Reading the SaaS–AI transition through web data

About this publication. This piece accompanies a webinar we produced for Eagle Alpha, the alternative-data community connecting investment firms with non-traditional data providers. It is a showcase of how our dataset (393M tracked hostnames, 3,000+ detected technologies, four years of point-in-time history) surfaces structural signals that traditional financial data misses. The numbers below come straight from our March 2026 crawl.

Four years of monthly crawl data have surfaced something we didn't expect to find: two separate technology populations are forming on the web, and they barely share a stack. One is the familiar SaaS application layer (Mailchimp, Salesforce Commerce Cloud, Heroku), whose web footprint has been contracting since 2022. The other is a fast-growing AI-native layer (Base44, Lovable, Replit) that is building without those tools, not migrating away from them.

That pattern predates the 2026 software sell-off. But the sell-off brought it into focus.

The web is growing faster than ever

Start with the headline number. Our March 2026 crawl counts 383 million unique domains. After filtering out parked domains, placeholder pages, and Host-not-found responses, 268.6 million remain. This is the working web, the part that actually serves content.
That count is accelerating. In a single month we detected 7.4 million newly activated sites against 4.6 million deactivated, a net gain of +2.8 million active sites. Whatever AI is doing to the software industry, it is not shrinking the web. The cost of creating a web presence has dropped, and more of them are being built.

AI adoption: broad but shallow

Of the 268.6 million active sites, about 2.65 million carry a detectable AI technology fingerprint: just under 1 percent of the filtered web. That number is a floor, not a ceiling. Our recipe coverage expands as new technologies become detectable, and a lot of AI usage hides behind backend API calls our crawler can't see.

But scale isn't depth. When we segment those 2.65 million sites by our Economic Footprint Index (EfI), the distribution skews sharply low. 91 percent score below 30: micro sites, test deploys, weekend builds. Only 8.7 percent show meaningful commercial activity, and just 5,300 sites (0.2 percent) reach enterprise-grade.

The AI layer of the web is broad but shallow. Its economic core is three orders of magnitude smaller than the headline count suggests.

The .ai TLD is the clearest startup-formation signal we have

The .ai extension, originally assigned to Anguilla, is now the fastest-growing tech TLD on the web. We detected 857,542 active .ai websites in March 2026, up 95 percent from May 2025. At current rates, .ai is on track to pass .io within months.

Demand isn't limited to registrations. ai.com sold for $70M in April 2025, the largest domain sale ever recorded, and 28 percent of Y Combinator and Techstars startups in 2025 chose .ai domains, up from under 3 percent in 2023. Anguilla's .ai registry revenue reached an estimated $85M in 2025, roughly 47 percent of the island's national budget.

The TLD is heavily concentrated in the US (72 percent) with Germany a distant second (9.8 percent). Vibe coding platforms (AI tools that generate websites from natural-language prompts) dominate the technology stack. Lovable alone accounts for 47 percent of all AI-detected .ai sites.

The most-detected AI fingerprint is not OpenAI

The most-detected AI technology in our database is Base44, with 1,062,608 sites: larger than Lovable, OpenAI, and Anthropic combined. Base44 is an AI app builder acquired by Wix in June 2025.

The count reflects domains where we detect Base44's fingerprint, not necessarily active businesses behind them. But the scale is telling. The no-code AI layer may be further along than headline coverage of ChatGPT and similar tools would suggest.

Two internets, near-zero overlap

Here is the cleanest pattern in the data. Lovable-built sites use Mailchimp on 0.2 percent of their domains. Sites that use Mailchimp today use Lovable on 0.2 percent of theirs.

The overlap in both directions is essentially zero. These are not companies mid-transition; these are sites built in different technology eras that have never encountered each other's stack. The AI-native cohort is not replacing traditional SaaS tools one by one. It is building without them. Across every tool we tested, AI-native builder sites use traditional email and CRM platforms at roughly a quarter to two-thirds of the rate we see on non-AI business sites.

Geography: the US leads, Germany follows

AI adoption is concentrating geographically. The US accounted for 34 percent of all AI-detected sites in July 2025; by March 2026 that share had grown to 39 percent (1.03M sites). Germany is the standout outside English-speaking markets, climbing from 18,000 AI sites in July 2025 to 218,000 by March 2026 (8.2 percent share). No other European market comes close. Brazil and Japan are emerging but trailing.

The data suggests a sequential pattern: the US leads, Germany follows, and the rest of Europe and major emerging markets sit further back on the curve.

Where SaaS is breaking, and where it isn't

Across 33 software products we track, four show material four-year declines in web footprint. All four sit in the application-layer SaaS cohort:

Not all application-layer SaaS is contracting. The pattern is consistent with a split along product type: point solutions that automate a single workflow are declining; infrastructure, document, and commerce platforms with deeper integration are holding or growing. Wix is up 28 percent over four years. Shopify is up 98 percent. Cloudflare +45 percent. DocuSign +110 percent.

Grouped together, the picture is clearer:

• Application Layer SaaS (Mailchimp, Heroku, SFCC, Freshworks): 1.70M sites in April 2022 → 1.18M in March 2026, −31%
• AI Native (Base44, Lovable, OpenAI, Anthropic, others): 2.65M sites, now 2.2× larger than Application Layer SaaS
• Transactional, infrastructure & commerce (Wix, Shopify, Cloudflare, SendGrid, DocuSign): roughly 15.7M sites, +26% to +110% over four years

The Salesforce divergence

The clearest single illustration of this split sits inside one company. Salesforce's core CRM product (Customer 360) has grown +133 percent over four years to 108,729 detectable sites. Commerce Cloud declined −57 percent over the same period, tracking alongside Mailchimp and Heroku.

The CRM data layer is growing; the application layer built on top of it is contracting. Within one company, the data moat is holding while the workflow tools are not. Oracle, SAP, and Microsoft sit in a similar category: companies whose proprietary data and infrastructure positions are difficult to replicate.

What this means

What our data shows is not that all software is declining. It shows that one specific category, application-layer point solutions that automate a single workflow, has been losing web footprint since 2022 or 2023. Products with deep enterprise integration, proprietary data, or infrastructure-level positioning continue to show stable or growing footprints.

At the same time, the internet is growing. Net +2.8M active sites per month. The sites being built on AI-native tools show near-zero overlap with the traditional SaaS stack. They are a distinct technology population, not a migration.

Five signals we're watching

1. The Salesforce CRM web footprint. Whether the +133% / −57% divergence continues, or spreads to the core product, is a key indicator.
2. .ai surpassing .io. At current growth rates, this happens within months. A direct signal of AI-native startup formation velocity.
3. Vibe-coding consolidation. Base44 (1.06M), Lovable (486K), Replit, Bolt, and Cursor are all competing for the same market. Which platforms retain users will shape the next generation of the web.
4. Geographic sequencing. Germany's jump suggests the adoption wave moves market by market. Western Europe and Brazil are next.
5. AI co-adoption inside SaaS. 4.4 percent of Mailchimp's remaining sites now show detectable AI. That number is the transition frontier; changes in its rate will tell us whether SaaS users are migrating or just diversifying.

The gap between reported earnings and actual technology adoption may be widening. Web intelligence is one of the few places that gap is visible in real time.

Source: Dataprovider.com, March 2026 crawl. All figures reflect unique domains after primary-hostname deduplication. Full methodology and definitions in the underlying report.