Hostinger’s AI Bet Now Runs on Other People’s Habits

Hostinger CEO Giedrius Zakaitis says people still treat AI tools as week-long projects even when the work takes minutes. The Lithuanian firm now treats hosting as background plumbing and wants those same customers to run the businesses they just built.

That handoff gives Hostinger a job it did not have when a WordPress install was the hard part. It has to secure and operate software written by people who never hired a developer.

Hostinger’s New Job Starts After Publish

Zakaitis, appointed on June 16, 2026 after 14 years at the company, started in customer success and later ran product and technology. He took over from Daugirdas Jankus, who stays on strategic projects. His brief is blunt. Anyone with an idea should be able to build it and run it without writing code.

He describes hosting as almost an abstraction. Ten years ago the first question was whether you could build the thing. Now building is easy, and the question has moved to whether you can actually run the business. That can mean an overnight report on what to do next, or a look at the competition, rather than a live homepage.

The money already follows that bet. Hostinger posted 51 percent revenue growth in 2025, reaching €275.4 million. That was the fourth straight year above 50 percent, and roughly four times the €69.6 million it recorded in 2022, a 58 percent compound annual rate. By the end of 2025 it served 4.6 million small businesses and solo operators in more than 150 countries, up from about 1.5 million in 2022. The June 16, 2026 appointment post put the live base at more than 5 million users, with nearly 900 staff. India, Brazil, Indonesia, the United States, and France lead by active users.

HOSTINGER REVENUE AND CUSTOMERS

2023 was the first year of EBITDA profit, at €2.4 million. From 2022 through 2025 the firm added €205.8 million in new revenue and 3.1 million net customers. Kodee, its support agent, handled 81% of customer support chats by the end of 2025, up from 50 percent at the start of that year, and saved about €9 million. The June 16, 2026 post said Kodee could then perform more than 500 admin tasks, including DNS, backups, and store work. Hostinger Reach, the email marketing tool, had 150,000 users by December 2025. Net Promoter Score sat at +59.

Jankus, still CEO when those 2025 figures came out, put the pitch this way: customers do not want to manage infrastructure or tools. They want to run their businesses. AI, in that telling, is how Hostinger takes the fiddly bits off their hands.

Why Customers Still Default to WordPress

The product map is already past “make a site.” On August 18, 2026, Hostinger folded its visual website builder and its AI app builder into one platform, AI Builder, with an agentic mode and a manual mode. The Hostinger AI Builder launch post said a prompt, PDF, screenshot, or link can start a project, and that stores, accounts, and email marketing sit in the same flow.

The mix of what people build has already moved. Hostinger said that a year before that launch, fewer than 4 percent of projects sat outside ordinary website categories. By August 2026, nearly one in five were software products, internal tools, or learning platforms, a fivefold rise in 12 months. The AI web-app product had more than 800,000 customers by the end of 2025. Zakaitis also says people now ship their own booking tools, quote calculators, and small SaaS apps instead of buying generic software.

Habit still sits on the other side of that chart. A much larger group than he expected still chooses WordPress because it is what they know, with years of plugins and muscle memory. Hostinger still turns up at WordCamp events and still publishes WordPress how-tos, because that workflow pays the bills. SEO work also keeps pulling some builders back toward WordPress even after they try a hosted AI builder, because search habits change slower than page generators.

Virtual servers are the other quiet shift. Zakaitis says AI agents do not belong on a laptop. Customers run them on a VPS with full access, which is a different support and security problem from a one-click blog.

OpenClaw Arrives With Guardrails Hostinger Wrote

OpenClaw, formerly known as Moltbot and Clawdbot, is a personal AI assistant that runs on your own machine or server. It talks over WhatsApp, Telegram, Slack, and Discord. It can read files, run shell commands, and act in a browser. The original design was personal use. Default settings were a poor fit for a public VPS.

Zakaitis says Hostinger saw huge demand from existing customers who wanted to try it, and that getting started was hard, so the company made install much easier. It also had to fix the security holes that came with that ease. Authorisation was not on by default. Hostinger changed the config, added token-based login, and put extra checks around the template. “As a company, we can identify potential attack vectors and put guardrails in place rather than simply handing customers a VPS and leaving them to figure everything out themselves,” he said.

The public help pages match that story. A new VPS can ship with OpenClaw pre-selected. Docker Manager can pull it from a catalog. In both paths Hostinger issues an auto-generated OpenClaw gateway token and hides the value until the customer copies it. Login to the dashboard requires that token. Docs warn people never to commit the token or API keys, and they point to a separate hardening guide and an SSL setup for HTTPS. For a new box, Hostinger recommends KVM2 or larger.

THE OPENCLAW DEFAULTS HOSTINGER CHANGED

• Gateway login: A random token is created at deploy, then required before the dashboard opens.
• One-click path: Docker Catalog or a pre-installed VPS replaces a raw install for non-technical users.
• Channel hooks: WhatsApp numbers and Telegram bot tokens sit in the same form as model API keys.
• Follow-up docs: SSL via Traefik, token recovery in Docker Manager, and a separate harden-and-secure guide.

That is the second job hiding inside “anyone can build.” The customer who can stand up an agent in a catalog click is the same customer who will not audit bind addresses, reverse proxies, or skill permissions. Cheap VPS plans are already how some builders run personal agents in production, including bookkeeping bots and multi-channel posting stacks. Hostinger’s answer is not to block the tool. It is to own the boring security work those users will skip.

On September 11, 2026, the company said OpenClaw 2.0 was already live on Managed OpenClaw, with major security, onboarding, and session upgrades after 16,000-plus merged pull requests from 900-plus contributors.

OpenClaw 2.0 is here — its biggest update yet. 🦞

16,000+ merged PRs. 900+ contributors. Major upgrades to security, parallel execution, shared sessions, memory, and onboarding.

Already live on Managed OpenClaw at Hostinger — tested and ready to go.

See what’s new 👇 pic.twitter.com/8dxoEIYnHp

— Hostinger (@Hostinger) September 11, 2026

Shared hosting has not gone quiet while that agent stack grows. Customers still fight malware on ordinary sites, which is the older version of the same problem: Hostinger sells simplicity, then inherits the mess when the simple path is also the leaky one.

FixGuard Won Silver by Patching AI Sites

In September, Hostinger tried to prove the “idea to live business” pitch with a timed build. The 21-Day Startup Challenge prizes totaled $3,500. Entrants had to be 18 or older, hold a Hostinger account, and work alone. They had to use four products in a real way: Hostinger Agent for research and planning, AI Builder for the MVP, web hosting for a public page, and a VPS for backend work. Registered builders got 1,000 Agent credits, 15 AI Builder credits, and one month of Unlimited Web Hosting plus a VPS KVM 1 plan.

THE SEPTEMBER BUILD WINDOW

1. August 24, 2026: Registration opens.
2. September 3, 2026: The 21-day build period begins.
3. September 24, 2026: Submissions close at 23:59 UTC.
4. September 29, 2026: Winners are announced.

Judges scored product and user experience at 25 percent, the business idea at 20 percent, business potential at 20 percent, meaningful use of all four Hostinger products at 20 percent, and creativity at 15 percent. The brief was useful over flashy. A small tool that worked could beat a grand project that did not ship.

THE THREE CASH WINNERS

• BedLedger, $2,000: A rent stack for small landlords in India, covering invoices, reminders, payments, and digital receipts.
• FixGuard AI, $1,000: A checker that finds broken forms, dead links, and accessibility faults on AI-built sites, including its own.
• LectureLift, $500: A study layer that turns recorded university lectures into notes, flashcards, quizzes, and clearer explanations.

FixGuard AI is the tell. The silver project is a repair kit for the websites the same AI stack produces. BedLedger is the other tell: a landlord in India doing rent collection without a hired developer, which is exactly the down-market user Zakaitis says venture-backed software firms are nudged to abandon.

He also recounts a customer who left hospital with a recovery plan and built a tracker that asked whether the stretches got done. Internal tools for bookings, inbound mail, and quotes show up in the same bag. The challenge was a marketing event. It was also a sample of the workload Hostinger now has to host, patch, and keep online after day 21.

Four-Year Plans Keep Hostinger With Small Customers

Zakaitis argues that Lithuania’s thin capital markets trained this behaviour. Companies there were often bootstrapped because they could not raise easily, so they had to be careful with cash. Hostinger, he says, has grown entirely through revenue and reinvestment and has never raised outside capital to fund operations. Tesonet bought shares in 2018 and ConHostinger in 2021, both as purchases from existing holders, not funding rounds. Hostinger’s about page says ConHostinger holds a minority stake of around 30 percent, with the rest held by founders and staff.

Early customers financed the next wave by buying four-year plans, which paid for servers, product work, and marketing. A small home market made global sales a math problem, not a later ambition. Largest user bases now sit in India, Brazil, Indonesia, the United States, and France, with the United Kingdom, Mexico, Colombia, Argentina, and Spain in the next tier.

He calls the absence of fund cycles a luxury. Public software firms, he says, keep talking about going upmarket because those accounts pay more. “But there are millions of people who want to start businesses. Individually they might not pay as much, but there are so many of them. We can build for those customers for the long term without somebody pushing us to move upmarket.”

That choice is why OpenClaw templates and a 21-day contest for amateur builders sit on the roadmap at all. A firm being pushed toward high-average-value accounts would not spend security engineers on a personal agent that was never meant for a public VPS. Hostinger can, because those millions of small accounts are the business.

Localisation is the other expensive habit the bootstrap years taught. Zakaitis says translating a site is not enough. Words and even emojis move. Support scores once fell in a market until the local team explained that people would not pick the heart-eyes emoji as a positive rating, because it read closer to “I love you.” Pricing has the same trap. Converting €20 into rupees is not a local plan if the resulting package is about 30 percent of average monthly income.

The Inbox Is Where Sovereignty Shows Up

Asked about Europe’s digital sovereignty debate, Zakaitis does not start with model weights. He starts with email. There are not that many European inbox providers, he says, and more customers now ask where their mail will be stored, because it is some of their most personal data. For Hostinger, sovereignty is a storage question on everyday services, not only an AI-infrastructure argument.

That fits the same pattern as OpenClaw. The customer does not want to become a systems person. They want to know who holds the mail, who holds the agent token, and whether the box will still be there next year. Hostinger’s pitch is that it will answer those questions so the user does not have to.

Zakaitis also says he has little idea what AI will look like in five years. He is sure demand will remain for simple ways to start and run a business. The product roadmap follows that bet: overnight advice, competitor notes, agents on a VPS, and a builder that is supposed to keep working after the first page goes live.

A Month Later, the Hour-Long Tool Sits Unused

Two surprises have stuck with him. The technology moved faster than he would have guessed three years ago. People and companies adopted it much more slowly than he would have guessed. User tests inside Hostinger keep showing the same stall. Staff can hand someone a tool and say the task will take less than an hour, often minutes.

If you’d asked me three years ago to describe where AI would be today, I probably would have underestimated the capabilities of the technology but massively overestimated how much people and companies would actually be using it. People are creatures of habit.

Giedrius Zakaitis, CEO, Hostinger

A month later, many still have not used it, because they still picture a week of work or a hired developer. “The technological frontier has moved, but people’s mindset hasn’t necessarily moved with it,” he said. The 21-day contest, the OpenClaw catalog, and the merged AI Builder are all attempts to close that gap in public, on the cheapest accounts, without a venture board asking why the company has not gone upmarket.

People still think they need a developer. They still imagine a week for a job that can take less than an hour. Hostinger’s growth says plenty of them are paying anyway. The unused hour in the test lab says the next million users will not arrive just because the build button got faster.