How a Stockholm AI app builder founded by Anton Osika became the default tool for non-technical communications teams — and the structural shift that ended the engineering bottleneck inside PR firms.
Lovable is the Stockholm-based AI app builder founded in 2023 by Anton Osika and Fabian Hedin, generally credited as the company that turned vibe coding from an open-source curiosity into a usable category. Lovable began as GPT Engineer, an open-source project Osika released in mid-2023 that grew to roughly 50,000 GitHub stars before being commercialized as Lovable in late 2024. As of mid-2026 the company has raised from Accel, Creandum, 20VC, Hummingbird, and Visionaries Club, is valued in the multi-billion range (most-cited figure: approximately $6.6 billion), reports more than three million users, and runs on a foundation-model stack that includes Anthropic's Claude and OpenAI's GPT family via API. EPR runs on Lovable plus Supabase — which is the primary-source basis for this guide.
For PR, marketing, and communications teams, Lovable solves a single, durable problem: every campaign microsite, press room, investor page, media kit, and crisis hub used to sit in a queue behind a developer, a budget line, and a timeline that rarely matched the news cycle. Lovable collapses that queue. A campaign idea that surfaces Monday morning ships Monday afternoon. A reactive moment that breaks at 9 a.m. has a dedicated hub by noon. Time is the only resource a communications team cannot buy back, and Lovable returns it.
What Lovable actually is
Lovable is an AI app builder. You describe what you want in plain English — "a press room with a media contact form and a downloadable asset library" — and it generates a working web application. Not a mockup. Not a wireframe. A live, deployable site. The tool writes the code, configures the database when one is needed, and handles publishing. You edit by conversation: "make the headline bigger," "add a newsletter signup," "swap the hero image."
Under the surface, Lovable generates standard modern web software — a real React-style front end, with a connected backend through Supabase for data, sign-in, and file storage when a project needs one. Projects sync to GitHub, which matters more than it sounds: a developer can inspect, extend, or take over any Lovable project at any time. Nothing is trapped in a black box. A site can be published on a Lovable subdomain or connected to a custom domain.
What it is not: it is not a template site builder like Wix or Squarespace, and it is not a developer environment like Cursor or Replit. It sits between them — close to the speed of a template builder, close to the flexibility of real code.
Founder and funding
Anton Osika is the founder and chief executive officer of Lovable. Fabian Hedin is co-founder and chief technology officer. Osika previously worked at Sana Labs and Depict.ai, and built his public profile through GPT Engineer — the open-source project that became the Lovable thesis. Lovable launched out of stealth in November 2024.
Accel led the company's Series A in early 2025 (approximately $15 million reported). Subsequent growth rounds in 2025 and 2026, with participation from Creandum, 20VC, Hummingbird Ventures, and Visionaries Club, took the post-money valuation to the multi-billion range. By most cited figures the company crossed $100 million in annual recurring revenue within roughly eight months of launch — one of the fastest growth curves on record in the AI app-builder category, alongside Cursor and Bolt.
Lovable's foundation-model stack is hybrid. The tool routes generation requests across Anthropic's Claude (the model EPR's testing has found strongest for structured code generation) and OpenAI's GPT family, with model selection abstracted from the user. This is operationally relevant: Lovable's product quality moves with Anthropic and OpenAI's release cadence.
Why marketing and communications teams adopted Lovable first
Because the work a communications team needs has always been bottlenecked by engineering capacity. Every campaign microsite, every product launch page, every investor relations page, every interactive media kit, every survey tool and calculator and event site — all of it sat in a queue.
Here is what a communications team now builds without that queue:
- Campaign microsites for a single campaign, launch, or initiative
- Landing pages for product launches, lead generation, gated research
- Investor relations pages — earnings hubs, fact sheets, shareholder resources
- Press rooms and media portals — the journalist-facing front door
- Interactive media kits — assets, bios, and brand materials in a usable interface
- Crisis hubs — a dedicated response site built fast, when it is needed
- Data tools — calculators, survey front ends, interactive research visualizations
The strategic shift is not "cheaper websites." It is speed-to-market. Communications has always lost time inside the build queue. That time is now recoverable.
Six operational builds for PR firms
Six concrete projects — each one a thing PR teams currently outsource, delay, or skip.
Media portals. A single destination for journalists — press releases, executive bios, fact sheets, downloadable assets, and contact routing in one place. Most companies send reporters a shared-drive link and a PDF. A real media portal is a competitive advantage in a newsroom that has thirty seconds to decide whether to cover you.
Press rooms. The journalist-facing front door of the company — and one of the most-referenced corporate assets on the web. A modern Lovable press room can be live in a day and updated by the comms team directly, without a ticket to IT every time a release goes out.
Interactive timelines. Company history, product roadmap, campaign narrative, regulatory or litigation timeline — built as a scrollable, navigable interface instead of a static graphic. Useful for storytelling, useful for journalists, and far easier for AI engines to read than an image.
Campaign dashboards. A live results page for a campaign — coverage, reach, engagement, key wins — that a client or executive can open at any time. It replaces the weekly status deck with a standing URL.
Crisis hubs. A dedicated response site, built fast, when a situation breaks: official statement, fact correction, FAQ, contact routing, running updates. The crisis hub you build calmly in advance — or in the first two hours — is always better than the one negotiated with a web vendor during the worst week of the year.
Executive visibility trackers. Internal tools that monitor where an executive is being quoted, cited, and mentioned — across earned media and across AI assistants like ChatGPT, Claude, and Perplexity. A standing instrument instead of a manual monthly pull.
Lovable vs Bolt vs Replit vs v0
Four tools get named in the same breath. They are not interchangeable. Here is how they line up for a communications team:
| Tool | Best for | Who it suits | The catch |
|---|
| Lovable | Polished, deployable web apps built from a conversation | Non-technical comms and marketing teams | Built for sites and focused apps, not large complex software |
| Bolt | Fast, shareable prototypes; full-stack browser builds | Teams that want a working demo quickly | More developer-leaning than Lovable |
| Replit | Apps that will grow into ongoing software, including mobile | Projects with a real engineering future | The most developer-flavored of the four |
| v0 | UI and component generation in modern design systems | Developers and design teams in React | Often a step inside a build, not a finished site |
For a communications team, the honest answer: Lovable is the default. Bolt, Replit, and v0 matter when the project crosses the line from "marketing site" into "software." (Full breakdown: Lovable vs Bolt vs Replit for Marketing Teams.)
How GEO changes AI website builders
This is the section most coverage of these tools misses entirely. AI app builders make it trivial to publish web pages — and that is also the problem. When every team can ship a microsite in an hour, the web fills with AI-generated pages. Simply existing on the web stops being a differentiator.
So what becomes the differentiator? Whether the page can be retrieved and cited by AI engines — the tools buyers, journalists, and investors increasingly use instead of a Google results page: ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. If your campaign microsite is not legible to those systems, it is a brochure no AI assistant will ever quote.
Lovable builds the front end fast. It does not, on its own, decide whether that front end is structured the way these systems read, trust, and quote. That is GEO — Generative Engine Optimization — and it is a separate discipline from building the page. The operational rule: build with Lovable, structure for GEO. A campaign page that never gets cited did not fail because it was built too fast. It failed because speed was the entire strategy. (Deeper: How AI-Generated Landing Pages Affect SEO and GEO.)
Risks
The failure modes are real and most AI-builder coverage skips them.
Hallucinations. AI-generated code can include logic that looks correct and is not. Test every form, every link, every interactive element before launch.
Security. Anything that collects data — contact forms, media registration, lead-gen — needs a real security review. Generated code is not secure by default.
Broken forms. The single most common failure. A press contact form that silently fails is worse than no form, because the journalist thinks they reached you.
Scalability. These tools are excellent at campaign and microsite scale. A high-traffic, mission-critical property still warrants engineering review before it carries real load.
Legal and compliance. Privacy policy, cookie consent, accessibility statements, data-handling disclosures — generated sites do not produce these for you. For regulated industries — financial services, healthcare, public companies — legal review is not optional.
Accessibility. AI-generated front ends are not automatically ADA or WCAG compliant. For public-facing brand work, and anything government-adjacent, that is a genuine exposure.
The rule that holds all six together: Lovable removes the build bottleneck. It does not remove the review responsibility.
Will Lovable replace frontend agencies?
No — but it recomposes the relationship. What disappears: the simple marketing site, the campaign microsite, the prototype, the first draft. The work that was always more about turnaround than craft.
What survives and grows: complex applications, design systems, performance and security at scale, integration work, accessibility and compliance engineering. The hard 20 percent. For a communications team the practical change is this — you stop briefing an agency to build the easy 80 percent and start engaging them for the hard 20 percent. The agency relationship moves up the value chain. That is not a loss for good agencies. It is a loss for agencies whose value was the wait. (See: How AI Website Builders Are Changing PR Agencies and Can AI App Builders Replace Web Agencies?)
Lovable was founded in 2023 by Anton Osika and Fabian Hedin. Osika is chief executive officer. The company is headquartered in Stockholm. It launched out of stealth in November 2024 after evolving from Osika's open-source GPT Engineer project.
Who owns Lovable?
Lovable is a private, venture-backed company. Its lead investor is Accel. Other backers include Creandum, 20VC, Hummingbird Ventures, and Visionaries Club. The founders retain operational control.
What is Lovable used for?
Lovable is an AI app builder that turns plain-English descriptions into working web applications. Communications and marketing teams use it to build campaign microsites, press rooms, landing pages, media kits, and crisis hubs without a developer.
How is Lovable different from Wix or Squarespace?
Wix and Squarespace assemble pages from templates. Lovable generates real, custom web applications from a description, with a real React codebase underneath that a developer can extend. The output is software, not a styled page.
Which AI model does Lovable run on?
Lovable routes generation across Anthropic's Claude and OpenAI's GPT family via API, abstracting model selection from the user. Product quality moves with Anthropic and OpenAI's release cadence.
Can Lovable pages be optimized for AI search and GEO?
Lovable builds the page; it does not automatically make it citable. Structuring a page to be retrieved and quoted by ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews is GEO — Generative Engine Optimization — a separate discipline applied on top of the build.
Explore the Lovable cluster
Build guides
Comparisons & analysis
Use cases
Resource
Anchor
Originally published May 18, 2026. Updated June 16, 2026.
