StackOverflow Visitor Decline Is *Not* Caused By ChatGPT

Nope, the popular rapport does not proove AI causes a decline in StackOverflow’s visits count. The decline started way before AI became a hot potato

Right now, there is a report circling through the internet called “Stack Overflow is ChatGPT Casualty: Traffic Down 14% in March”. Before you read this article please take a look at the report. I like that it provides some raw data, because it’s always nice to see such and analyse, but I’m concerned by the conclusions.

The main point of the article is that ChatGPT and GitHub Copilot are causing a decline in StackOverflow visitors.

First thing that got my attention is calling StackOverflow a message board. StackOverflow is a Q&A portal. It’s not a message board.

So I was curious if the article defends the title thesis. The thesis is repeated in the article, so the title is not only a marketing effort.

Another red flag I’ve noticed is that in key takeaways StackOverflow data are described in year to year basis and in percent, while GitHub Copilot and ChatGPT in monthly absolute numbers of visitors.

It makes these values impossible to compare. I don’t believe it is wrongdoing, it’s not a scientific article. However, it was a red flag that there might be something wrong with the reasoning.

The first chart is really confusing:

We see an almost vertical line for ChatGPT, a noticeable increase in Github.com visits and…a slight increase in StackOverflow visits…. 🤔

I thought StackOverflow is loosing users…

Also, as a sidenote, visits to GitHub can not be connected directly with GitHub Copilot.

Let’s look closer at the StackOverflow data:

The pink line seems to be quite steady. There is even growth in March 2023. So where is the loss?

Let’s look at the second chart: