2026 AI Commerce
Readiness Index

Executive Summary

Only 2 in 100 ecommerce sites are ready for AI to add to cart and complete checkouts. The other 98 are not ready for the shift that is already underway.

Aidō Lighthouse™, the first AI commerce readiness platform, analyzed the state of ecommerce infrastructure of 345+ retailers across ten industry categories (beauty, fashion, electronics, food, travel, and more). Using our proprietary D/U/T Framework: Discoverability, Understandability, and Transactability, we're sharing a benchmark report on the state of ecommerce readiness for AI.

AI is no longer just helping consumers find products, it is beginning to compare options, add to cart, and initiate transactions on their behalf. It is a question of "when," not "if" browser agents, LLM-powered shopping assistants, and voice-driven purchasing will drive the next wave of ecommerce.

For many retailers, the question "are we ready for AI?" hinges on their visibility in generative chat platforms like ChatGPT and Claude. This is only part of the solution. Discovery is not enough. Understanding is not enough. The agent must be able to transact.

Methodology

How We Measure AI Readiness

Every score in this report was generated by the Aidō Lighthouse AI Commerce Readiness platform. Each scan runs 110+ checks across discoverability, understandability, and transactability. No scores were estimated or inferred; all scans were conducted directly against live sites.

This report draws on 345+ scans conducted through the Aidō Lighthouse platform across ten industry categories. The sample spans global retailers, direct-to-consumer brands and marketplace platforms across North America, Europe, and Latin America.

Finding 01

First-mover advantage in agentic commerce is almost entirely unclaimed.

Across 345+ scans, the average overall readiness score is 48.1 out of 100. Only 2% of sites score 80 or above — the threshold for AI-Ready status. 34% fall into the Critical tier, with scores below 40. The remaining 64% are in the Developing band: partially ready, but not yet capable of supporting autonomous agent transactions.

Finding 02

All three dimensions fall short — and each has a different path to improvement.

All three dimensions score below the AI-Ready threshold. Each falls short for different reasons and each requires a different kind of fix.

Discoverability · avg 48.9 — Structural access is still a problem. Nearly half of sites score below 50 on discoverability. Many actively block AI crawlers, lack structured sitemaps, or have no product feeds. This is not a solved problem, it is simply a more familiar one, with established fixes: robots.txt configuration, XML sitemap generation, and product feed publication.

Understandability · avg 63.8 — The strongest pillar, but not a passing grade. Understandability is the highest-scoring dimension, a reflection of years of SEO investment in structured data and semantic markup. But 63.8 is not a passing grade. Many sites have partial Schema.org implementation: Product markup without Offer or Review data, or JSON-LD that covers categories but not individual product pages. AI agents operating on incomplete structured data will make errors.

Transactability · avg 28.8 — A commerce infrastructure problem, not a content one. Transactability is the lowest score because it is a fundamentally different kind of gap. Discoverability and understandability can be improved through content and configuration changes. Transactability requires infrastructure: cart APIs, checkout pathways, tokenised payment support, and authentication designed for machine access. These were never built for agents, because agents were never expected to check out. 30% of commerce stacks score zero. There is no programmatic path for an agent to act on.

Aidō's Live Payment Test makes this tangible: watch a real agent attempt discovery, add to cart, and checkout on your site in real time. For most retailers, it is the first time they have seen their commerce infrastructure from an agent view.

Discovery–Transaction Gap: Notable Examples by Sector

The starkest illustration of the three-pillar gap: sites that score near-perfectly on discoverability — fully visible to AI — but return near-zero on transactability. Being found is not the same as being buyable.

Sites identified by sector only. Source: Aidō Lighthouse scan data, March 2026.

Finding 03

1 in 3 retailers have no transaction infrastructure for AI.

The transaction gap is not evenly distributed. At the extreme end, 29% of sites — nearly 1 in 3 — score zero on transactability. These retailers are fully invisible to any agent attempting to act, regardless of how well they score on discovery or understandability.

For AI agents, a low transactability score is not a signal to try harder but is a dead end. There is no fallback. The agent moves to the next result.

Critically, many of the retailers with zero transactability score well on discoverability and understandability. Their products are findable. Their content is structured. But when an agent attempts to act, the stack does not support it. Visibility without transactability is a missed transaction.

Finding 04

Your industry sets the average. It does not set your ceiling.

Across ten industry categories, the average overall score ranges from 58 (Beauty & Cosmetics) to 33 (Travel). However, within every category, there are sites performing significantly above and below the industry mean. The ceiling is higher than the average suggests. And the floor is lower.

Score Distribution by Industry

The spread within each category is as revealing as the averages. Beauty & Cosmetics has the broadest distribution with sites at both extremes. Electronics has a tight cluster in the mid-range. Fashion is heavily skewed toward Critical. Understanding your category distribution is as important as knowing your own score.

Score tiers: Excellent ≥80 · Good 60–79 · Fair 40–59 · Poor <40 · Source: Aidō Lighthouse platform, March 2026

Beauty & Cosmetics achieves the highest average overall score at 58, with the best single-site scores in the dataset. The category's strength reflects years of investment in product content, rich media, and structured data, originally driven by SEO and social commerce. That investment translates into AI readiness.

Jewelry & Accessories ranks second overall (55) with the highest transactability score of any category (55). This reflects the category's early investment in product configurators, ring-size selection tools, and custom order APIs, all of which map well to agent-compatible transaction flows.

Food & Beverage shows the highest discoverability score of any category at 83 — driven by strong sitemap and feed implementation across grocery and delivery platforms. But transactability sits at 7, reflecting the complexity of delivery logistics, inventory volatility, and the absence of agent-compatible checkout flows.

Consumer Electronics & Tech sits fifth at 48 overall, closer to the middle of the pack than the category's digital maturity might suggest. The ceiling is high: some electronics retailers score in the 90s. But the floor is low, with a long tail of specialist retailers that score below 20.

Travel scores lowest overall at 33, with the worst transactability in the dataset at 6. The category's commerce model (multi-step booking flows, dynamic pricing, and identity verification) is structurally incompatible with current agent transaction models. This is a product gap, not a content one.

Finding 05

Scale and brand recognition offer no protection.

One of the most striking patterns in the data is the absence of a correlation between brand size and AI readiness score. Globally recognised retailers, including household names with billions in annual revenue appear in the Critical tier. Smaller, digitally-native brands appear in the AI-Ready tier. Brand scale is not a proxy for infrastructure readiness.

Large retailers have invested heavily in digital experience, but that investment has been directed at human-facing interfaces: mobile apps, personalised recommendations, visual search. The infrastructure required for agent-compatible commerce (open APIs, machine-navigable checkout, tokenised payments) was not part of that investment roadmap, because agents were not yet the customer.

Illustrative examples, identified by sector only. Source: Aidō Lighthouse scan data, March 2026.

Finding 06

Protocols unlock readiness but they do not create it.

The emergence of commerce protocols such as UCP, ACP, MCP, WebMCP is one of the most significant developments in agentic commerce infrastructure. These protocols define how AI agents interact with commerce systems: how they authenticate, how they query product data, how they initiate transactions. They are the language of agent commerce.

But protocols are a layer on top of infrastructure not a substitute for it. A protocol cannot create a cart API that does not exist. It cannot generate Schema.org markup that was never implemented. It cannot make a checkout flow machine-navigable if it was designed exclusively for human interaction.

The readiness work required to support a protocol (structured product data, accessible APIs, machine-navigable checkout) is precisely the same work measured by the D/U/T Framework. Retailers who build that foundation will be well-positioned to benefit from protocols as they mature and proliferate. The two are complementary: protocols raise the ceiling; readiness determines whether you can reach it.

What Comes Next

The window is narrowing.

Agentic commerce is not a hypothetical. Browser agents, voice-driven purchasing, and LLM-powered shopping assistants are already live.

What this benchmark reveals is that the ecommerce infrastructure required to support autonomous agent transactions has not been built. The retailers who move first will not just be AI-ready. They will be ahead of a shift that is already underway.

The retailers who capture AI-generated commerce will not necessarily be the ones with the biggest brands. They will be the ones with the most agent-compatible infrastructure.