Stack Overflow is a website that has nurtured a generation of programmers. For twenty years, it was the first tab people opened whenever they encountered a bug in their code. Today, its traffic and the number of questions are falling so sharply that it’s no longer just a decline—it’s a shift in an era. Let’s take a closer look at what’s happening to it—and why this affects more than just programmers.
Numbers
Number of new questions per month: a peak of over 200,000 in 2014, less than 50,000 by the end of 2025. In other words, the figure has fallen back to roughly the 2008 level—fifteen years of growth have been wiped out.
The year-over-year comparison is particularly telling: in December 2024, there were 25,500 new questions, compared to 42,700 a year earlier—a drop of about 40% (Stack Overflow Trend Analysis). According to Similarweb, traffic continues to decline each month (PPC Land Review).
Different sources use different calculations, so focus on the order of magnitude and the trend, rather than the exact numbers. The trend is the same across the board.
What Happened
The decline began even before ChatGPT—around mid-2021—but accelerated sharply after November 2022. The reason is simple and mundane: the behavior pattern has changed.
In the past, when a developer encountered an error, they would go to Google, find a similar question, and read the answers and comments. Now, they simply paste the error into the chat and immediately receive a response tailored to their code. No registration, no waiting, and no risk of getting a “this is a duplicate, closed” message instead of help.
Not only are there fewer questions, but there are also fewer people answering them. This creates a vicious cycle: fewer new answers mean fewer reasons to visit the site.
A Paradox Worth Understanding
The models that now provide answers in place of Stack Overflow were trained largely on Stack Overflow. Millions of questions and answers, written by people for free in order to earn reputation within the community, became training data for the services that have taken the site’s audience away.
The site shared its knowledge—and saw a decline in traffic. The company responded with licensing agreements: access to training data is now sold to large AI companies, and there’s also an enterprise product for internal knowledge bases. In other words, Stack Overflow isn’t going away—it’s simply ceasing to be a free public platform that relies on advertising revenue from traffic.
Why this isn't just about programmers
Stack Overflow is just the first major example. The model is the same for any website whose value boils down to “here’s the answer to the question”:
- reference books and encyclopedias;
- "how-to" websites;
- recipes, schedules, converters, reference guides;
- A significant portion of corporate blogs simply repeat what is already common knowledge.
The same thing is already happening in regular search: Google provides answers right in the search results, and some queries no longer reach websites. We've discussed this separately— Google shows answers instead of links.
What Artificial Intelligence Doesn't Take Away
It's helpful to focus not on what disappears, but on what remains:
- My data. Your prices, your timelines, your case studies with numbers. The model doesn't know them because they aren't on the internet.
- Action, not information. Buy, sign up, calculate, complete the process. You can forward a reply, but not a transaction.
- Tools. People go back to places where things actually work, not just where they're written down. We've done a few free tools based on this very logic.
- Relevance. What changed yesterday: availability, prices, schedule, recent changes to the law or the platform.
- Trust and Reputation. Who is responsible for the outcome? Chat will not sign the contract or visit the site.
- Locality. The conditions in your city, your delivery services, your reality.
What should a website owner do about this?
- Check what your traffic consists of. If the majority of the content consists of informational articles on general topics, this section will continue to shrink.
- Shift your focus to commercial pages. "Order," "prices," "case studies"—people go there anyway because they need to choose a service provider, not look up definitions.
- Write down what is missing from the model. Personal experience, analysis of real-world projects, statistics, mistakes, and local specifics.
- Build your audience directly. Newsletter, messenger, community—these are channels that aren’t affected by how Google changes its search results tomorrow.
- Don't rely solely on search. Recommendations, links from client websites, local search results, and advertising.
Don't rush to bury it
Talk of the “death of Stack Overflow” is just as exaggerated as the talk of the death of email was in its day. Complex, rare, and new problems still find their way to people: artificial intelligence performs well when the answer already exists, but noticeably worse when no one has written it yet.
The irony is that if there are critically few human responses, future models will have nothing to learn from. This is an open question to which no one yet has an answer.
In Brief
A website whose value lies in providing answers to questions loses its audience. A website whose value lies in action, data, a tool, or reputation remains in demand.
If your website is currently more like the first type, that's a good reason to review its structure—don't wait. We We create turnkey websites and We're showing our support available; the project can be calculated in three steps.
Telegram