Methodology

The panel is Vergio's continuously monitored set of Shopify stores: the ones every panel-population statistic on Vergio Data (everything except the universe metrics below) is measured across. A store counts toward the panel only while its monitoring status is active; a store that has churned or paused monitoring drops out of the panel immediately, though it remains part of the broader universe count.

The universe is a different, larger number from the panel: the cumulative count of every Shopify storefront domain Vergio has ever observed, tracked and vertical-classified regardless of whether Vergio actively monitors it today. It includes stores that may have since closed, and it only grows over time as Vergio's own discovery keeps finding new stores. The universe is never used to estimate how many Shopify stores are live right now; that question is out of scope for Vergio Data.

Every statistic published on Vergio Data traces to one of the metrics below, including the Vergio Agentic Commerce Index (VACI), the headline measure of how ready Shopify stores are for agentic commerce. Definitions are rendered directly from Vergio's own metric catalog, the single source every page and API response reads from.

Panel size(panel)

count of active monitored stores at window end

Panel composition(panel)

share by monitor tier / by vertical / by market (three one-level families; sub-30 buckets fold into "other" so each family sums to 100%)

Crawler block rate(panel)

% of definitive robots.txt reads with blocksAiShopping = true

llms.txt adoption(panel)

% of definitive reads with hasLlmsTxt = true

Product schema adoption(panel)

% of definitive reads with hasProductSchema = true

FAQ schema adoption(panel)

% of definitive reads with hasFaqSchema = true

Vergio Agentic Commerce Index(panel)

VACI v1.0 = mean(100 - crawler_block_rate, llms_txt_rate, product_schema_rate, faq_schema_rate), 1 decimal

Agent-Ready rate(panel)

% of stores definitive on ALL four facets passing: NOT blocking AND product schema AND (llms.txt OR FAQ schema)

Agent-Ready score distribution(panel)

share of stores at each component count 0..4; denominator = stores definitive on ALL four facets (same as agent_ready_rate)

App category adoption(panel)

% of definitive app_fingerprint reads with >= 1 detected app in the category

App market share(panel)

among stores with >= 1 detected app in category C, % using app A; top 8 per category + "Others" fold

Review platform share(panel)

among stores with a detected review platform, % per platform

Ad pixel adoption(panel)

% of definitive fingerprint reads with the network's PIXEL present; "any" = >= 1 of the four

Live ads rate(panel)

among stores with a definitive ad-library read this window, % with active_ad_count > 0

Stores that started advertising(panel_cohort)

constant-cohort stores whose ads-live SIGNAL first_observed_at falls in the month (NULL first_observed_at excluded)

Monthly store changes(panel_cohort)

constant-cohort stores with first_observed_at in month, one row per type

Market split(panel)

% by first-party market classifier (method mix stored on the row)

Tracked Shopify domains(universe)

count(*) of store_universe ("cumulative Shopify domains observed")

Vertical split(universe)

% by our classifier's vertical label (unknown -> "unclassified")

Country split(universe)

% by store_universe.country (registry baseline, dated; 12-month vendor-fallback rot rule)

Monthly store launches(universe)

store_candidates with discovery_kind = 'new_launch' and catalog_live_at in month ("new stores observed by Vergio"; by-vertical uses vertical_guess)

The Vergio Agentic Commerce Index (VACI) is the mean of four component rates, each measured across the same monitored panel: 100 minus the crawler block rate, plus the llms.txt adoption rate, plus the Product schema adoption rate, plus the FAQ schema adoption rate, all divided by four. The result is rounded down to 1 decimal. This is VACI version 1.0.

The crawler block rate checks whether a store's robots.txt blocks any of these AI shopping and answer crawlers at the site root:

• oai-searchbot
• perplexitybot
• claude-searchbot
• googlebot

Googlebot is included in this list because it also feeds Google's AI shopping surfaces, not only classic web search.

Vergio's app fingerprint reads for at least one installed app in each of these categories:

• reviews
• subscriptions
• upsell
• loyalty
• email and SMS
• helpdesk
• quiz
• page builder
• search
• A/B testing
• session replay
• popup

Vergio measures Shopify advertising activity two different ways. The ad pixel adoption rate is an on-site fingerprint sensor: it reads a store's page code for the tracking pixel of each network (meta, tiktok, pinterest, snap, plus an "any" row for at least one). The stores-that-started-advertising count is a separate ad-library signal sensor: it looks for a first-observed date on a distinct ads-live signal per network (meta, google, tiktok, bing, snap, plus an "any" row for at least one). The two sensors track different network lists and answer different questions: a pixel on a page shows the store can track ads for that network; an ads-live signal shows Vergio observed a live ad for that network in the ad library.

Ad-library reads are tier-skewed: not every monitoring tier reads the ad library on the same cadence, so the live ads rate's denominator is smaller than the full panel. The exact denominator is disclosed inline next to that figure on the advertising page.

store_candidates with discovery_kind = 'new_launch' and catalog_live_at in month ("new stores observed by Vergio"; by-vertical uses vertical_guess).

The by-vertical breakdown of monthly launches uses the vertical guessed at discovery time, not a store's later, more complete vertical classification, so an early guess that turns out wrong is never corrected retroactively in this split.

Vergio assigns every store a market (an ISO-3166 country code) using a strict order of evidence: first, the country ccTLD of its primary domain; if that gives no signal, the country implied by an unambiguous single-country currency it transacts in (a currency like the euro or US dollar that circulates across many countries gives no signal here); if that also gives no signal, the country on file in Vergio's vendor registry; and if none of those resolve, the market is recorded as unknown. Swiss francs fold to Switzerland and British pounds fold to the United Kingdom under this rule.

This month, 12% of the panel classified by domain, 5.8% by currency, 81.4% by vendor registry, and 0.6% remain unknown.

A country's published split suppresses once more than half of its stores are relying on a vendor-registry market that has not been refreshed in over 12 months.

A statistic is suppressed for a month when either its real sample falls below 30 qualifying stores, or more than 20% of the panel lacked a fresh, definitive read on the underlying facet. A suppressed figure shows the last valid value and the month it was last valid, instead of a fabricated current number. Month-over-month movement is separately suppressed when the constant cohort behind it has fewer than 300 stores, or when that cohort's mix of monitoring tiers diverges from the whole panel's mix by more than 15 percentage points.

A published month can be corrected, but only forward: a correction must carry a methodology version strictly higher than the one already published for that month, and the originally published row is kept in place, never deleted. Pages always render the highest methodology version on record for a given metric and dimension, and a corrected figure carries a note naming the correction date and the value it replaces.

This run excluded 0 constant-cohort advertising-signal rows and 2 constant-cohort store-change rows from their respective monthly counts, because those signal rows carried no usable first-observed date.

Vergio's AI-shopping-readiness capture only began in mid-2026, so months before that are never backfilled with an estimate. Where a metric is published for an earlier month, it is computed retrospectively: only over stores that were already being monitored on or before that month's end, never over the full current panel. This keeps every retrospective figure honest to what Vergio could actually see at the time, at the cost of a smaller retrospective sample than the current live month.

• v1.0 (July 16, 2026): Initial methodology.

Every statistic, chart, and definition on Vergio Data is released under a Creative Commons Attribution 4.0 license. Reuse, republish, or build on them freely, including commercially, as long as you credit Vergio and link back to the page you cited.

Every statistic on Vergio Data lives at a stable anchor on its page: hover a figure to reveal its # link, or copy the page's URL once you have scrolled to it. To cite a number, link to that page's URL followed by the stat's anchor (for example, vergio.ai/data/agentic-commerce#{anchor}), name Vergio Data as the source, and quote the value next to the month shown beside it.

For questions about any figure, definition, or methodology on Vergio Data, contact [email protected].