The Cross-Border MSME: Visible to Platforms, Invisible Elsewhere

Ninety-Seven Percent of Exporters, One-Third of the Value

Start with the number that frames everything else in this study. Small businesses are 97% of all identified US goods exporters by count, according to the SBA Office of Advocacy's analysis of Census Bureau data, but they account for only about 35% of total US export value, $541.6 billion of $1,554 billion in 2021. That is not a rounding gap. It means the typical American exporter is small, but the typical export dollar is not moving through a small business at all.

The SBA brief is careful to note the harder truth sitting underneath that split: most small businesses do not export at all. The 97% figure describes the population of businesses that already export, not small business as a whole. So the real question is not why small exporters capture a smaller share of value than large ones, that is expected given scale, but why so few small businesses cross into exporting in the first place, and whether that barrier is capability, capacity, or simply not being found by a buyer on the other side of a border.

The typical American exporter is small. The typical export dollar is not.

The Website Test Most Exporters Still Fail

One answer sits in a joint survey by the Export-Import Bank of the United States and the National Small Business Association, cited in the same SBA brief: 54% of small exporters surveyed do not have a website capable of processing an international order, and 69% do not have a site available in a foreign language at all. This is a smaller, older survey, and we tier it emerging rather than established for that reason, but it points at something concrete and fixable: a seller can be findable and still be un-buyable, because the checkout, the currency, or the language is not there when a foreign buyer arrives.

A separate, older US Chamber of Commerce and Google survey, fielded in 2019 and cited secondarily in the same SBA brief, found small exporters were far more digitally equipped than non-exporters, 92% of exporters reported using digital tools generally, against a majority of non-exporters describing themselves as unfamiliar with digital tools. Among the digitally-equipped exporters, 31% sold through ecommerce marketplaces, 43% used online payment processing, 41% used online marketing or advertising, and 29% used online translation tools. We are tiering this survey contested rather than established, not because the finding is implausible, but because it is now seven years old in a category that moves fast, and it survives here only as texture on the digital-readiness gap, not as a current baseline. The same survey found 61% of small businesses believed greater digital tool adoption could help them overcome cross-border trade barriers, the largest named barriers being international finance and payment issues (38%), tariffs and customs (33%), logistics and distribution (32%), and unclear regulations (31%). Those are not visibility problems. They are the operational floor a seller has to clear before visibility work pays off at all.

Where the Money Already Moved: Marketplaces

While a majority of small exporters were still missing basic web infrastructure, global ecommerce demand consolidated onto a small number of large platforms. ECDB market intelligence, reported by WorldEF, put marketplaces at an estimated 83.4% of global ecommerce gross merchandise value in 2025, up from about 81% in 2023, leaving owned, first-party stores at roughly 16.6% of the total. We tier this contested because it is a single vendor's market-intelligence estimate rather than an official statistic, but the direction, share continuing to consolidate toward marketplaces, lines up with everything else in the evidence.

Amazon's own Small Business Empowerment Report claims independent US sellers on its marketplace exported more than 580 million items globally in 2025, averaging more than $375,000 in annual sales per seller. We tier this contested, not because it is necessarily wrong, but because it is a self-published vendor report about the vendor's own platform, a direct conflict of interest, and its full figures could not be independently re-verified in this pass. A February 2026 independent US survey from the Small Business & Entrepreneurship Council found 30% of small businesses sell on Amazon and 74% agreed that digital platforms make it easier to compete with larger firms, a smaller, more conservative, and independently fielded figure that we tier emerging. Read together, these numbers describe the same shape from three different angles: marketplaces are not a side channel for a cross-border MSME anymore, they are close to the default distribution surface, whatever their exact share turns out to be.

Buyers Have Already Gone Self-Directed

The buyer side of this story is moving in parallel. Gartner's May 2026 press release found 69% of B2B buyers still turn to a sales representative to validate a synthesized insight, which only makes sense set against the broader trend Gartner has reported across several 2025 and 2026 releases, that a rising majority of B2B buyers prefer a completely digital, rep-free purchase path. We are tiering the full trend contested, because only the most recent of Gartner's three cited releases could be independently retrieved and confirmed as methodologically comparable to the earlier two, but the direction is consistent with the OECD's finding that AI application use among platform-connected SMEs in ten OECD countries rose to 39% in 2025 from 26% in 2024, generative AI use specifically rising to 26% from 18%.

That OECD figure carries its own caveat worth stating plainly: the survey sample is drawn from SMEs that are already customers of major digital platforms including Amazon, Intuit, Kakao, Rakuten and Sage, so it likely overstates the digital maturity of the broader, less platform-engaged small business population, a limitation the OECD's own researchers flag. Even the platforms' own top adoption barriers, maintenance costs (40%), lack of time for training (39%), and hardware costs (32%), are ordinary operating constraints, not technology skepticism. Buyers and sellers are both moving toward self-service and AI assistance at the same time, from opposite ends of the same transaction.

The Click That Isn't There Anymore

Pew Research Center's analysis of 68,879 real Google searches from a panel of 900 US adults, captured in March and April 2025, is the strongest single piece of evidence in this study for what a synthesized answer does to a seller's traffic. When a search produced an AI summary, appearing in about 18% of the searches studied, users clicked a traditional search result in only 8% of visits, against 15% without a summary present, roughly half the click-through rate. Direct clicks on the AI summary itself happened in just 1% of visits, and 26% of AI-summary sessions ended with no further click at all, compared with 16% of sessions without one.

Bain & Company's analysis of Sensor Tower panel data, covering more than 100 million daily on-device events, found a parallel pattern building inside ChatGPT specifically: shopping-related prompts nearly doubled as a share of all queries in six months, from 7.8% to 9.8% between January and June 2025, a 25% category gain layered on top of roughly 70% overall growth in ChatGPT prompt volume, and click-through rates on the outbound shopping links inside those answers rose from 2.2% to 5.7% over the same window. Adobe Analytics data from the 2025 holiday season, cited via Shopify, found traffic referred from AI sources to US retail sites grew 693% year over year, with AI-referred shoppers converting 31% more and bouncing 33% less than shoppers arriving from other sources. Read together, the picture is not simply decline. Raw click volume through classic search is shrinking wherever an answer gets generated instead of listed, but the clicks that do arrive through an AI answer appear to carry real intent. Both things are true at once, and a seller who only tracks total sessions will misread which one is happening to them.

The Question No One Has Answered Yet

Here is the limit of what this evidence can tell a small exporter. Every behavioral study of the AI-answer layer we found, Bain's analysis of Sensor Tower panel data, Adobe's holiday analytics, Pew's browsing panel, is a vendor or panel-based measurement, not an official census, and none of them break results out by seller size or seller country. No public research yet answers whether AI Overviews, ChatGPT, Perplexity, Gemini or Copilot surface a small or foreign seller at a rate anywhere close to how they surface an incumbent with existing brand recognition and a large indexed footprint. It is entirely possible the AI-answer layer compounds the same incumbent advantage that classic search already carries. It is also possible it does the opposite, and gives a well-documented small seller a cheaper way to get quoted than ranking ever was. Nobody has published the data to say which.

That is not a gap in our research effort, it is a gap in what exists to measure. Treat any confident claim that AI search 'favors' or 'punishes' small sellers, from us or from anyone else, as a guess dressed as a finding until an independent, seller-size-aware study says otherwise.

It is entirely possible the AI-answer layer compounds the incumbent's advantage. It is also possible it does the opposite. Nobody has published the data to say which.

Where Trust Still Breaks, Before Discovery Even Matters

Visibility only matters once a buyer trusts what they find. The 2019 US Chamber of Commerce and Google survey named the barriers small businesses actually feel when they try to sell across a border: international finance and payment issues (38%), tariffs and customs (33%), logistics and distribution (32%), and unclear regulations (31%), with 61% of respondents believing better digital tools could help clear those barriers. We could not locate current, methodologically sound cross-border payments research that ties payment friction empirically to lost buyer conversion at the discovery stage. One widely cited industry statistic on payment-speed expectations could not be independently re-verified this pass and was dropped rather than published on trust alone.

What the surviving evidence does support is the shape of the problem, not its exact size: a buyer deciding whether to trust an unfamiliar, foreign, small seller is weighing operational signals, can this seller actually take my payment, ship on a timeline I can plan around, handle customs, at the same moment they are weighing discovery signals, is this seller who they say they are. A seller who fixes only the visibility half of that equation without also closing the payments, logistics and regulatory gaps the Chamber survey names is optimizing the wrong bottleneck first.

Reading Your Own Terrain Instead of Guessing at It

Put the pieces together and a cross-border MSME today is contending with three surfaces moving at three different speeds: a marketplace layer that has already absorbed the large majority of global ecommerce demand, an owned-website layer that most small exporters have not yet brought up to a basic international standard, and an AI-answer layer that is demonstrably reshaping click behavior in ways nobody, us included, can yet say favor the small seller or the large one. Betting a whole strategy on any single one of those surfaces, ranking in classic search, winning a marketplace's algorithm, chasing an AI-answer citation, is a bet made without the data to back it.

This is exactly the gap the Visibility Corpus exists to close: a market-by-market, surface-by-surface map of where a given buyer population's attention actually sits today, read before deciding where to spend, not after. Search Surface Optimization is how a seller earns presence across that whole map rather than one piece of it, and a Machine-Readiness Score is how they track whether that presence is actually growing where it counts, marketplace, own site, or the answer an AI engine gives, for their specific vertical and target market. The exporters this evidence describes are not failing because they lack effort. They are flying without instruments across a terrain that changed shape twice in the last two years.