Discoverability vs. desirability: How retail brands can adapt to agentic commerce

Ecommerce companies should focus on structured data, the idiosyncrasies of their category, building long-term brand equity and determining what it is about their brand that makes them a destination in their own right.

(This article is an excerpt from CTI's Agentic Commerce Landscape Review)

Discoverability and desirability. Boston Consulting Group describes these two qualities as crucial in an agentic future. Another way of thinking about them is ‘marketing to machines’ vs. ‘marketing to humans’.

Discoverability (marketing to machines) is all about how well your brand or product shows up in AI search, social search, YouTube, or can be read, sold or recommended by an agent.

Desirability (marketing to humans), on the other hand, is all about the premium on your brand and how well trusted you are (in an environment where new tech and macroeconomics may affect consumer confidence).

These concepts are not new to marketers who are steeped in the ideas of outbound vs. inbound, of availability and mental availability, differentiation and distinctiveness, and mass marketing. But, arguably, the more the media landscape is fragmented, the more important the distinction becomes between these two drivers of discoverability and desirability.

This is a trend we highlighted (‘brand as the biggest moat’) in our 2026 trends report.

So, how machine readable is your website?

Google offers advice on building machine-friendly websites, and it has an experimental agentic browsing category within Lighthouse (its website audit/scoring tool).

There are a number of factors that can affect how well an agent can read or browse your website. Beyond applying to join the Universal Commerce Protocol (US only at time of writing) and ensuring your product data in Google Merchant Center is accurate, they include:

• Stability: reducing Cumulative Layout Shift (CLS), so things don’t jump around during loading and confuse an agent.
• llms.txt: an emerging convention that summarises website content.
• Accessibility: including accessibility tree and semantic HTML
• WebMCP integration: this experimental browser standard turns your website's UI elements into reliable APIs.

Product details pages (PDPs) in particular represent an interesting opportunity for retail brands who want to prepare for agentic commerce.

Ecommerce and retail media analyst Roger Dunn points out that PDPs fare relatively poorly in Adobe’s Citation Readability scoring, which measures how easily an LLM can parse your website. Whilst top performing brand landing pages score highly (79.1%) and category pages, too (70.6%), the top performing PDPs are only 63.5% readable.

Given that PDPs are often where the buying decision is made, and given also that AI traffic converts very well (with fewer returns), making these pages more machine readable is a good first step.

It’s still early days. And it’s still horses for courses

The big brands and retailers are obviously already thinking about their own AI assistants, and also how deeply to integrate with LLMs. And they are also thinking about those bigger positioning decisions in a tough economic climate – such as premiumisation and loyalty. Not to mention what their owned experience feels like, from customer support to fulfilment.

At this stage, it would seem prudent for brands, not just in retail, to prepare for agentic compatibility / machine readability – if not for actual sales via the likes of Google and Amazon, then for the possibility of a visibility uptick in search.

What we can say for certain is that agentic technology is already influencing the visibility of retail brands across the web. In response, ecommerce companies should focus on structured data, the idiosyncrasies of their category, building long-term brand equity and determining what it is about their brand that makes them a destination in their own right.