AI coding platforms are building the boring internet
Across a crawl of 20.7 million hosts, Village identified 87,003 live sites managed by Lovable, Replit, Base44, v0, and Bolt. Most of the visible ones are real businesses or working products, yet none rank among the web's 10,000 most-visited sites.
AI coding platforms have moved beyond prototypes, but their output has yet to achieve meaningful distribution. Across a crawl of 20.7 million hosts, we identified 87,003 live sites managed by five leading platforms: Lovable, Replit, Base44, v0, and Bolt. Among the most visible of these, 74.9% operate as real businesses or working products. None, however, rank among the web’s 10,000 most-visited sites, and only 682 reach the top one million.
At Village, we crawl the web’s most popular sites and detect the technologies they use, helping investment firms track software-vendor adoption at web scale. Our analysis of AI coding platforms shows that people are making the same things they have always made on the web: stores, schools, games, utilities, financial products, client portals, and specialized business software.
Findings
- Lovable dominates. Its 63,868 managed sites represent 73.4% of all sites detected across the five platforms. Replit follows at 9,575 sites, then Base44 at 8,553, v0 at 4,692, and Bolt at 315.
- The most visible sites are mostly working software. Across 1,250 high-visibility sites, we classified 398 as real businesses and 538 as working products. Just 42 were parked or broken, and 35 looked like demos or personal projects.
- Publishing is abundant, but distribution is scarce. No managed site we detected appeared in the CrUX top 10,000. Seven appeared in the top 50,000, 21 in the top 100,000, and 682 in the top 1 million. Publishing software is easy, but attracting a sustained audience is hard.
How sites rank
The five cohorts contain 87,003 managed sites. Lovable contributes 63,868 of them: about 6.7 times as many sites as Replit and more than the other four platforms combined.
To evaluate how much traffic these sites receive, we used Google’s CrUX (Chrome User Experience Report), which uses anonymous Chrome telemetry to provide a relative, navigation-based measure of site popularity. Lower rank bands indicate greater reach and traffic. Across this entire cohort, 62.3% of sites sit in the 10m–50m band, and only 0.8% reach the top 1m.
The cohort contains no top-10,000 sites. It contains seven sites in the 10,000–50,000 band, another 14 in the 50,000–100,000 band, and 186 in the 100,000–500,000 band. In total, just 207 of 87,003 sites make the top 500,000.
Lovable has both the largest footprint and the largest number of highly ranked sites: 480 of the 682 sites in the top 1m are Lovable deployments. But even Lovable’s distribution is overwhelmingly concentrated in the long tail. Building sites has become vastly easier, but attention remains highly concentrated.
What the most popular sites do
To go beyond deployment counts, we classified the 250 highest-visibility managed sites from each provider as measured by CrUX traffic. We had GPT-5.6-Sol inspect each live website directly and classify it.
The most vibe-coded sites are functioning products
Across the 1,250-site sample, 936 sites (74.9%) were classified as either a real business or a functional product. Real businesses alone account for 31.8%, whereas functional products account for another 43%. Lead-generation sites represent 9.4%, content or affiliate properties 8.1%, demos or personal projects 2.8%, and parked or broken sites 3.4%.
This weakens the popular claim that these platforms are used mainly for demonstrations: their most visible sites include many working products.
Bolt has the highest “real business” share at 47.6%, followed by v0 at 30.4% and Base44 at 29.6%. However, the Bolt sample covers 250 of only 315 detected sites, whereas the Lovable sample covers 250 of more than 63,000. These results describe each platform’s 250 most visible sites; they do not measure the quality of all its sites or estimate the average project on the platform.
People are building ordinary internet businesses
The largest categories across the equal-sized samples are e-commerce and marketplaces (10.6%), education (9.8%), gaming and gambling (9.1%), SaaS and business software (9.0%), media and content (8.2%), and consumer utilities (8.0%). AI tools account for just 4.3%.
The category mix is surprisingly broad, but there are visible differences by platform. Replit has the highest AI-tool share at 8.8% and is also strong in gaming and SaaS. Bolt skews toward education, e-commerce, healthcare, and professional services. Lovable has a broad consumer-facing mix spanning commerce, media, utilities, gaming, and finance. v0 leans toward gaming, education and SaaS.
Sites use similar services, but defaults differ
Once a site is live, its surrounding services and technologies offer a second clue to commercial maturity: analytics, conversion tracking, authentication, payments, reliability tooling, and customer engagement. We measured browser-based signals across all 87,003 managed platform-site records, combining evidence across detected hosts for each site.
The resulting technology map shows both a common commercial baseline and sharp platform-specific defaults.
Analytics are more common than payments and monitoring
The most common commercial layer is measurement. Google Analytics appears on 23–33.2% of sites depending on platform. Google conversion measurement appears on 16.5–21.5%, Meta Pixel on 12.4–20.1%, and Google Tag Manager on 10.9–16.8%. In general, many sites contain tools for measuring where visitors come from and whether they sign up or buy. Some of these tools may be platform defaults.
Visible monetization and operational tools are far less common. Stripe appears on 1.4% of v0 sites and 5.4% of Replit sites. Sentry-instrumented code appears on 1–1.9%, PostHog on 0.3–1.7%, and Datadog browser signals on 0–0.3%. These are lower bounds because our crawler may not reach checkout, authentication, or post-login routes. The typical managed deployment is much more likely to have basic marketing analytics than visible tools for payments, product analytics, or monitoring.
Each platform has a different technical profile
v0 is effectively a Next.js ecosystem. We detect Next.js on 4,685 of 4,692 likely managed v0 sites, or 99.9%. In contrast, Next.js has at most a 3.5% share on the other platforms. v0 also has the highest visible Cloudflare Web Analytics rate at 4.8%, plus relatively elevated Firebase and Auth.js signals.
Replit appears to be more extensively configured by users. Within the cohort, Replit has the highest observed rates for Google Analytics (33.2%), Stripe (5.4%), and Clerk (4.6%), as well as the highest share of AI tools in the classified sample. That combination suggests that Replit users choose more of their own application services.
Lovable’s scale shows up in absolute integration counts. We see 15,989 sites with Google Analytics, 12,824 with Meta Pixel, 10,732 with Tag Manager, 1,873 with Stripe presence and 1,194 making active Supabase requests. The percentages are not always the highest, but the number of sites makes Lovable the largest observable distribution channel for most third-party technologies.
Base44 and Bolt appear more self-contained. Base44’s runtime remains tightly connected to its own backend services, whereas Bolt’s small 315-site denominator means that individual integrations can move its percentages substantially.
What this may mean for investors
The data support four interpretations.
- Deployment count is a measure of supply rather than success. Lovable’s lead is strategically meaningful, but the scarcity of top-ranked properties shows how little raw creation volume says about durable demand.
- The selected sample includes much more than AI tools. AI tools are only 4.3% of the high-visibility sample. The larger opportunity is to become the base infrastructure for ordinary vertical software, commerce, education, media, gaming and professional workflows.
- The control plane may be more defensible than the generated code. Code can be exported; managed databases, authentication, deployment, domains, analytics, and operational workflows are harder to replace casually. The most important platform question is therefore not “How many projects were generated?” but “How many useful products remain managed 6 or 12 months later?”
- There is substantial attach-rate whitespace. Analytics appear on roughly a quarter to a third of sites, but payments, product analytics, and observability remain uncommon in the public signals. Platforms that make these capabilities native and keep them useful as applications scale may be able to expand revenue beyond generation.
Appendix
How we counted managed sites
We looked for sites that still depend on each platform’s current serving layer, runtime, or control-plane integration rather than sites that merely contain exported code or an old badge. We chose to ignore stale indicators such as generator metadata (easy to preserve after an export) and DNS verification (records can go stale).
The table below documents the platform-specific rules. The ‘managed’ designation is strongest when supported by direct serving evidence. Base44 can also be detected through active platform runtime calls, whereas for v0 we could not confirm continued management because public evidence cannot always distinguish a v0-controlled project from a v0-authored project manually redeployed on Vercel.
| Platform | Current cohort | Detailed detection methodology | Exclusions and limitations |
|---|---|---|---|
| Base44 | 8,945 hosts / 7,515 domains | A host enters the cohort when we observe at least one exact Base44-controlled signal: (1) live *.base44.app; (2) direct CNAME to base44.onrender.com or a Base44 project hostname; (3) JS runtime path/global base44SharedInstances; (4) origin-scoped configuration such as base44_app_id, base44_server_url, base44_functions_version, base44_from_url or base44_analytics_session_id; (5) a request to an app-specific route such as /api/apps/{id}/analytics/track/batch, /public-settings/by-id/{id} or /entities/User/me; or (6) Base44’s OpenID issuer/runtime globals. These signals indicate that the live frontend remains connected to Base44 infrastructure, even when the frontend is hosted elsewhere. |
We reject generic Render signals—Render IPs, rndr-id, and x-render-origin-server—because Base44 shares that infrastructure with unrelated Render customers. We also reject media.base44.com, the shared Supabase bucket, and Base44 image URLs when observed alone. The result measures active Base44 platform usage rather than frontend hosting alone; only about 1,900 hosts expose an attributable Base44 hosting CNAME or default hostname. |
| Lovable | 68,358 hosts / 63,868 domains directly managed; 77,921 hosts in the proxy-inclusive candidate set | The strict cohort requires a successful root response from one of the serving addresses observed consistently across live *.lovable.app deployments: 185.158.133.1, 185.41.148.1, or 185.41.148.2. We corroborate or recover sites hidden behind a reverse proxy using (1) *.lovable.app; (2) CNAME to cname.lovable.app or another project under .lovable.app; (3) Lovable TLS SANs; (4) x-lovablehtml-render-cache or x-lovablehtml-snapshot-key; (5) injected first-party routes /~flock.js, /~api/analytics, /__l5e/events.js, and __l5e assets; and (6) globals/storage such as __lovable_cloud_auth_js_version, __lovableEvents, __lovable_anonymous_id, and __lovable_session. |
Direct edge matching has the clearest hosting interpretation. The additional 9,563 hosts showing active runtime signals outside that edge may include Cloudflare or other reverse-proxy deployments—which Lovable supports—as well as some exported applications that retain injected files. We do not count lovable-uploads, links to lovable.dev, CSP declarations, or legacy gpteng.co assets alone. |
| Replit | 10,361 hosts / 9,575 domains strict; 11,699 hosts / 10,802 domains inclusive | The strict custom-domain rule requires a replit-verify DNS TXT association that currently resolves to 34.111.179.208 or 34.117.33.233. This combines platform-specific ownership evidence with present routing, eliminating abandoned TXT records. We add current first-party deployments through (1) successful *.replit.app hosts; (2) direct CNAMEs ending in .replit.app; (3) replit-cluster; (4) Replit TLS SANs; and (5) exact deployment integrations such as replit-dev-banner.js, replit-badge-v3.js, Replit Pill, the feedback widget, REPLIT_BEACON_VERSION, or ReplitPill. Edge-only successful pages produce the upper estimate. |
The crawl contains 14,541 hosts associated with replit-verify, but many are historical; we only count the 10,361 still routed to Replit in the strict figure. We reject TXT alone, generic Google infrastructure, declared scripts that never execute, and common scaffold directories such as attached_assets or public-objects. Some exported sites retaining platform scripts may remain in the 11,699-host inclusive cohort. |
| Bolt | 333 hosts / 227 domains | The rule is a successful 2xx/3xx root response plus at least one serving-layer marker: (1) hostname ending in .bolt.host; (2) direct/terminal CNAME exactly site-dns.bolt.host; or (3) exact response header x-powered-by: Bolt.new. The header is particularly useful for Cloudflare-fronted custom domains where the public CNAME and Bolt TLS certificate are hidden; it added 11 live hosts beyond the DNS/default-host baseline. A targeted _bolt-verify.<domain> TXT value beginning bolt-verify- can be used as corroboration. |
We exclude bolt.new/badge.js and the preview script when they occur without hosting evidence because they can survive an export or manual Netlify deployment. Generic Netlify headers, IPs, and x-nf-request-id are also insufficient. _bolt-verify is not a standalone positive because it may remain after verification; 61 of 225 known custom-domain registrable domains still exposed it. |
| v0 | 4,957 hosts / 4,692 domains, classified as likely managed | We first establish v0 authorship with exact generator metadata: generator=v0.app, generator=v0.dev, or a recognized mixed v0 generator value. We then require both a successful 2xx/3xx root and current Vercel serving evidence on the same host: (1) *.vercel.app; (2) direct CNAME to cname.vercel-dns.com; (3) server: Vercel; or (4) x-vercel-id/x-vercel-cache. This reduces the 6,820-host authored cohort to sites that remain on the hosting environment used by v0 publishing. Legacy cname.v0.dev, old “Built with v0” routes, official v0 catalog membership, and legacy v0-*.vercel.app names are supplementary evidence. |
This cannot distinguish a project still controlled through v0 from an exported v0 project manually redeployed to Vercel. Vercel evidence by itself is never counted. We reject the shared hebbkx1anhila5yf.public.blob.vercel-storage.com bucket, blob.v0.dev assets, favicon/scaffold patterns, Next.js chunk hashes and v0- hostname prefixes alone. Since v0 removed its badge and automatic prefix, there is no longer a public high-recall control-plane identifier. |
Top 10 sites for each platform
The tables below show the ten highest-visibility sites we found for each platform. CrUX stores rank bands, not exact positions inside a band, so treat rows in the same band as peers rather than a precise leaderboard.
Base44
| Site | Description | CrUX band | Detection evidence |
|---|---|---|---|
| anigum.com | Account-gated anime discovery and viewing site; the live root stopped at Cloudflare verification. | 10,000–50,000 | base44_from_url storage only; weaker-confidence match |
| snapinsta.com.br | Portuguese-language downloader for Instagram videos, photos, Reels, Stories, and IGTV. | 10,000–50,000 | base44SharedInstances; base44_from_url; analytics-session storage |
| ftf-values.base44.app | Trading-value lists and calculators for the Roblox game Flee the Facility. | 50,000–100,000 | Live Base44 host; base44SharedInstances; base44_app_id |
| naira-point-plus.base44.app | Login-gated rewards platform offering points for tasks and referrals, redeemable for cash. | 50,000–100,000 | Live Base44 host; base44SharedInstances; base44_app_id |
| prorewards.io | Indian rewards app offering coins for games, app trials, surveys, and tasks. | 50,000–100,000 | base44SharedInstances; base44_app_id; app-specific analytics request |
| colegiopceirafaelgaleth.com | Ecuadorian online high-school programs offering an official secondary-school qualification. | 100,000–500,000 | Base44 runtime; analytics request; analytics-session storage |
| cup-predictor.com | Free World Cup 2026 bracket builder with leaderboards and private leagues. | 100,000–500,000 | Base44 runtime; base44_app_id; public-settings request |
| elarepitas.com | Community and download catalog for Minecraft 4D skins, add-ons, and mods. | 100,000–500,000 | Base44 runtime; base44_app_id; base44_from_url |
| fantasensiia.base44.app | AI tool generating sensitivity and DPI settings for the game Free Fire. | 100,000–500,000 | Live Base44 host; base44SharedInstances; base44_app_id |
| www.gardiners.uk.com | Website for a UK civil-engineering, construction, landscaping, and electrical contractor. | 100,000–500,000 | Base44 runtime; base44_app_id; base44_from_url |
Lovable
| Site | Description | CrUX band | Detection evidence |
|---|---|---|---|
| christianlyricz.com | Christian worship lyrics, chords, playback, and presentations in Telugu, Hindi, and English. | 10,000–50,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| connectplug.com.br | Redirects to CPlug, a Brazilian POS, ERP, delivery, and restaurant-management platform. | 10,000–50,000 | Detected host served directly by Lovable edge during crawl |
| pedir.delivery | Redirects to Multipedidos, an ordering and operations system for restaurants and delivery businesses. | 10,000–50,000 | Detected host served directly by Lovable edge during crawl |
| taskreward.cc | Malaysian marketplace where users perform paid tasks that generate social-media engagement. | 10,000–50,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| in.adda.io | Redirects to ADDA, an apartment and residential-community management platform. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| ear-nova.online | Rewards site offering payment for watching ads, completing microtasks, and referring users. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| freecaseevals.com | Auto-accident lead-generation questionnaire matching potentially eligible visitors with attorneys. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| skillmoney.online | Account-gated rewards site advertising cash for games, ads, referrals, and leaderboard activity. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| tiomono.app | Paraguayan online casino offering slots, live games, and a signup bonus. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
| transfertci.com | Côte d’Ivoire application for mobile-credit and data transfers; showed a login page and an app download during the crawl. | 50,000–100,000 | Lovable edge; injected /~flock.js; /~api/analytics |
Replit
| Site | Description | CrUX band | Detection evidence |
|---|---|---|---|
| bolaodacopa.ai | Brazilian World Cup prediction-pool platform with automatic scoring and rankings. | 10,000–50,000 | Direct replit-verify TXT; current A record points to Replit |
| the.journey.com | Staff-only marketing asset-builder login for the Journey boutique-hotel loyalty program. | 50,000–100,000 | Direct replit-verify TXT; current A record points to Replit |
| lotodigital.org | French-language multi-country African digital lottery application. | 50,000–100,000 | Direct replit-verify TXT; current A record points to Replit |
| onthon.com | Bangladeshi marketplace for social accounts, channels, websites, and game accounts. | 50,000–100,000 | Direct replit-verify TXT; current A record points to Replit |
| andalmet.es | Andalusia-focused weather site providing forecasts, current conditions, and regional news. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
| students.arrahmah.org | Student login portal for Ar-Rahmah Institute. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
| cointrend.live | Account-gated cryptocurrency staking service branded Elite Staking Network. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
| digitallottery.org | Second domain serving the same visible Loto Digital African lottery experience. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
| directcard.com | Consumer credit-card application and account-management site focused on credit building. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
| drunkers.store | Mexican store selling digital games, Game Pass, gift cards, and activation codes. | 100,000–500,000 | Direct replit-verify TXT; current A record points to Replit |
Bolt
| Site | Description | CrUX band | Detection evidence |
|---|---|---|---|
| concours.agriculture-fpa.ma | Moroccan application portal for agricultural vocational-training admissions. | 500,000–1m | Successful root response; CNAME site-dns.bolt.host |
| retomundial.hubmetaverso.com | Company-gated World Cup challenge that requires an employer-provided access link. | 500,000–1m | Successful root response; CNAME site-dns.bolt.host |
| www.aatmashuddhi.com | Jain community portal containing spiritual readings, events, programs, and religious travel schedules. | 1m–5m | Successful root response; CNAME site-dns.bolt.host |
| bachelor-friday-coun-3rvi.bolt.host | Bengali novelty calculator counting how many Fridays a user has spent without a spouse. | 1m–5m | Live Bolt deployment; x-powered-by: Bolt.new |
| bellytox-lifestyle.bolt.host | Health and wellness habit-tracking community; currently opens on a terms-acceptance gate. | 1m–5m | Live Bolt deployment; x-powered-by: Bolt.new |
| www.capepatio.com | South African retailer of premium cast-aluminium patio and outdoor furniture. | 1m–5m | Successful root response; CNAME site-dns.bolt.host |
| chatmoney-members-ar-f0ta.bolt.host | Sign-up funnel for an Africa-focused course and community teaching AI monetization strategies. | 1m–5m | Live Bolt deployment; x-powered-by: Bolt.new |
| app.cleanedgemanpower.com | HR and workforce portal for job applications, employees, supervisors, resignations, and inventory. | 1m–5m | Successful root response; CNAME site-dns.bolt.host |
| www.corhs.org | Colorado reptile and amphibian rescue, adoption, education, and wildlife organization. | 1m–5m | Successful root response; CNAME site-dns.bolt.host |
| www.evopmetaverse.io | Tokenized virtual world featuring land, businesses, games, NFTs, marketplaces, and staking. | 1m–5m | Successful root response; CNAME site-dns.bolt.host |
v0
| Site | Description | CrUX band | Detection evidence |
|---|---|---|---|
| www.logoai.com | AI logo, brand-identity, asset-kit, image, and video design platform. | 50,000–100,000 | generator=v0.app; server: Vercel; x-vercel-cache |
| www.bygens.app | Token-trading application; the current public experience is an account login. | 100,000–500,000 | generator=v0.dev; server: Vercel; x-vercel-cache |
| cbrilliance.io | Digital-asset ecosystem combining trading, project discovery, shopping, education, and community. | 100,000–500,000 | generator=v0.app; x-vercel-cache |
| darlink.ai | 18+ platform for customizable AI companions, role-play, images, and video; the root hit Cloudflare verification. | 100,000–500,000 | generator=v0.dev; x-vercel-cache |
| gainzalgo.com | Paid TradingView indicator and trading-signal service for cryptocurrency, forex, and equities. | 100,000–500,000 | generator=v0.app; server: Vercel; x-vercel-cache |
| www.goteamgames.com | Corporate team-building tools, currently centered on company World Cup prediction pools. | 100,000–500,000 | generator=v0.app; server: Vercel; x-vercel-cache |
| maestro.org | AI-first higher-education learning platform; currently presents login and early-access signup. | 100,000–500,000 | generator=v0.app; server: Vercel; x-vercel-cache |
| mypec.app | Pragati Engineering College student-information portal with separate faculty and student logins. | 100,000–500,000 | generator=v0.app; server: Vercel; x-vercel-cache |
| pastescript.com | Public text and code paste-sharing application with templates and a script generator. | 100,000–500,000 | generator=v0.dev; server: Vercel; x-vercel-cache |
| seatsidekick.com | World Cup ticket-price comparison and alert service, now displaying a post-event thank-you page. | 100,000–500,000 | generator=v0.app; server: Vercel; x-vercel-cache |