The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
I built an automated German AI publication with Astro — here are the early results
Over the past few months, I’ve been working on a German-language publication focused on artificial intelligence, large language models, AI software, technical explanations, and industry developments.
The project is built with
Astro and supported by a largely automated content and publishing workflow.
I wanted to share some of the early performance data because the differences between Google Search, Bing, and AI-powered search have been especially interesting.
Google Search Console
Across the analyzed period, the website generated:
-109 organic clicks
-20,233 Google Search impressions
-an average CTR of approximately 0.54%
The monthly development looked like this:
February: 7 clicks and 399 impressions
March: 12 clicks and 1,662 impressions
April: 16 clicks and 2,088 impressions
May: 18 clicks and 4,039 impressions
June: 41 clicks and 9,151 impressions
July, partial month: 15 clicks and 2,894 impressions
The most encouraging signal is the growth in visibility.
Google impressions increased from 399 in February to 9,151 in June, which represents growth of more than 22 times within a few months.
Clicks also increased from 7 in February to 41 in June.
The CTR is still relatively low, but this appears to be mainly connected to ranking positions. Many pages are currently receiving impressions while ranking outside the top results, often between the lower first page and the following pages.
That means the next major opportunity is not simply publishing more content. It is improving how existing pages perform.
The most important areas are currently:
-titles and search snippets
-internal linking
-search-intent alignment
-topical authority
-content differentiation
-ranking positions
-click-through rate
Bing organic search
Bing generated:
-39 clicks
-2,347 impressions
-an average CTR of approximately 1.66%
Bing currently provides much less total visibility than Google, but the click-through rate is noticeably higher.
Google offers the larger traffic opportunity, while Bing already appears to send comparatively qualified visitors from a smaller number of impressions.
Microsoft Copilot and Bing AI citations
The most surprising part of the experiment has been the performance in Bing AI and Microsoft Copilot.
AI citations developed as follows:
March: 57 citations
April: 165 citations
May: 212 citations
June: 384 citations
July, partial month: 113 citations
The number of pages cited by Bing AI also increased:
March: 36 cited pages
April: 72 cited pages
May: 82 cited pages
June: 115 cited pages
By June, content from more than 100 pages had already been used as a source by Microsoft’s AI systems.
This is especially interesting because some pages appear to receive AI citations before they achieve strong positions in traditional search.
That suggests AI visibility may develop on a different timeline from classic organic rankings.
Several individual pages also accumulated dozens or even hundreds of citations, particularly content with a strong practical or educational angle.
The formats that currently appear to perform best include:
-tutorials
-technical explanations
-glossary-style content
-implementation guides
-comparison-focused content
-practical problem-solving articles
I am intentionally not sharing the exact topic clusters yet, as part of the project is still experimental.
The “Crawled — currently not indexed” category is the main area I am monitoring. These pages may need more time, stronger internal links, clearer differentiation, or better alignment with search demand.
The 404 pages also need to be reviewed individually to ensure that removed URLs are handled correctly and are no longer referenced internally or included in the sitemap.
Lighthouse and technical performance
The Lighthouse results were:
Performance: 98
Accessibility: 93
Best Practices: 100
SEO: 100
The measured performance metrics included:
First Contentful Paint: 1.2 seconds
Largest Contentful Paint: 1.5 seconds
Total Blocking Time: 0 milliseconds
Cumulative Layout Shift: 0
Speed Index: 4.0 seconds
Astro has been a very strong fit for this type of content platform.
The combination of static rendering, lightweight pages, and limited client-side JavaScript provides an excellent technical foundation.
At this point, site speed is not the main bottleneck. The more important challenges are authority, rankings, CTR, content quality, and distribution.
What I have learned so far
The biggest lesson is that content volume alone is not enough.
A scalable publishing system can create a large number of pages, but that does not automatically lead to traffic.
The real work starts after publication.
Another important observation is that traditional search and AI search do not behave in exactly the same way.
Google appears to test many pages gradually before giving them stronger visibility.
Microsoft Copilot, by contrast, has already cited a surprisingly broad selection of content.
This may indicate that AI search visibility can emerge earlier than meaningful organic traffic.
The project is still early, but the direction is encouraging.
The strongest signals so far are:
Google impressions increased more than 22 times from February to June
Google clicks increased from 7 to 41 per month
Bing organic CTR reached approximately 1.66%
Bing AI citations increased from 57 in March to 384 in June
the number of pages cited by Bing AI increased from 36 to 115
several individual pages accumulated a high number of AI citations
Lighthouse Performance reached 98
SEO and Best Practices both reached 100
the technical foundation remains fast despite the growing amount of content
The main question now is whether growing visibility and AI citations will translate into stronger rankings, more clicks, and sustainable revenue over the coming months.
Would you like to see more insights into Astro, SEO, automated publishing, and AI search visibility?
Follow me for future updates.
You can also find high-quality Astro themes here:
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
AI is making it easier for anyone to create designs.
But not everything should be DIY.
I used the same prompt on Claude, ChatGPT, and Figma Make.
Three different results.
If you’re not a designer, how do you know which one is actually good?
How do you know what hurts usability, accessibility, or conversion?
That’s why professionals exist.
AI is a great starting point, but expertise is what turns an AI-generated design into a product people actually enjoy using.
Even if you start with AI, let an expert refine it before you launch.
👇 Which one would you choose? Claude, ChatGPT, or Figma Make?
Great job 👏
Most of my collages are built from dozens of images that I source and combine in Photoshop. But sometimes the exact element I have in mind simply doesn't exist, or I can't find the right angle, lighting, or composition.
That's where AI comes in.
For this piece, I used Envato's AI Image Generator to create a custom starting point when I couldn't find the image I needed. I then combined it with assets from Envato Elements, using Photoshop to composite everything together, refine the lighting, color grade the scene, and create the final artwork.
Once the collage was complete, I used Envato's AI Video Generator to bring it to life with subtle movement, then finished the animation in Premiere Pro using additional Envato Elements assets.
For me, AI isn't about pressing a button and calling it art. It's another creative tool—one that helps me bridge the gap between the idea in my head and the final artwork.
From AI Slop to Autonomous Pipeline: Drilling for Data with Base44 & Envato 🛢️⚡
The Backstory: The Panic & The Concept
I spend my days building raw, headless, zero-touch B2B data pipelines. But recently, a high-ticket client asked the one question that terrifies every backend builder: "Can I see your portfolio?" I didn't have one. I had 48 hours to prove I wasn't just talking.
I spun up Base44. It generated a structural layout in seconds, which was sick, but the content? Absolute corporate AI slop. Fake statistics, soulless design, empty words. It was a dead skeleton. I needed to override the AI and give it a heartbeat, which led me to the core theme of this project: Drilling.
I grew up in Moreni, Romania—the 3rd place on Earth to extract oil back in 1929. My great-grandparents literally drilled the earth for raw fuel. Today? I drill the modern web for unstructured data. So, I turned the Base44 template into a digital drilling rig. As you scroll down the depth gauge (from 0m to 1420m), you dive deeper into my real, live architectures.
The Envato Glow-Up: Injecting the Soul
A cool metaphor is nothing without the right aesthetic. I wanted the site and the video to feel like a high-tension tech thriller. I hit Envato Elements to completely override the AI's generic look:
The Visuals: I ripped out the plain background and injected a dark, industrial grainy texture via CSS to give it that gritty, underground feel. I swapped the AI placeholders with my actual live work: geospatial heatmaps, running Make.com webhooks, and my NLP LLM-as-a-Judge Streamlit app.
The Sound Design (SFX): I wanted the viewer to feel the pressure. I used Envato for the dark cinematic synthwave track, heavy sub-boom drops for the panic moments, digital error buzzers when the AI failed, and rapid-fire UI clicks.
The Voice: I used a cinematic AI voice generator to narrate the exact moment my ancestors' history connected with my modern tech stack.
The Video: Learning DaVinci Resolve in 5 Hours
Here’s the craziest part of this submission: I didn’t know how to edit videos before this weekend. I wanted the pitch video to be as dynamic as my code, so I literally learned DaVinci Resolve from absolute zero in 5 hours just for this challenge. I built a 4-way split-screen grid and edited a rapid-fire "ladder" sequence of me downloading Envato assets, synced perfectly to mechanical keyboard clicks every 0.2 seconds. It’s raw, fast, and shows exactly how I operate.
The Result
I didn't just build a pretty template. I built a fully functional B2B infrastructure. The contact form at the bottom of the site is hardwired via webhooks straight into my Make.com pipelines.
AI gave me the speed and the skeleton. Envato gave it the aesthetic and the soul. My ancestors drilled for oil. I automate the future.
You can check my page here: https://garrulous-build-logic-flow.base44.app/
Post on SocialMedia X:https://x.com/ydgsgiwhs/status/2077907190224961774?s=20
#HumanByDesign #GiveItAGlow #EnvatoChallenge #BuildInPublic #Base44
Trending
Claude
Claude has entered the design space. How are you using Claude Design?
Contra University
Learn from expert creatives how to earn more using next-gen AI tools.
creativeaiflow
Creative AI workflows are evolving. What tools do you use, and what are their strengths and weaknesses?
freelancerlife
Freelancer life is wins, pivots, and everything in between. What’s yours right now?