Artificial Intelligence has been the talk of the town for years now. At first, people tried to ignore it as simply a corporate buzzword, but its prevalence in our daily lives continues to grow, whether we like it or not.
Why use AI for building a portfolio?
It’s not perfect, but it can help you get started
While AI continues to arguably get smarter and better, it is making a strong case for itself as a valuable tool in certain aspects. AI has become a key tool for so many people to find jobs, track applications, get their resumes and cover letters proofread, or even write them both from scratch.
Beyond that, there’s also another way to use AI to assist you in your career: Making your portfolio. Let’s be honest, it’s a tough job market out there, and if you’re someone from a non-tech background, it can be understandably difficult to build a decent portfolio website from scratch.
There are a lot of free website templates out on the Internet for design, so in most cases, you pick from a common pool of templates that everyone is using. This likely rules out your portfolio standing out from a design perspective, which might be more problematic for those who have a creative and visual portfolio to display. Even if you do find a unique and rare design template, you still need to know basic web design understanding on the front-end to manually add your portfolio and adjust the design according to your personal needs.
That’s where AI can step in, if you want it to.
Google Gemini, ChatGPT, or Claude?
One prompt to test them all.
Once you decide you want AI to simply build your portfolio website, the immediate question is, which one? Google Gemini, Claude, and ChatGPT are the three most popular AI platforms in the world at the moment. So, I tried all three to see which one gave me the best results in a single prompt and got me closest to what I wanted.
Now, my first pick was using Perplexity Pro, as it gives me access to pro versions of ChatGPT, Claude, and Gemini models. However, there are still some limitations applied to these models when using them through Perplexity.
So, to get the full native experience of using each AI model, and without paying for either of them, I had to see what the free AI models would offer.
My prompt for building a portfolio
Minimum effort, maximum expectations
My prompt for all three AI models was the same, and if you fancy trying it out for yourself, here it is:
“Build me a personal portfolio site for a tech journalist. Here’s my bio: [Add bio]
Make the portfolio clean and minimal. Make sure you add multimedia elements, such as feature images of articles alongside the headlines.
Also, add hyperlink to the articles on both the headline, and feature image. Make the design look personal, and it should have the feel of a portfolio that is ready to show potential employees.
Add my contact details such as my email address [specify your email address] and social media account usernames for X (@yourusername) and Instagram (@yourusername). Here are my published articles: [add your published work]”
As you can see, I wanted AI to build my portfolio by displaying my published articles for How-To-Geek. You can easily replace article-specific requests with what is relevant to the content that you want to showcase.
ChatGPT
Playing it safe
I started with ChatGPT, and right off the bat, the AI began to code the HTML. It was almost instant, and it felt like the AI model was waiting with the code locked and loaded to fire off as soon as I requested the portfolio.
I took the entire HTML code, pasted it into a .txt file, and saved it as a .HTML file so that I could open the file on my browser offline. That’s the best option I feel, for those who want a finished product from AI. Once you’re happy with how it looks, you can take the HTML code to whichever domain you are hosting your website on. If you have not purchased a domain for your portfolio, you can use a free hosting site like Vercel.
Now, as you can see from the images, ChatGPT did a decent job. It fundamentally understood the assignment and kept the design clean. However, it was not able to fetch the feature images from the articles, and went with different images in the portfolio. Now, most of the images look relevant to the story, but I don’t know where these images were pulled from, and if these are under Creative Commons or if they are copyrighted. So that’s a key problem to deal with manually. Another thing is the bio, which I did not ask ChatGPT to paraphrase, but it did. However, I would understand the argument that I also did not tell ChatGPT not to paraphrase it.
The contact information is a bit sloppy for my taste, as it does not have any label for email or social media platforms at the top. To be fair, it’s done well right at the bottom in the “Get in Touch” section, so I can simply follow up with ChatGPT to remove it from the top.
Overall, ChatGPT produced a good base to work with, given it was the first prompt. There’s definitely a lot of room for back-and-forth prompts to fix minor stuff. So, there’s potential, and if you have a pro version, you are likely to get better results from the get-go.
Claude
Impressive, and takes several design liberties
When I gave the prompt to Claude, it took its time. Within two minutes, Claude gave the code, and I also got to see how it looks right there on the chat window. I did take the code and made my own .HTML file, so that I can properly see how it looks in a browser.
Claude’s results were worth the wait. Unlike ChatGPT, Claude was able to fetch the feature images, and it kept my story deck intact. ChatGPT summarised each article in its version of the portfolio, which again was a hit or miss and needed manual intervention.
Claude was simply more accurate. The entire design felt like it has some personality to it, the fonts were creative, there was color on the website beyond the images, and the layout of articles flows similarly to the style of How-To-Geek.
Now, Claude also paraphrased the biography section, which would require a follow-up prompt to keep it verbatim. Claude also turned a part of the biography into a pull quote, which is neither accurate nor looks good. Get in Touch section’s paragraph was a nice and welcome change, and the overall layout of this design impressed me.
If you want to use Claude, go for it; it will not disappoint. From the looks of it, it will require even fewer follow-up prompts to make it ready to show in your job applications.
Google Gemini
Discouraging experience and disastrous results
Next up, it was Google Gemini. I’ll be honest, I didn’t have full confidence in Gemini to impress me after seeing the portfolio that both ChatGPT and Claude made.
I was right.
Google Gemini struggled to even get the ball rolling. It took over 10 minutes for Gemini to process the prompt. I kept waiting until I realized this is going to take forever. I refreshed the page and gave the prompt again. I waited a couple of minutes before Gemini gave me an error.
I tried again, and I got the same error again. When I asked Gemini what the issue was, it surprisingly gave me the portfolio code as a reply. Here’s Gemini’s explanation to what happened:
“It looks like there was a temporary technical hiccup on my end while trying to process that extensive list of links and formatting requirements all at once. I apologize for the “broken record” moment!”
For the record, the extensive list of links was 14 articles. Claude and ChatGPT did not find issues dealing with the minimal formatting guidelines I provided in the prompt.
Nevertheless, I copied the code over to see the results, and well, it would have been better if Gemini had asked me for another prompt and more clarity instead of producing the portfolio.
As visible in the gallery, Gemini just did not understand the assignment well enough to provide anything close to what Claude and ChatGPT are capable of. Feature images are missing, there’s hardly any design element to this page, and the less we talk about this, the better.
Claude nailed it, and Gemini needs to up its game
In conclusion, Claude is the clear winner. With a prompt as simple as mine, I was happy to see Claude filling in the gaps in terms of design and giving the portfolio a personality. ChatGPT did nothing wrong, and it can get your portfolio looking how you want it to be, with a few more follow-up prompts.
As for Gemini, the experience was bad, and the result would discourage anyone from even following up and trying to make the AI understand when there are clearly better options available.
