Your next competitor for a sale may not be another brand. It may be your customer's AI.
Thirty commerce and technology leaders spent a morning at Batch LDN on one question: what does it take to compete when the buyer is becoming a machine? This page is what they concluded.
For a decade, online growth meant buying more of the same: more traffic, more campaigns, more discounting. That machine is flattening. A new intermediary now sits between you and your customer, an AI that finds products, evaluates them and increasingly buys, often without the customer ever visiting your site.
Commerce & Tech Leaders
Morning, One Question
Speakers, On the Record
London · 11 Feb 2026
A buyer's agent can find your products, compare them across the market and complete the purchase, often without a human ever loading your storefront. Most commerce dashboards cannot see it happening, because agent traffic does not behave like human traffic.
— Sartaj Rajpal, Leads AI Research, Profound
Buyers ask ChatGPT, Perplexity, Gemini and Google’s AI mode what to buy. Traditional SEO is a floor, not a strategy. Brands models do not cite risk becoming invisible to a growing share of demand.
As agents take on more of the decision, the signals they read, authentic reviews, verified product data, credible third-party discussion, become a brand’s most durable advantage.
The organisations moving fastest are not the ones with the largest budgets, but the ones with the culture to experiment, fail honestly and ship.
Leads AI Research, Profound
For three decades, being found online meant ranking on Google. That is fragmenting. Answer engines are becoming the first place people go for product recommendations, and a model decides who to cite in roughly three seconds: it fans one query into several, pulls in twenty to fifty candidate links, then makes a trust decision about which five to eight to cite. What disqualifies a page: content trapped in JavaScript, slow loads, mass-produced filler, cryptic URLs. What earns a citation: semantic URLs, answer-first content, fresh dates, hard data and genuine third-party discussion.
used an AI tool for shopping or product research in the previous seven days
Profound survey, ~2,300 people
Of AI citation outcomes explained by traditional SEO metrics
Profound, ~1,300 pages
annual AI-influenced purchase volume, on Rajpal's current estimate
Profound
On where citations come from, Rajpal was emphatic: Reddit is the single most-cited source across the major models, YouTube second, with roughly a fifth of citations from earned media. Volume and authenticity of recent discussion matter more than star ratings or view counts.
RAMP, a US fintech, had a strong Google presence but was effectively invisible to AI models in its accounts-payable category, ranking nineteenth with a 3.2% AI visibility score. Rather than flood models with content, the team published two comparison pages built for how models read: semantic URLs, structured data, real data on the page.
AI visibility in the accounts-payable category, within weeks
citations within the month (47 by week two)
visibility lift, passing eleven competitors, from two pages
Rajpal's projection: by 2030, AI agents could direct 30% of all consumer spending. He traced a fifteen-month arc from experiment to infrastructure, from Perplexity's one-click buy in late 2024, through ChatGPT shopping and agentic checkout in 2025, to Google's checkout mandates and answer engines serving ads by early 2026.
CEO, Trustpilot
Blair's argument was economic before it was technical: high-trust customers spend more, stay longer and recommend you, so the lifetime-value gap between high-trust and low-trust customers is, in his words, orders of magnitude. What has changed in the AI era is the mechanism. Brands can no longer broadcast credibility one-to-many; models assemble it from signals across the whole web, rewarding recent, high-volume, genuine review data.
people now using LLMs to find information, much of it with commercial intent
Adrian Blair, Trustpilot
fake reviews removed from Trustpilot in 2025, tied directly to authority
Adrian Blair, Trustpilot
most-cited domain globally across major LLMs, January 2026
Adrian Blair, Trustpilot
Founder, Nobody's Child & Fabacus
Xeni approached trust as verified, machine-readable data. Fabacus captures more than two hundred attributes per product, from composition and traceability to compliance, and his point was that this same verified data is what lets an agent identify and trust a product, and tell a legitimate one from a counterfeit. Without it, an agent has no reliable way to judge.
He also made a counter-intuitive move: anticipating that agentic commerce will raise the cost of digital acquisition, he signed for several new physical stores, citing a direct correlation between local physical presence and online sales in the same area. The omnichannel model, he argued, becomes more important in an agent era, not less.
Former head of innovation, Google
Biehler's argument was that the barriers to AI-era growth are usually cultural, not technical. When he asked the room why innovation stalls, the answers were consistent: fear of failure, business-as-usual ROI thresholds, and approval chains that take months when an experiment should take a week. He recalled Bill Gates asking a research team their failure rate; told it was 40%, the response was that this was not high enough. In genuinely innovative teams, a high failure rate signals appropriate risk-taking.
What he told leaders to actually do: start with purpose, not technology; communicate the vision constantly, in customer terms, not just financial ones; model psychological safety by celebrating the act of contributing before judging the answer; protect thinking time, because AI lets teams fill every hour with execution and remove the slack strategy needs; and know when to stop a project that is not the best use of resources, even one with visible champions.
Synthesising the morning, the room kept returning to the same four questions. This is the framework that emerged from The Exchange.
Be retrievable and citable by AI models across the funnel. Assess visibility for the queries that matter, close the gap between Google ranking and AI citability, and invest in semantic, answer-first content on the sources models actually cite.
Build a verifiable, machine-readable layer of trust. Map review volume, recency and distribution across cited platforms; treat structured product data as a commercial asset, not a compliance cost; keep claims consistent with the actual experience.
Build the human infrastructure to move at the pace competition now demands. Communicate a clear, non-financial purpose, create explicit safety for honest failure, protect strategic thinking time, and reward value created, not effort spent.
Prepare to transact with agents, not just people. Track AI-referred traffic separately, take a clear position on how your brand shows up in agent-mediated buying on price and availability, and plan for a near future where agents complete a meaningful share of transactions.
The window for differentiation is still open, but the baseline rises every month. The AI Readiness Assessment is a rapid, interview-based conversation that produces a prioritised view of where to focus across the four pillars, and what to fix first. It ends with a ranked roadmap you own.
A twelve-page briefing from The Exchange · London: a written summary
of everything the speakers said, with every figure attributed to the
person who presented it and to their source.
Ask Sol anything about digital products, AI, engineering, or growth, and get answers drawn from years of Net Solutions thinking and experience.