April 27, 2026

The conversion gap reversed. TechCrunch reports that AI-driven traffic to US retail sites grew 393% year-over-year in the first quarter of 2026, and visitors arriving from those sources are now worth more than the humans browsing alongside them. Adobe's Q1 analytics, drawn from over a trillion visits to US retail sites, found AI-referred shoppers converted 42% better than non-AI traffic in March 2026. One year earlier, the same population converted 38% worse. That is an 80-percentage-point swing in twelve months. Revenue per visit from AI referrals ran 37% above non-AI traffic, and AI-sourced visitors spent 48% more time on site while browsing 13% more pages. A consumer survey accompanying the analytics found 39% of respondents have used AI for online shopping and 85% said it improved the experience. The agentic shopper is not a worse customer arriving with worse intent. The agentic shopper is a customer who already decided what to buy.

Retailers are not yet ready for that customer. The same Adobe analysis flags an optimization gap that explains why so much AI traffic still lands on pages built for humans. About 25% of homepage and category page content is not optimized for large language models, and 34% of individual product pages are inaccessible to AI. Retailers who compound the conversion advantage will be the ones whose product detail pages, category taxonomies, and structured metadata can be parsed by an agent without ambiguity. The traffic kept coming. The infrastructure to monetize it consistently did not.

American Express moved to absorb some of the residual risk. PYMNTS and Fortune both report that AmEx debuted its Agentic Commerce Experiences Developer Kit on April 14 and announced industry-first Agent Purchase Protection covering eligible charges from registered AI agent errors. Luke Gebb, AmEx's EVP of global innovation, framed the calculation directly: "To date there have probably been as many press releases on agentic commerce as transactions, but no doubt it will happen." The company can make the liability commitment because it already runs dispute infrastructure for human chargebacks, and its developer kit issues tokenized agentic credentials only to verified agents that pass authentication before purchase. Visa, Mastercard, and Stripe have shipped competing infrastructure in recent months. Amex is the first card network to commit publicly that it will eat the cost when agents err, which is the question every consumer authorizing an agent eventually asks.

Google Cloud answered a different question. On April 22 at Cloud Next 2026, Google committed $750 million to its partner ecosystem to accelerate agentic AI deployment, the largest single partner investment any hyperscaler has announced. Accenture has already built more than 450 agents on Google Cloud. Deloitte called it the firm's largest commitment yet on any cloud AI platform, with 100 agents deployed. KPMG put in $100 million of its own capital. PwC committed $400 million for security and compliance agents. NTT DATA dedicated 5,000 engineers. BCG, Bain, and McKinsey received early Gemini access. Kevin Ichhpurani, president of Google Cloud's global partner ecosystem, framed the prize plainly. Agentic AI will create a roughly $1 trillion global market, and partners capture up to $7.05 in services revenue for every $1 spent on Google Cloud. The implication for retailers and B2B sellers is that the agents evaluating their catalogs over the next two years will increasingly be built by these consultancies on infrastructure they have a financial stake in standardizing.

Marketing teams have been operating partially blind to all of this. PPC Land reports that HUMAN Security expanded its Agentic Visibility capabilities on April 21 and integrated them natively into Adobe Experience Platform. The HUMAN 2026 State of AI Traffic report finds automation now growing eight times faster than human traffic. Gartner, cited in HUMAN's report, projects AI agents could drive 80% of internet traffic by 2035. OpenAI generates roughly one website visit per 198 crawls, against Google's one per six, which means agent-mediated retrieval is consuming retailer content at scale without delivering the human visits retailers measure. John Searby, HUMAN's Chief Strategy Officer, captured the gap. "Marketing and commerce teams are being asked to adapt to a fundamental shift in how customers discover and interact with brands, but they lack the tools to see what's actually happening." The Adobe integration is the practical fix. Marketers can now distinguish human, bot, and agent traffic from inside their existing analytics workflow, set policy on what gets allowed, blocked, or redirected, and optimize separately for human and agent surfaces.

Advertising is moving in parallel. The Current reports that Stagwell became the first global marketing network to adopt The Trade Desk's Koa Agents, an agentic AI capability for media planning, buying, optimization, and measurement, integrated through Open Agentic Kit. Marketers describe goals in natural language. Agents execute, optimize, and adapt campaigns in real time, replacing days of manual setup with continuous machine adjustment. Stagwell expects closed beta availability for select clients later this summer. The pattern matches what is happening on the buy side. Agents capture intent. Agents execute against constraints. The question for retailers is not whether their advertising will be agent-mediated. It is whether their catalogs, payment rails, and infrastructure can absorb the demand those agents redirect.

Four announcements lined up too cleanly to dismiss as coincidence. American Express committed to backing agent errors, removing one of the last consumer trust barriers. Google Cloud and its consulting partners committed roughly $1.25 billion in combined capital to enterprise agents. HUMAN Security gave marketers a way to see the AI traffic Adobe says is now their most valuable cohort. Adobe documented the conversion crossover that makes all of this commercially urgent. Each piece on its own is a press release. Together the rails, the capital, the measurement, and the behavior all moved in the same week. Retailers who finish 2026 with non-LLM-readable product pages will not have failed to predict the future. They will have failed to act on it after the data made it impossible to ignore.