Research brief · 26 July 2026

What people actually build with Lovable

A data-backed workflow priority for PeachJam: which generated applications matter most, what end-to-end journeys to reproduce, and which failure modes those journeys should expose.

50MLovable projects built to date
14,300+users in Lovable’s May 2026 survey
48.2%websites plus business-operations software
92.2%covered by the eight proposed golden journeys

Short answer

Start with business presence, then the software that runs the business.

Lovable’s own data says websites are the largest category (26.4%), followed by business operations (21.8%). Consumer apps, analytics dashboards and e-commerce bring the top five to 76.1%. This is a much better PeachJam roadmap signal than copying whatever happens to be visually prominent in Lovable’s template gallery.

Recommendation: do not begin by adding dozens of templates. Build eight executable, multi-turn golden journeys that run on both Lovable and PeachJam. Score the resulting app behavior, iteration safety, data ownership, publishing and mobile output—not whether the assistant said it completed the task.
First

Public business website

Responsive pages, lead capture, SEO, publish, custom domain and social previews.

Second

CRM / operations tool

Authentication, roles, relational CRUD, filters, reporting, imports and notifications.

Third

Consumer account app

Onboarding, mobile use, profiles, user-owned records, persistence and media uploads.

Primary evidence

Lovable’s published use-case distribution

The classification below comes from Lovable’s 2026 Build Economy study. The methodology combines anonymized platform activity from January 2025–May 2026 with a May 2026 survey of more than 14,300 users. The category chart itself is identified as user-survey data.

Bar lengths are normalized to the largest category. Source: Lovable Build Economy 2026.

What “success” means to users

25.8% want a launched product with active paying users; 17.7% want a tool their team uses daily; 14.5% want a polished professional brand presence; 12.3% want a working prototype; and 6.4% want a client handoff.

Who PeachJam should optimize for

80% identify with non-technical roles, 80% build solo, and 45.7% are founders or co-founders. Error states, approval boundaries and recovery must be understandable without requiring browser devtools or database expertise.

Prioritization

Prevalence is the baseline; failure yield changes the order.

Websites and business tools are both common and cheap to exercise. E-commerce, marketplaces and AI tools have smaller shares but disproportionately stress payments, secrets, role boundaries, asynchronous jobs and recovery. They belong in the core suite, not the long tail.

Priority Workflow family Lovable share Why it belongs Primary defect yield
P0Public business website26.4%Largest category; fastest parity signalResponsive layout, assets, forms, routing, SEO, publish
P0CRM / business operations21.8%Daily-use software and strongest internal-tool patternAuth, roles, relational CRUD, data isolation, integrations
P0Consumer account app11.8%Top-three category and mobile-heavyOnboarding, session lifecycle, ownership, uploads, persistence
P1Analytics dashboard9.1%Common standalone category and repeated ops featureImports, chart correctness, filters, dates, empty/loading/error states
P1E-commerce storefront7.0%Revenue-critical and officially supportedCart, checkout, webhook idempotency, inventory, email
P1Productivity utility6.2%Common solo-builder and passive-income shapeCalculation correctness, persistence, export, share links
P1Marketplace5.4%Smaller share but highest multi-role complexityBuyer/seller permissions, search, messaging, moderation, payments
P1AI / chatbot product4.5%Strategic for an AI builder and integration-heavySecrets, streaming, citations, quotas, retries, provider failures
P2Content and education5.0% combinedImportant but partially exercised by website flowsSSR/prerender, content structure, enrollment, media, discoverability

Executable corpus

Eight golden journeys cover 92.2% of the published categories.

Each journey should be run as the same sequence of prompts on Lovable and PeachJam. Assertions must inspect the generated app, data and deployed behavior. Preserve the artifacts and screenshots from every turn so regressions can be attributed.

01

Local-service website and lead capture

Build → refine from an attached brand reference → add a validated contact form → add metadata → publish → update.

  • Assert desktop and iPhone navigation, form validation/submission, accessible labels, 404 handling and no horizontal overflow.
  • Assert title, description, Open Graph image, sitemap/robots behavior and published/preview parity.
  • Make a follow-up visual change and verify the form and mobile navigation survive.
Build a polished mobile-first website for Cedar Dental with Home, Services, Team and Contact pages. Add a lead form with name, email, phone, requested service and consent. Use a calm cream-and-sage visual system, real empty/error/success states, and accessible navigation.
02

Purpose-built CRM / operations tool

Build relational data → add manager/member roles → import leads → add pipeline and reporting → trigger email notification.

  • Use two accounts to prove row/workspace isolation and role restrictions rather than checking route visibility only.
  • Exercise create, edit, filter, sort, pagination, CSV import/export, optimistic updates and refresh persistence.
  • Verify schema changes preserve existing data and that notifications are emitted once.
Build a CRM for a five-person agency. Managers can manage all leads and members can update assigned leads. Include a leads table, Kanban pipeline, company and contact records, tasks, weekly dashboard, CSV import/export and an audit trail.
03

Consumer habit / wellness app

Sign up → onboarding → profile/avatar → create and complete habits → logout/login → second-user isolation.

  • Verify session restore, protected redirects, email confirmation/recovery and delete-account behavior.
  • Check daily boundaries and streaks across timezone changes; test empty, loading and offline/error states.
  • Use the installed iPhone Pro profile for the primary evidence, not an arbitrary narrow viewport.
Build a mobile-first habit tracker with email sign-up, a short onboarding flow, profile photo, daily habits, streaks, weekly history, reminders and private per-user data. Include a friendly empty state and dark mode.
04

Analytics dashboard over imported data

Upload a CSV → map columns → show KPIs/charts → filter dates → drill into rows → export a report.

  • Test malformed/empty files, large datasets, decimal and timezone boundaries, missing values and duplicate uploads.
  • Assert chart values against the underlying rows, not screenshots alone.
  • Verify filters survive navigation and links can be shared without leaking private data.
Build a sales analytics dashboard that imports an orders CSV, maps its columns, shows revenue, orders, average order value and conversion, supports date and channel filters, drills into source rows, and exports the filtered report.
05

E-commerce storefront

Catalog → search/product detail → cart → authentication → test checkout → webhook → order email → inventory update.

  • Verify cart persistence, price authority, sold-out races, duplicate webhook delivery and failed/cancelled payment recovery.
  • Prove entitlements/order state from provider-backed events, not a client-side success redirect.
  • Check secrets never enter generated client bundles and order emails are sent once.
Build a small ceramics store with categories, search, product variants, inventory, cart and customer accounts. Add Stripe test checkout, order history, an order-confirmation email and an admin order/inventory view.
06

Single-purpose productivity utility

Build calculator/generator → save named scenarios → export → share → revise calculation rules without losing records.

  • Use golden numeric cases, invalid inputs, rounding, locale formatting and keyboard-only interaction.
  • Test local-first behavior separately from signed-in cloud persistence.
  • Assert generated PDF/CSV contents and share-link access rules.
Build an ROI calculator for sales teams with configurable costs and gains, monthly/yearly views, named scenarios, comparison charts, PDF export and a read-only share link. Include exact validation and rounding rules.
07

Two-sided marketplace

Buyer/seller onboarding → listing with images → search/filter → favorite/message → moderation → purchase or booking.

  • Use buyer, seller and admin accounts to test ownership and forbidden actions.
  • Exercise private/public storage, listing status transitions, concurrent booking and abuse reporting.
  • Confirm search and filters work after pagination, refresh and mobile navigation.
Build a local workshop marketplace with buyer and host accounts. Hosts publish workshops with images, capacity and dates; buyers search, favorite, message and book. Add moderation, booking history and role-specific dashboards.
08

AI knowledge assistant

Connect server-side AI → upload source files → stream an answer with citations → enforce limits → recover from provider failure.

  • Assert secrets and system instructions stay server-side; validate file ownership and citation provenance.
  • Exercise timeout, retry, cancellation, partial streams, quota exhaustion and duplicate submission.
  • Provider-routing or request/response changes in PeachJam itself remain operator-reviewed.
Build a support knowledge assistant where signed-in teams upload PDFs and ask questions. Stream answers with source citations, keep each workspace private, show usage limits, and provide clear timeout, retry and no-answer states.

Cross-cutting probes

The workflow suite should find platform bugs, not merely app bugs.

Recent community reports repeatedly describe a gap between an impressive first draft and production: auth hangs or redirect loops, permissive data policies, Stripe success without entitlements, preview/live differences, missing environment variables, sync failures and repeated AI “fix” loops. These reports are anecdotal, so they are used as test hypotheses, not prevalence statistics.

Iteration safety

A narrow follow-up must not replace routes, erase data code, remove auth, regress responsive behavior or “fix” unrelated files.

Identity and ownership

Test first login, refresh, expiry, recovery, logout, two users, roles and forbidden direct API calls.

Database lifecycle

Test schema creation, safe migrations, existing-data preservation, row ownership, imports and concurrent writes.

Preview → production

Compare URLs, assets, redirects, OAuth callbacks, secrets, service workers, metadata and backend state after publishing.

Money and async events

Test provider readback, webhook replay, delayed events, cancellation, failed payment and idempotent fulfillment.

Recovery and observability

Interrupt builds, time out providers, disconnect integrations and prove the user sees a truthful, actionable recovery path.

Mobile and accessibility

Use the representative iPhone profile, keyboard navigation, focus restoration, labels, reduced motion and real content lengths.

External connectors

Test admin permissions, project linking, credential rotation, unlink/delete effects and staging/production separation.

GitHub ownership

Prove repository creation, two-way sync, external commits, conflicts, disconnect/reconnect and deployment from owned code.

Current repository comparison

PeachJam’s control-plane coverage is broad; its generated-app corpus is narrow.

This is based on origin/main@4fe1d501. PeachJam already has strong browser coverage for its dashboard, project editor, publishing, cloud, sharing, settings and collaboration surfaces. The gap is representative apps built and exercised as real user workflows.

What already exists

  • 41 frontend Playwright spec files spanning auth, editor, Cloud, publish/share, billing, settings, collaboration and route behavior.
  • Production/runtime smokes for preview, custom domains, GitHub export, email, images, collaboration and managed-backend lifecycle.
  • 14 curated template seeds: eight websites and six apps, including internal tools, services, events, portfolios and project/product management.
  • Real generated-app smoke fixtures for authenticated notes/todos, plus simple client-side todo and hello-world/image builds.

Material missing coverage

  • No representative golden journey for e-commerce, marketplace, consumer profile/media, analytics import, or AI/chatbot products.
  • No generated-app payment lifecycle proving checkout → provider event → entitlement/order state.
  • No multi-role generated CRM/marketplace proof using two users and direct forbidden-operation checks.
  • No single harness that runs an identical multi-turn prompt trajectory on Lovable and PeachJam and compares observable contracts.
Best next implementation: introduce scenario manifests for journeys 01–03 first, with prompt turns, required observable contracts, actor fixtures, mobile evidence and artifact paths. Keep deterministic fixtures, paid provider staging runs and production smokes as three separate evidence levels.
PhaseDeliverableExit condition
AShared journey manifest and scorerOne scenario can run on both products without product-specific assertions hiding mismatches.
BWebsite, CRM and consumer-app journeys60% of Lovable’s published category mix has behavioral coverage.
CDashboard, store and utility journeys82.3% coverage; import/export and revenue paths exercised.
DMarketplace and AI assistant journeys92.2% coverage; multi-role, streaming, secrets and failure recovery exercised.
ETemplate expansion informed by journey failuresTemplates reuse proven capabilities instead of becoming visual demos of unverified behavior.

Repository evidence: template seeds, stateful runtime smoke, builder continuity smoke, and test/smoke scripts.

Method and limits

What is measured, what is inferred

Measured

Category shares, builder roles, motivations and success definitions come from Lovable’s published 2026 study. The official template sitemap was independently inventoried on 26 July 2026: 177 leaf templates; the largest families were portfolio (41), services (28), internal tools (21), music (14), events (13), blog (10) and SaaS (10).

Inferred

The recommended PeachJam ordering combines category prevalence with likely defect yield. Template counts are treated only as product-investment evidence, not user demand. Community discussions are used to select failure probes, not to estimate how often failures occur.

Lovable’s categories are broad. “Business ops” can include CRMs, inventory, HR and approval systems; “consumer apps” can include wellness, social and personal finance. The golden journeys intentionally choose representative stress cases, not a claim that every app in a category has the same shape.

Sources

Primary sources first