This project marked one of the first steps in adopting AI within Hostinger’s products. What began as a Hackathon exploration evolved into a solution for a critical challenge faced by WordPress users: the struggle to start writing content. Many users abandoned their websites after feeling overwhelmed, ultimately leading to churn.
By integrating LLM-powered assistance, we enabled users to overcome this barrier with ease. The outcome exceeded expectations—WordPress plugin usage reached an all-time high as users immediately recognized the value of the feature.
In this article, I share how our team defined the MVP, measured impact, iterated based on feedback, and transformed an experimental idea into a tangible feature that delivered real value for both users and the business.
Project Background
In Q3 2023, Hostinger began to shift its focus toward leveraging AI to make website management simpler and more accessible for all users. As part of this initiative, our team explored opportunities where AI—particularly large language models (LLMs)—could address user pain points and create tangible value.
Through data analysis and behavioral insights, we noticed a concerning trend: users from the Learners and WordPress Newbies segments often became inactive shortly after setting up their websites. This drop-off raised important questions about what was blocking them from moving forward.
We formed several hypotheses to explain this behavior:
Deployment challenges: Some users struggled during the initial setup and didn’t know what to do next.
Mismatch in expectations: Others successfully deployed their site, but the outcome did not match their vision, leading to frustration.
Content creation barrier: Even for users with a beautiful, functional site, many lost motivation when it came to writing and publishing their first pieces of content.
This last point stood out as one of the most critical barriers. Despite completing the technical setup, users stalled at the very moment where they were supposed to bring their websites to life. We saw this as a clear opportunity for AI to step in—not by building the site itself, but by helping users take the first and most intimidating step: Creating a Content.
In summary,
As a WordPress Newbie, I need an easier way to start writing a content, so I can publish posts confidently and stay motivated to keep building my website.
Building the MVP (v0)
To validate our hypothesis quickly, the team decided to build a lightweight AI Content Creator feature directly inside the Hostinger WordPress plugin. Instead of over-engineering the solution, we focused on testing the core assumption: would users adopt AI assistance to overcome the initial struggle of starting a blog post?
We scoped the MVP to be as lean as possible so it could be shipped in less than a sprint. The initial flow was intentionally simple:
The user writes a short description or topic idea.
The AI generates a draft article.
The draft is automatically saved to the user’s WordPress post drafts.
After releasing the MVP, adoption was surprisingly high, validating that users saw immediate value in the feature. However, user feedback revealed several key usability gaps:
Lack of feedback: Users weren’t sure whether the generation process had completed.
Poor discoverability: Some users didn’t know where to find the generated draft in the dashboard.
Mismatch in tone: A portion of users felt the generated content didn’t align with their personal branding.
Despite these shortcomings,
This experiment successfully demonstrated clear demand for AI-assisted content creation.
The insights gathered gave us a strong foundation to refine the feature, ensuring it not only worked but truly fit into users’ workflows and expectations.
Post-MVP Iteration (v1)
Once the MVP validated our hypothesis and confirmed user demand, the next step was to refine the experience and address the usability gaps uncovered during the first release. Our goal was to ensure that the AI Content Creator evolved from a “quick win” into a truly reliable feature that both beginners and advanced users could integrate into their workflows.
Based on the initial feedback, we planned a set of key enhancements:
Redesign the page with scalability in mind: Creating a stronger foundation to support both current and future improvements.
Preview before publishing: Allowing users to review generated content before it was placed in the WordPress Posts menu.
Guidance and next steps: Clearly showing what users could do with the draft (e.g., edit or publish directly in the WP Editor).
CSAT integration: Adding a lightweight survey after each generation to continuously evaluate AI output quality.
Personalization controls (lower priority): Giving users options to adjust content parameters so they didn’t feel “stuck” starting from a blank page.
Improved Flow
End Result
This iteration not only improved clarity and usability but also attracted new user segments beyond beginners. What initially started as a “starter content” helper began to show value for professional users, who were looking for efficiency and deeper integration into their publishing process.
Key learnings from this iteration included:
SEO expectations: Professional users expected the generated content to be optimized for search engines.
Tone flexibility: Users wanted the ability to mix tones (e.g., professional + friendly) for better alignment with personal branding.
Rich media integration: There was growing interest in adding AI-generated images alongside text to save even more time.
This phase marked the transition from a simple experiment to a feature with clear potential for long-term growth, shaping the roadmap for further iterations and refinement.
Post-MVP Iteration (v1.5)
Building on the success of the earlier version, we continued to refine the AI Content Creator by introducing features that responded directly to user feedback while retaining everything that worked well from previous iterations. The focus for this stage was to move the feature beyond “starter content” into a tool that could support more professional publishing needs.
The enhancements introduced in this version included:
Multiple tone selection: Giving users more flexibility to tailor the voice of the content to their personal or brand style.
SEO-focused generation: Automatically creating SEO keywords and meta descriptions alongside the article draft, helping users optimize their posts more efficiently.
Image integration: Adding relevant images from Unsplash as a quick-win solution to complement the generated text, addressing the common request for richer media support.
Product creation: Supporting a product post creation workflow for WooCommerce websites, helping merchants quickly set up product descriptions with AI assistance.
End Result
This iteration represented a significant step forward in aligning the feature with the expectations of both beginners and professional users. However, it also marked the temporary conclusion of the agile development cycle, as the team shifted focus to other high-priority projects.
Even so, we ensured continuity by implementing robust tracking and CSAT collection. This allowed us to keep gathering insights on how users were engaging with the feature and what improvements would be most valuable when the team resumed development.
Redesign and Reimagine
As the product matured and development resources became available again, I took the initiative to redesign the AI Content Creator interface. By this stage, the feature was being used by multiple personas with different expectations—ranging from beginners who valued simplicity to professionals who demanded more control and flexibility. My challenge was to create a design that balanced both ends of this spectrum without disrupting the established flow or requiring significant redevelopment effort.
The redesign process was guided by a clear set of goals:
Streamline the input process: Reduce friction and make starting with AI as effortless as possible.
Enhance transparency in generation: Provide clear status indicators and progress feedback, following the UX principle of recognition over recall so users could understand system states at a glance.
Improve preview usability: Refine the layout to make generated content easier to scan, while ensuring users could still access detailed editing controls when needed.
Ensure accessibility and responsiveness: Adapt the design to work seamlessly across different devices and screen sizes, supporting a diverse user base.
Final Result
This redesign shifted the feature from a functional tool into a more polished and user-friendly experience, capable of serving both entry-level users and advanced creators. By keeping the core flow intact while elevating usability and accessibility, the redesign not only supported existing adoption but also laid the foundation for future scalability.
Exploring Another Realm
After establishing the AI Content Creator as a plugin within the WordPress admin dashboard, we began exploring ways to bring the feature closer to where users actually work. The idea was to embed AI assistance directly into core WordPress flows—such as the Block Editor and Add Product functionality—so users could generate or refine content without switching contexts.
The main design challenge was ensuring that the integration felt seamless and natural, blending into existing WordPress interfaces rather than appearing as an external add-on. To address this, I conducted exploratory research into how users typically create and edit content in the Block Editor, as well as how they add and manage products within their stores.
From these explorations, several key insights emerged:
Beyond creation: Users didn’t just want to generate new content—they also wanted AI to help edit, refine, and polish existing text.
Seamless interactions: The AI needed to feel like a natural extension of the editor. Whether calling up a block, editing inline text, or tweaking product details, the interaction should be as effortless as existing WordPress functions.
Context awareness: Each menu had its own workflow and conventions, meaning the AI integration had to adapt to those contexts instead of forcing users into a new way of working.
Design Explorations
Block Editor: Create a new post
Block Editor: Edit with AI
WooCommerce: Add Product
Although these explorations were produced shortly before my layoff, they represent an important forward-looking direction for the product: embedding AI directly into users’ primary workflows to make it not just a supporting tool, but a natural part of their day-to-day content creation and product management.
Impact
The introduction and evolution of the AI Content Generator significantly improved the user experience for WordPress beginners and professionals alike. By lowering the barrier to starting content creation, the tool helped users overcome one of the most common drop-off points—staring at a blank editor with no clear next step. The result was a more engaging and satisfying experience, reflected in positive feedback and continued adoption.
Beyond just solving the initial writing struggle, the iterative improvements—such as tone personalization, SEO-ready content, and media integration—transformed the feature into a valuable tool for ongoing content management, not just first-time setup. Usage remained consistently strong, and the fact that the feature is still in active use today highlights its long-term relevance and impact.
Importantly, this project also became a stepping stone for Hostinger’s broader AI adoption strategy. The success of the AI Content Generator informed and inspired the development of a more advanced AI WordPress agent, designed to support even more tasks and workflows directly within the platform.
This journey shows how a small, hackathon-born experiment evolved into a core product capability, shaping both user satisfaction and the company’s AI roadmap.
Retrospective and Things To Learn
Looking back, the AI Content Creator project taught us an important lesson: sometimes it’s better to experiment, execute, and iterate quickly rather than spend too much time trying to perfect the plan upfront.
By shipping the MVP early, we were able to validate the idea directly with real users instead of relying only on assumptions. Each iteration gave us concrete feedback—from CSAT scores, support tickets, and usage tracking—that we could act on immediately. This allowed us to introduce impactful improvements like tone selection, WooCommerce product post support, and image integration in a matter of weeks, not months.
If we had waited until we had a “perfect” plan before release, we might have missed the momentum and user excitement around the feature. Instead, our rapid cycles gave users value right away, while also helping us learn what mattered most to them. In the end, this approach not only shaped the product into something more aligned with user needs, but also laid the foundation for bigger initiatives like Hostinger’s AI WordPress Agent.
The project reinforced that in product design and development, progress often comes from building, testing, and improving in real-world conditions—rather than aiming for perfection on paper.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M 2.857 0 C 1.281 0 0 1.281 0 2.857 L 0 17.143 C 0 18.719 1.281 20 2.857 20 L 17.143 20 C 18.719 20 20 18.719 20 17.143 L 20 2.857 C 20 1.281 18.719 0 17.143 0 Z M 3.08 7.598 L 6.049 7.598 L 6.049 17.143 L 3.08 17.143 Z M 6.281 4.576 C 6.281 5.527 5.513 6.295 4.563 6.295 C 3.612 6.295 2.844 5.527 2.844 4.576 C 2.844 3.625 3.612 2.857 4.563 2.857 C 5.513 2.857 6.281 3.625 6.281 4.576 Z M 14.192 17.143 L 14.192 12.5 C 14.192 11.393 14.17 9.969 12.652 9.969 C 11.107 9.969 10.871 11.174 10.871 12.42 L 10.871 17.143 L 7.906 17.143 L 7.906 7.598 L 10.75 7.598 L 10.75 8.902 L 10.79 8.902 C 11.188 8.152 12.156 7.362 13.598 7.362 C 16.598 7.362 17.156 9.339 17.156 11.911 L 17.156 17.143 Z" fill="rgb(0, 0, 0)" height="20px" id="nI7FUoXPE" transform="translate(2 2)" width="20px"/></svg>)
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M 9.3 0 C 4.163 0 0 4.163 0 9.3 C 0 14.438 4.163 18.6 9.3 18.6 C 14.438 18.6 18.6 14.438 18.6 9.3 C 18.6 4.163 14.438 0 9.3 0 Z M 15.45 4.29 C 16.556 5.64 17.227 7.365 17.242 9.24 C 16.98 9.184 14.355 8.651 11.711 8.985 C 11.494 8.46 11.291 7.995 11.014 7.425 C 13.95 6.225 15.281 4.519 15.45 4.294 Z M 14.565 3.371 C 14.422 3.574 13.226 5.182 10.402 6.24 C 9.101 3.847 7.657 1.882 7.44 1.59 C 9.96 0.982 12.615 1.639 14.569 3.371 Z M 5.921 2.122 C 6.131 2.411 7.549 4.376 8.865 6.716 C 5.149 7.702 1.875 7.687 1.522 7.684 C 2.036 5.22 3.697 3.169 5.921 2.122 Z M 1.357 9.311 C 1.357 9.229 1.357 9.15 1.361 9.067 C 1.71 9.075 5.557 9.124 9.525 7.939 C 9.754 8.385 9.971 8.835 10.17 9.285 C 7.297 10.095 4.687 12.416 3.401 14.621 C 2.134 13.211 1.357 11.351 1.357 9.311 Z M 4.425 15.577 C 5.254 13.882 7.507 11.692 10.71 10.597 C 11.824 13.496 12.285 15.926 12.405 16.62 C 9.851 17.707 6.78 17.411 4.425 15.574 Z M 13.74 15.896 C 13.657 15.412 13.237 13.087 12.195 10.234 C 14.685 9.836 16.871 10.489 17.141 10.575 C 16.789 12.784 15.517 14.692 13.736 15.9 Z" fill="rgb(0, 0, 0)" height="18.599999999999994px" id="FDJfVDwST" transform="translate(2.7 2.7)" width="18.599999999999994px"/></svg>)
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M 13.852 0 C 15.48 0 16.8 1.32 16.8 2.947 L 16.8 6.09 C 16.729 6.086 16.657 6.082 16.586 6.082 L 16.571 6.082 C 16.196 6.082 15.735 6.172 15.405 6.337 C 15.03 6.51 14.704 6.769 14.43 7.11 C 13.987 7.657 13.721 8.396 13.657 9.225 C 13.654 9.251 13.654 9.274 13.65 9.3 C 13.646 9.326 13.646 9.345 13.646 9.371 C 13.642 9.416 13.642 9.461 13.642 9.506 C 13.642 9.577 13.639 9.649 13.642 9.724 C 13.687 11.602 14.7 13.106 16.504 13.106 C 16.605 13.106 16.702 13.102 16.8 13.091 L 16.8 13.856 C 16.8 15.484 15.48 16.804 13.852 16.804 L 2.947 16.8 C 1.32 16.8 0 15.48 0 13.852 L 0 2.947 C 0 1.32 1.32 0 2.947 0 Z M 3.086 4.009 L 3.097 4.013 C 3.592 4.125 3.84 4.29 3.84 4.89 L 3.84 11.91 C 3.84 12.51 3.589 12.675 3.094 12.788 L 3.082 12.791 L 3.082 12.896 L 5.062 12.896 L 5.062 12.791 L 5.051 12.788 C 4.556 12.675 4.305 12.51 4.305 11.91 L 4.305 5.299 L 7.534 12.893 L 7.717 12.893 L 11.04 5.085 L 11.04 12.083 C 10.999 12.555 10.747 12.701 10.301 12.803 L 10.29 12.806 L 10.29 12.908 L 13.736 12.908 L 13.736 12.806 L 13.725 12.803 C 13.279 12.701 13.024 12.555 12.979 12.083 L 12.975 4.89 L 12.979 4.89 C 12.979 4.29 13.23 4.125 13.725 4.013 L 13.736 4.009 L 13.736 3.908 L 11.029 3.908 L 8.516 9.81 L 6.004 3.908 L 3.086 3.908 Z M 16.8 11.561 C 15.859 11.284 15.188 10.245 15.255 9.019 L 15.255 9.019 L 16.796 9.019 L 16.796 11.561 Z M 16.56 6.476 C 16.646 6.476 16.725 6.487 16.8 6.51 L 16.8 8.662 L 15.293 8.662 C 15.349 7.402 15.803 6.491 16.56 6.476 Z" fill="rgb(0, 0, 0)" height="16.803750000000008px" id="hJbYoYIUX" transform="translate(3.6 3.6)" width="16.80000301949815px"/></svg>)
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M 1.8 0 C 0.806 0 0 0.806 0 1.8 C 0 2.366 0.266 2.899 0.72 3.24 L 8.52 9.09 C 9.161 9.57 10.039 9.57 10.68 9.09 L 18.48 3.24 C 18.934 2.899 19.2 2.366 19.2 1.8 C 19.2 0.806 18.394 0 17.4 0 Z M 0 4.95 L 0 12 C 0 13.324 1.076 14.4 2.4 14.4 L 16.8 14.4 C 18.124 14.4 19.2 13.324 19.2 12 L 19.2 4.95 L 11.76 10.53 C 10.481 11.49 8.719 11.49 7.44 10.53 Z" fill="rgb(0, 0, 0)" height="14.400003051757778px" id="cPCyOUUsV" transform="translate(2.4 4.8)" width="19.199999999999996px"/></svg>)