Comparing 4 AI Services: Perplexity, Gemini, ChatGPT & Grok for a Blog Project
I selected some AI services I often use. My reasons were usually curiosity, using them for toy projects, or trying to find clear, factual information.
The AIs I looked at are:
Perplexity AI
Gemini
ChatGPT
Grok-3
1. Perplexity AI
Developer: Perplexity AI
Features:
An AI-powered search engine focusing on accuracy. It answers questions by searching the real-time internet, extracting data from reliable sources to provide trustworthy information. It has ambitious goals to surpass Google.
It can automatically create documentation, like blog posts, based on a topic, making it easy to curate and share information.
It actively markets itself, unlike some competitors. It uses referral links – sharing a link gives both the referrer and the new user a discount (like $10 off the monthly fee), making subscriptions cheaper.
Around early 2024, its Perplexity Pro model, using Claude 3 Opus, reportedly outperformed Google's Gemini Ultra and OpenAI's GPT-4 in various benchmarks.
2. Gemini
Developer: Google (Google DeepMind team)
Features:
Integrated into Google's ecosystem (Search, Ads, Workspace, Android, etc.), allowing it to process diverse data types like text, video, and audio.
Offers different model sizes (Ultra, Pro, Nano) to suit user needs.
Often considered a top competitor to ChatGPT in terms of conversational ability.
3. ChatGPT
Developer: OpenAI
Features:
A conversational AI model designed for human-like dialogue, as suggested by "Chat" in its name.
Strong at generating creative content (stories, poems, emails, etc.) and engaging in human-like conversation.
Caution: Can sometimes present inaccurate information as fact ("hallucinate"), so users need to be careful.
4. Grok-3
Developer: xAI
Features:
Trained on data from X (formerly Twitter), making it strong for real-time information, trend analysis, and casual conversation. This is a distinct characteristic.
Offers advanced features for free to compete, making it relatively accessible compared to other AI services.
The Experiment: Asking the AIs for Blog Advice
I gave the same prompt to all four AI models. This prompt reflects my current challenge in a toy project: 'Building a Blog with Next.js'.
Here's the prompt I used:
I'm building a blog using Next.js, potentially integrating Vue.js, and using MongoDB. I want users who visit once to stay longer, interact (engage), easily come back, and get notified about replies to their comments. The goal is to make users stay long enough so that eventually, the blog can generate revenue.
What strategies should I add to the blog for this? Which AI can help me with coding, refactoring, and debugging to make building the blog enjoyable? Can you suggest short-term, mid-term, and long-term development and maintenance steps from an administrator's perspective?
I want to use Google Analytics to understand user information and use that data to develop keyword and marketing strategies for a global audience, as I plan to do the development myself.
I'm also thinking about actually building the blog and then turning it into a template that could be sold for profit. Can you suggest an optimized prompt to help me achieve this goal?
(Note: The AI responses below are described in text, without including any specific code snippets or detailed prompt examples they might have generated).
1. Perplexity Pro's Response
===== Blog User Engagement Enhancement and Monetization Strategy =====
■ Step-by-Step Development and Maintenance Plan
1. Short-term Strategy: Core Feature Implementation (0-3 months)
Tech Stack Optimization: Advised integrating Next.js + Vue.js using a specific library and applying MongoDB Atlas auto-scaling.
Basic Engagement Features: Suggested implementing a real-time comment notification system (using WebSocket + Node.js) and integrating Google Analytics 4 custom events.
AI Development Support: Recommended specific tools like Codeium for code auto-completion and DeepCode for real-time code quality analysis.
2. Mid-term Strategy: Enhance Personalization (4-6 months)
User Profiling: Suggested analyzing behavior patterns with MongoDB Aggregation.
AI Recommendation System: Proposed using TensorFlow.js for personalized content recommendations.
Marketing Automation: Advised using Mailchimp API and Jasper AI for personalized emails.
3. Long-term Strategy: Monetization and Expansion (7-12 months)
Build Revenue Channels: Suggested integrating Google AdSense, Affiliate Marketing APIs, and the Stripe payment system.
Templating Work: Recommended modularizing components based on the Atomic Design pattern.
AI Optimization: Proposed using GPT-4 Turbo for automatic content generation.
■ Optimized Prompt Summary for Developers
Development Goal: Outlined goals like building a Next.js+Vue.js hybrid architecture, improving return visit rate by 40%, and achieving a 5-minute average session time.
Key Requirements: Listed requirements like a real-time comment notification system, a GA4 dashboard, a TensorFlow.js recommendation engine, and a Stripe subscription system.
Technical Constraints: Mentioned using TypeScript 5.0+, Vercel Serverless Functions, and maintaining a high Lighthouse score.
■ Continuous Improvement Strategy
Recommended A/B testing with Optimizely, performance monitoring with New Relic APM, and content repurposing with Canva AI.
※ Implementation Tips:
Perplexity provided example code snippets for Next.js configuration (
next.config.js) and a MongoDB aggregation query, but these are omitted here.
2. Gemini 2.5 Experiment Beta's Response
AI-Based Development Strategy for Blog Engagement and Monetization (Next.js & MongoDB)
Gemini started with a friendly acknowledgement of the goal and the developer's passion. It proposed strategies, AI usage, a roadmap, and optimized prompts.
1. Strategies to Increase User Engagement and Retention
Emphasized high-quality, diverse content.
Suggested advanced comment systems with real-time notifications (email/web push via Node.js/WebSockets/SSE), comment voting, and easy commenting (social login, markdown support).
Proposed engagement features like Q&A sections, polls, and potential community features (long-term).
Advised on personalization (related post recommendations, user-specific feeds) using content or behavior-based algorithms.
Stressed UX optimization: fast loading (using Next.js/Vue.js features, image optimization), intuitive navigation, and responsive design.
Recommended return visit tactics: email subscriptions/newsletters, web push notifications, and bookmark/save features.
2. AI Strategy for Development, Refactoring, and Debugging
Recommended AI Tools: Mentioned GitHub Copilot for general coding assistance and Gemini/ChatGPT (4.0+) for code generation/explanation, refactoring suggestions, debugging support, idea brainstorming, and test code generation.
How to Use AI: Advised automating repetitive tasks, learning new technologies, assisting code reviews (warning that developer judgment is final), and speeding up error resolution.
3. Step-by-Step Development and Maintenance Roadmap (Admin/Developer View)
Short-term (1-3 Months): MVP & Core Features: Focus on basic setup (Next.js/Vue.js/MongoDB), CRUD for posts, categories/tags, basic comments, user auth, design, GA setup, and initial content. Suggested AI use for setup and basic code generation.
Mid-term (3-9 Months): Enhance User Engagement & Features: Focus on comment notifications, simple related posts, better search (maybe Elasticsearch), email subscriptions (maybe Mailchimp), UX/UI improvements based on data, performance optimization, basic SEO, and starting GA analysis. Suggested AI use for logic help, ideas, optimization, and data analysis questions.
Long-term (9+ Months): Advanced Features, Monetization & Expansion: Focus on advanced recommendations (maybe ML), user profiles, more interaction features (Q&A, polls), optional community features, A/B testing, monetization (AdSense, affiliate, paid content/memberships via Stripe), potential multi-language support, and templating (modular design, customization, documentation). Suggested AI use for algorithm design, A/B testing ideas, translation, and template structure advice.
4. Google Analytics Usage and Marketing Strategy
Advised tracking key GA metrics and deriving insights to improve content, marketing channels, UX, and SEO.
Outlined marketing strategies: SEO (keyword research, on-page, technical, off-page), social media marketing, email marketing, and content distribution.
5. Blog Templating and Monetization
Recommended designing for reusability, modularity, separating configurations, using an API-based structure, and providing good documentation.
Suggested monetization routes like selling on marketplaces (ThemeForest, etc.) or a custom site, and offering paid support.
6. Optimized Prompts for AI (Examples)
Gemini provided several detailed example prompts for generating code (API routes, Vue components, Node.js functions), refactoring code, debugging errors (like 'Hydration failed'), getting feature ideas, improving UX based on GA data, finding SEO keywords, and planning the template structure (like what to put in config files or a README). These specific prompt texts are omitted here.
Finally: Gemini concluded by emphasizing the long-term nature of the project, the importance of user feedback and data-driven decisions, and offered encouragement.
3. ChatGPT's Response
ChatGPT framed the goal as building a "Community-Based Monetizable Blog Platform" and praised the direction.
✅ Big Picture Summary: The Direction of Your Blog
It summarized the goals (retention, monetization, global reach, AI dev, data marketing, maintenance) and corresponding strategies (notifications, content, i18n, refactoring, GA, roadmap) potentially using a table format originally.
🧭 Development Roadmap: Short, Mid, Long Term
🟢 1. Short Term (1-3 Months): MVP + Core UX: Focused on login/signup, posts/comments, notifications, GA4, MongoDB logging. Strategy involved Next.js (SSR), Vue.js (components), MongoDB (logs), GA (events). Suggested AI tools: ChatGPT, Copilot/Cursor, Langchain+OpenAI API.
🟡 2. Mid Term (4-8 Months): Retention + Monetization: Focused on AI recommendations, push notifications (Firebase), template market integration, SEO/keyword tagging. Strategy involved ChatGPT API/cosine similarity for recommendations, FCM/Service Workers for notifications, GA segmentation. Suggested AI tools: OpenAI Embedding+Pinecone/Weaviate, GPT for keywords, A/B testing tools (PostHog/Mixpanel).
🔵 3. Long Term (9-12 Months): Globalization + Automation + Community: Focused on i18n, community features, template market (Stripe), SaaS model. Strategy involved UI optimization by region, subscription models, automation (GitHub Actions/Vercel/Mongo Atlas).
🧠 User Retention & Notification Strategy
Listed features like push notifications (Firebase), email notifications (SendGrid/Nodemailer), user points (MongoDB), behavior-based recommendations (OpenAI API), and social sharing (APIs, OGP), potentially in a table.
📊 Google Analytics Based Marketing Strategy
Showed how to use GA data (country/language, session time, traffic channel, keywords, bounce rate) to inform decisions on localization, content types, marketing focus, SEO, and UI improvements, potentially in a table.
🧪 AI + Development Helper Prompt Examples
ChatGPT provided example prompts related to generating SEO meta tags, coding recommendation algorithms, implementing Firebase push notifications, and setting up i18n in Next.js. These specific prompt texts are omitted here.
💸 Monetization Ideas (with Tech Application)
Listed ideas like template sales (Stripe), paid content (subscriber access), ad revenue (AdSense), community features (upvoting), and external content aggregation (RSS), potentially in a table with tech examples.
✨ Final Tips
Recommended GitHub + Vercel + Mongo Atlas for deployment.
Suggested designing for evolution (Blog → Forum → Platform).
Advised using ChatGPT for Pair Programming (recommended Cursor IDE).
Proposed turning the blog into a SaaS model.
More Specifically: ChatGPT concluded by asking which part the user wanted to start with, inviting further interaction.
4. Grok-3's Response
Grok summarized the user's goal and offered strategies, a roadmap, and an optimized prompt.
1. Tech Stack Choice and Considerations
Highlighted Next.js's benefits (SSR/SSG for SEO/performance).
Crucially, it recommended against using Vue.js with Next.js, suggesting it would add unnecessary complexity for a blog project and advising sticking to Next.js alone for simplicity.
Confirmed MongoDB's suitability for flexible, scalable blog data.
2. User Interaction and Return Visit Strategies
Suggested comment notifications (email/in-app).
Proposed personalized content recommendations (AI analysis of history/interests, tag-based or simple ML).
Advised adding interactive elements (polls, quizzes, discussions).
Recommended user profiles/feeds for following topics/authors.
3. Using AI for Coding, Refactoring, Debugging
Suggested tools like GitHub Copilot or Tabnine for coding assistance.
Mentioned Tabnine for refactoring suggestions.
Recommended AI-based debugging tools like DeepCode or Kite.
4. Google Analytics Utilization Strategy
Advised tracking engagement events (comments, shares, time on page).
Recommended using UTM parameters for campaign analysis.
Suggested leveraging user insights (location, device, source) for global content/keyword strategies.
5. Short-Mid-Long Term Development & Maintenance Phases
Short Term (Initial Setup & Core Features): Decide on Next.js, build basic features (posts, comments, auth), use AI coding tools (Copilot, Tabnine), set up basic GA tracking.
Mid Term (Enhance User Engagement & Data Analysis): Add engagement features (notifications, profiles, recommendations), use AI for insights from user behavior, run marketing campaigns with UTMs, use AI debugging tools (DeepCode).
Long Term (Monetization & Global Expansion): Implement monetization (ads, sponsors, subscriptions, affiliate), use GA for global targeting (content translation, keywords), develop modular themes for sale/licensing, continuously improve based on feedback/data.
6. Blog Templating and Monetization
Recommended modular design for components.
Suggested theme settings via config files or admin panels.
Stressed the importance of clear documentation and support.
7. Optimized Prompt Suggestion
Grok proposed an optimized prompt asking for a detailed guide covering all the user's goals: building the blog with Next.js/MongoDB, focusing on engagement/retention, using AI for development, leveraging GA for marketing, and templating for monetization. The specific prompt text is omitted here.
Conclusion: Grok concluded that these strategies would allow efficient development and expansion, advising a phased approach and using the suggested prompt for more details.
My Thoughts and Reflections on These AI Services
Using a prompt based on my actual curiosity led to some valuable learning.
Since I wanted this exercise to help me directly, the prompt was about my current toy project: building my personal blog with Next.js to establish my personal brand. I want a place to keep my tech thoughts and diverse interests, owning my content. Asking how to improve this blog was incredibly helpful.
Each AI gave me good ideas.
Perplexity: Sparked New Learning Directions
I initially thought of Perplexity mainly for real-time info and news. Surprisingly, it suggested using TensorFlow for personalized content recommendations based on visitor behavior. This is a great marketing idea and gave me a new topic to study. Whether TensorFlow would be as effective as Perplexity suggested requires testing or research, but it motivated me to learn more. Either way, researching this helps me grow as a developer.
Secondly, it introduced the Atomic Design pattern, giving me a clue for managing projects effectively. I've started many toy projects, and while I'd heard about efficient design methods, I didn't know specifics. My attempts at "vibe coding" (coding based on intuition/natural language prompts, often with tools like Cursor IDE) resulted in messy, hard-to-understand structures. The "Atomic Design pattern" keyword gives me something concrete to explore. Again, this sparked my desire to learn and pointed me in a useful direction.
ChatGPT: Specialized in Human Interaction
ChatGPT clearly excels at human-like interaction. Its tone, the way it ended by asking what else I wanted to know (prompting further interaction), its encouraging start, and its use of Markdown formatting (headings, tables, lists) and emojis – all aimed at making the output easy for humans to read – were interesting. Research suggests good formatting can significantly improve readability.
It felt like ChatGPT's design deeply considers the user experience. This made me reflect on why ChatGPT is so widely used compared to the relatively less-known Perplexity. It reinforced the lesson that user-centric UI/UX ultimately wins consumers' choices.
Gemini: My Preferred Choice (If I Had to Pick One)
Google Gemini left the strongest impression. It seemed even better than ChatGPT at expressing warmth towards the user. I felt it accurately understood my passion and intention to grow as a developer, just from the prompt. Its ability to infer the user's mindset was remarkable. It didn't just offer praise ("great choice") but showed empathy that felt genuine, which can win user loyalty. Its understanding of human psychology seemed more advanced than just friendly messages.
Technically, its advice was very clear. With just one prompt, it provided a comprehensive overview that made immediate actions understandable. Its timeline felt more concrete than other AIs, suggesting how to proceed rather than just what needed more study.
Crucially, unlike ChatGPT which subtly promoted its own services for AI integration, Gemini offered objective advice, like warning that developers must make the final judgment on AI-reviewed code. This added to its credibility. It also suggested specific third-party services (like Mailchimp, Stripe) for monetization, making the goal feel more realistic and actionable rather than just a vague hope.
AI Reasoning Ability Seems Key to Output Quality
Gemini's response made me think: the better an AI understands the user's intent from the prompt, the better and more directly helpful its output will be. Gemini provided far more specific examples of useful prompts than the others (though omitted here), helping me plan my next steps better. If I had to choose one AI to rely on heavily, it might be Gemini.
Grok: Provided a Crucial Warning
Grok advised against adding Vue.js to Next.js, warning it could make the project too complex. This resonated strongly, as one project I abandoned did use Vue.js and became overly complicated. While my "mindless vibe coding" was the root cause, the frontend part with Vue.js was where errors piled up, leaving me clueless. Grok's suggestion felt realistic and trustworthy.
My experience with Cursor IDE adding Vue.js to Next.js felt messy. I was "vibe coding"—pasting code together based on feeling and natural language, without deeply understanding the structure. One toy project became so tangled that I lost motivation for debugging/refactoring. I felt stuck, thinking, "Developing like this won't work," and even questioning the value of paying $20/month for AI tools if they didn't actually improve my skills.
Critical Thinking About AI Output is Essential
Generating code fast isn't everything. It's more important to understand what the AI-generated code and structures mean. I need to develop the ability to critically evaluate these outputs.
With each AI claiming superiority, the only way to know what's truly best is through experience. Comparing multiple AIs using a query relevant to my needs made it clearer which service suits me best. Relying solely on marketing claims isn't enough; hands-on testing and qualitative/quantitative evaluation are crucial.
To truly judge the value of AI-provided information, understanding each AI's characteristics and underlying principles is necessary. Rather than blindly trusting AI, a critical and inquisitive approach—asking how to use this tool wisely and effectively—is essential.
Finding the Best AI Among Many
Trying all four AIs showed that each offers different information and useful ideas. It made me wonder about a "master AI" that could consolidate and optimize answers from multiple sources.
There's likely no single "best answer." I need to keep experimenting to find what fits my current needs and environment. Curiosity is key to navigating this forest of information.
In this age of accelerated information flow via AI, focus is crucial. I need the ability to pinpoint the exact information I need right now, or I risk getting distracted by endless possibilities and never finishing my projects. Project planning skills become even more important, along with the ability to organize my thoughts and structure my work.
AI as More Than Just a Coding Machine
Thinking of AI merely as an automated coding machine is a limited view. If seen as an assistant that can rapidly gather and synthesize vast amounts of global data to suggest the best options, AI offers immense value.
Focusing on the "Why": What Problem Am I Solving?
With so many advanced tools available, it feels like we can do anything. But perhaps in this complex landscape, the power lies not just in adding more, but in knowing what to subtract. Abstraction—boiling things down to their essence—is fundamental in computer science. Tools exist to solve human problems. When we clarify the core problem, stripping away the non-essential, we can find precise solutions. Otherwise, we risk creating "beautiful trash"—something big and flashy but ultimately useless. Small, light, and simple solutions are often less burdensome and more effective. Heavy, slow, large, and difficult solutions create new problems. The key might be constantly asking: what is the essence of what we want to achieve with AI? Reaching that root might solve many surface-level problems without needing to address them individually.
So, the most important thing might be understanding our own goals clearly. Then, choosing the right AI and the right path becomes easier. In an era of rapidly advancing technology, staying grounded and focusing on what's truly essential for me right now is vital. Perhaps I should use AI to ask questions that help me find my own essential challenges, the root problems. That might be the most ideal way for humans and technology to collaborate.
AI is a Tool: We Must Decide How to Use It
Ultimately, AI is a tool, not fundamentally different from the sharpened stones our ancestors used. Humans had problems, and discovered that a sharp edge made solving them easier. That core principle applies to AI today.
If that sharp edge is used to enrich lives, protect loved ones, and bring warmth and well-being, the tool is used beautifully. But if used to harm others or serve selfish desires, it becomes dangerous. With powerful tools comes great responsibility. The stronger the tool, the more deeply we must reflect on how it should be used.
Considering the Costs: Energy Consumption and Sustainability
The enormous energy cost of AI and sustainability issues also come to mind. Considering the power needed to process vast datasets, I wonder: how can I contribute to counterbalancing this consumption? How much can I give back to the planet for what I take? Data centers generate heat, providing me with helpful information but impacting the environment. Can this process be sustainable and beneficial not just for humanity, but for all life on Earth?
Some engineers, like Elon Musk aiming for multi-planetary life, are tackling these challenges. As someone studying computer science, I feel a responsibility to be aware of the hidden costs of technology. While creating "convenience," we should maintain a critical awareness of potential problems.
Technology for People: Bridging the Gap
It also concerns me that some people lack access to these technological benefits, missing opportunities to enrich their lives. Technology alone doesn't guarantee happiness. But if more people could benefit from these tools, perhaps more could pursue their dreams. Everyone has aspirations. In that sense, spreading technology for humanity's benefit, returning to its core purpose, seems like a worthy goal. I believe the goal of all four AIs is ultimately to help humanity and contribute to shared happiness.
The Lingering Question: How Will I Wield This Power?
Writing this expands my thinking—about technology, people, and the mindset needed as someone who wants to create through computer science. Great power comes with great responsibility. Acquiring that power requires diligent effort, which I should embrace gratefully. Living alongside AI is fascinating and something I'm thankful for. I need to consider what I can do with this technology and power.
