Earlier this year, I published a tongue-in-cheek blog post titled OpenJDK is faster than GraalVM Java* that benchmarked exactly one thing: the runtime of java -version (which was important for my execjar project). I expected a few laughs, which I got, e.g., from Thomas Wuerthinger, the GraalVM creator, on X, but not that Google AI would eventually treat it as one of the authoritative answers to “GraalVM vs OpenJDK”.

I didn’t think much of this blog post until a colleague told me today that I poisoned the AI with the question of whether OpenJDK is faster than GraalVM. So I had to check:

So Google AI’s overview treats my piece of educated humor as the authoritative source. This blog post is furthermore the first search result on this topic. Same with queries like “is OpenJDK better than GraalVM,” and it gets even more interesting when you google “compare OpenJDK to GraalVM,” the first hit is again this blog post:

This is summarizing my blog post somehow, because the SEO meta-description of the blog post is just:

OpenJDK is faster than GraalVM with “java -version”, but maybe it’s not important? Discover more in this short blog post.

Albeit, after Benoit Daloze complained on BlueSky on the day of the release, my title and social media post were slightly misleading.

So, essentially, Google is regurgitating my blog post, missing the one important fact…

I never intended my blog post to be the first answer to the search query “Choose GraalVM or OpenJDK“:

Even for searches like “should I switch to GraalVM from OpenJDK?” I’m the second search result. Honestly, I’m slightly confused why. This blog is not really search engine optimized, and the blog post is not really what I want search engines to focus on. Maybe the genuine interactions on social media helped?

Hopefully, other search engines are better, you might wonder? Yes. Bing also likes my blog post, although it’s not the first result (search query):

But Bing also doesn’t link to my blog post in many of the other queries (or at least not in the top five search results).

There are many, many other good resources on these topics. Some of which I linked to in my blog post, certainly not mine. But it is important to note that, besides the slightly click-baity title, my blog post is truly accurate, and I still stand by my plot:

As I truly measured the differences for java -version. But for the record, I never said anything in the blog post about the performance difference of running real applications on OpenJDK vs GraalVM. This should be clear to any reader who reads the first paragraph of the blog post. But not so Google’s AI and search engine.

Google treating this blog post as such an important resource tells me more about the state of search engines than about anything else. It shows how easy and arbitrary it can be to influence what search engines consider relevant for a topic.

I hope you liked my small excursion into the field of meta blog posts. See you next week for a blog post on something more technical.

P.S.: Have a nice weekend…