes puts it, the obsession with making websites “pretty” is actively costing brands agent-driven revenue.

Pretty Design vs Functional Design: the Real Cost

Jes Scholz: “Are those round corners that important that you’re going to add another second of load time, that you’re going to annoy AI agents, that you’re going to hurt your conversion rates? If you give executives that level of information, I would presume most will turn around and say “No, of course not.”

Jes highlights a common pattern: UI experts propose functional designs, but executive review cycles introduce cosmetic complexity that adds JavaScript, slows rendering, and breaks agent compatibility. The solution is bringing data to those conversations. If you can show that a design choice costs half a second of page load speed and reduces agent conversions by a measurable percentage, the decision becomes obvious.

New standards like Google’s proposed Web MCP protocol will help by letting sites communicate their capabilities to agents through structured APIs in the page headers. But adoption will be slow, just as schema adoption has been. The brands that move first on agent-readiness will have a significant competitive advantage.

AI Agents Will Reinforce Brand Leaders, Not Replace Them

Jes Scholz: “Unless you’ve told your agent through explicit prompts or through your previous purchasing behavior that you are extremely price conscious, it’s not going to follow those patterns. It’s going to follow the patterns of brand loyalty. What we’re looking at is a re-emphasis of market leaders.”

One of the most reassuring findings: AI agents are not going to destroy brand loyalty by chasing the cheapest price. Instead, agents follow existing market patterns. If a user is new to a category, the agent defaults to recommending the market leader based on reviews, citations, and digital presence.

For challenger brands, this creates an opportunity. Because agents can only see the digital reality, a brand that dominates online through citation targeting, strong review management, and consistent listings can position itself as the market leader in AI contexts, even if it isn’t the offline leader.

‍

Share of Voice: the Only Metric That Matters

Jes Scholz: “I would never look at AI search alone. This is not a ChatGPT question. This is a wider internet question and you need to be in a position where you’re monitoring multiple surfaces and understanding your Share of Voice across those entireties of the different corpuses.”

Jes advocates for Share of Voice as the single most important GEO metric, combining traditional search and AI search visibility. She warns against tools that only track ChatGPT visibility, since most AI agents will also run through Google’s Gemini, Chrome browser integrations, and other platforms. The key is understanding your share of voice across the full landscape.

This thinking maps directly to Uberall’s Location Performance Optimization (LPO) framework: Connect visibility, reputation, engagement, and conversion into one measurable system. If your Share of Voice goes up but market share doesn’t follow within a few months, something in the chain is broken.

Stop Reporting Vanity Metrics

Jes Scholz: “We as digital marketers have been blinded by metrics at the expense of forgetting KPIs. We’ve been communicating metrics to our executive teams that are just vanity metrics. Why do you care about impressions in Google Search Console? Why do you care about sessions in GA4?”

Jes makes a sharp distinction between metrics and KPIs. Impressions, sessions, and clicks are directional signals, not business outcomes. Many digital marketers have become lost in the data swamp and forgot that the point is revenue.

She recommends reporting on Share of Voice at the top of the funnel and market share (total sales and revenue) at the bottom. When those two correlate, with Share of Voice as a leading indicator, you know your strategy is working. For multi-location brands using GEO Studio, this means tracking prompt-level visibility and tying it to real business outcomes, not celebrating impression counts.

‍

What to Do Now: Fix Your AI Search Tracking

Asked what a CMO should prioritize in the next 12 months, Jes’s answer is precise: Fix your AI search tracking before you try to optimize anything.

Most brands have introduced bias into their AI tracking by pulling queries from Google Search Console (where they already rank) and measuring those in AI tools. The result is an inflated Share of voice that doesn’t correlate with actual market share. The fix is to start with unbiased grounding queries based on real category entry points and competitor analysis, then track across all meaningful AI platforms, including AI Mode, AI Overviews, and ChatGPT.

Only when you see a clear correlation between your AI Share of Voice and your market share do you have a reliable metric to optimize against. Until then, you’re flying blind.