Competing for the Shortlist: How AI Is Rewriting Customer Choice

For twenty years, businesses competing online pursued one objective: be found. Search was the front door to the internet, and companies invested billions to appear at the moment a customer expressed intent. Earn the click, win the visit, make the sale. 

AI is rewriting that sequence. Not by killing search; the evidence does not support that conclusion. It is inserting itself between a customer’s expression of intent and the options they ultimately consider. The competitive question is shifting from “will customers find us?” to “will machines put us on the shortlist?” 

I should declare an interest, because it is also my qualification. I have spent my career in the industries that build consideration sets: price comparison, product discovery, the businesses that turn a vague intent into three good options. What AI is now automating is not an activity adjacent to my sector: it is my sector’s product. So, what follows is partly what two decades of building shortlists teach about the next two decades of getting onto them. 

The evidence, without the hype 

Adoption is real and accelerating: Pew Research, in a survey fielded in February 2026, found that 49% of US adults have used an AI chatbot and 44% have used ChatGPT. The more consequential figure sits underneath, as 60% of US adults say they have read AI-generated summaries in search results.  

Exposure to AI-mediated answers therefore extends well beyond deliberate chatbot use. And when a summary appears, behaviour changes, with Pew’s analysis of nearly 69,000 actual Google searches finding users clicked a conventional result on 15% of visits when no AI summary was shown, and 8% when one was. 

Scale deserves equal discipline: across a sample of 74,752 websites, Ahrefs measured AI assistants at roughly 0.3% of total traffic, against about 28% from Google. Adobe recorded a 693% year-on-year jump in AI-driven traffic to US retail sites over the 2025 holiday season, and 393% in the first quarter of 2026, which is extraordinary growth from a small base.  

But the number that should hold a board’s attention is neither of those. As of March 2026, Adobe found visitors arriving from AI sources converting 42% better than traditional channels, having converted 38% worse only a year earlier. That 80-point swing in relative conversion performance over twelve months is not merely a traffic story. It suggests that AI-referred visitors are arriving further through the decision process. 

The rational posture is neither dismissal nor wholesale reorganisation. It is to prepare for a channel whose strategic importance is growing much faster than its volume. 

Being found is no longer the finish line 

The framework I would offer any leadership team has five stages: found, understood, compared, trusted, chosen. Each can be audited. 

Found: can machines reach you at all? Many businesses spent years blocking automated traffic and are now blocking the automation that carries customers. An explicit agent-access policy is becoming as basic as a sitemap. 

Understood: is your commercial reality machine-legible? Products, prices, availability, locations and policies need to exist in structured form, through schema markup and product feeds, not only in prose written for humans. ChatGPT’s shopping systems weigh price, availability, reviews and product characteristics, and let merchants supply feeds precisely so that this information stays current. 

Compared: can a machine establish how you differ? An advantage that is obvious to a salesperson, or buried in brochure copy, never enters an evaluation conducted in milliseconds. Attribute completeness and freshness decide whether your differentiation exists at all. 

Trusted: machines triangulate. Merchant claims are checked against reviews, third-party data and prior behaviour. Inconsistency between sources reads as risk, and risk is quietly excluded. 

Chosen: the final stage, and the fundamental change. Search presented options and sent the customer elsewhere to decide. AI participates in the decision. Increasingly, the consideration set is built before anyone reaches your website, if they reach it at all. 

What the shortlist business already learned 

The comparison industry internalised three lessons that the wider economy is about to learn. Machines can read adjectives, but they compare attributes; structured data therefore matters more than persuasive copy when a shortlist is assembled. Freshness is trust, because one stale price disqualifies a merchant faster than any competitor can. And the intermediary that adds structure has a defensible role, while the one that adds only traffic is increasingly exposed. 

That last lesson matters most now. Adobe’s visibility research found AI traffic surging while many retail sites still had significant machine-readability gaps, particularly on individual product pages. Someone has to become the machine-readable layer of every market. Data quality has stopped being an IT hygiene topic. It is now a marketing discipline. 

Selection, not yet transaction 

One boundary deserves respect. In a 2026 Gartner survey of US consumers, 31% were willing to let AI narrow their choices for household supplies, while only 11% were willing to let it make the purchase decision. The evidence so far points to selection, not autonomous transaction, as the near-term commercial battleground. That makes intuitive sense: choosing a holiday, a laptop or an insurance policy involves trade-offs people still want to resolve themselves. 

Yet even selection creates a new problem. Businesses spent years asking “is this a bot?” and blocking accordingly. In an agentic world, some bots represent customers: Visa is developing its Trusted Agent Protocol precisely to help merchants distinguish authorised shopping agents from malicious bots and other automated traffic.  

The useful question is becoming “whose bot is this, and should I trust it?” That is where discovery converges with digital identity and trust. 

Businesses spent the last twenty years learning how to be found by algorithms. 

They will spend the next twenty learning how to be chosen by them.