Stack Overflow Statistics (2026): 106 Stats on AI and Traffic

A data roundup on Stack Overflow in 2026: how many questions it holds, how fast new questions have fallen, how developers now find answers with AI, and what they need before they trust an AI answer. Every figure comes from the Stack Overflow Developer Survey 2026 or the public Stack Exchange API, and every one was fetched on October 30, 2026.

The short answer: Stack Overflow is still the second most used community platform among developers, at 68.6% of respondents, just behind GitHub (Community data), and it holds 24,244,920 questions and 36,100,748 answers (Stack Exchange API). But developers now go to AI first. 69.9% name “ask an AI agent” as one of their first moves when they hit a problem, and 64.1% say AI means they visit Stack Overflow less for simple questions (Community data). New questions have collapsed: the API counts 141,728 non-deleted questions asked in September 2020 (API, Sep 2020) and 969 in September 2026 (API, Sep 2026). The twist for anyone working on AI visibility: 79% of developers say source attribution matters when they decide whether to trust an AI answer (Knowledge data).

Key takeaways

• 69.9% of developers ask an AI agent as one of their first three moves to solve a problem, second only to online search at 82.7% (Community data)
• 64.1% visit Stack Overflow less for simple questions because of AI, and 28.9% ask fewer questions (Community data)
• 969 non-deleted questions were asked in September 2026, against 136,849 in September 2019 (API, Sep 2019, API, Sep 2026)
• 83% use at least one AI tool at work, led by coding assistants at 65.9% (AI data)
• 79% consider source attribution important or very important for trusting AI answers (Knowledge data)
• 15.9% now visit Stack Overflow more, specifically to verify answers, while only 1.2% contribute more (Community data)
• 74.6% used Anthropic models for work in the past year and 74.3% used OpenAI (Technology data)
• Only 6.6% trust AI with important work decisions; 48.0% trust it when they can easily verify the output (AI data)

How these numbers are measured

Stack Overflow data comes from two very different places, and they should not be mixed up.

Definitions used on this page

• Developer Survey 2026: Stack Overflow’s annual survey, fielded from June 23 to August 5, 2026. The results use 30,903 valid responses from 169 countries. Each question has its own sample size (shown as n), because respondents could skip questions and some questions were only shown to people who use AI.
• Respondents vs. percent: most survey tables are “select all that apply”, so percentages in one table can add up to more than the total number of respondents.
• Stack Exchange API counts: live totals from Stack Overflow’s public API. Question counts by month use the API’s filter=total on questions created in a date range.
• Non-deleted questions: the API only counts questions that still exist. Questions that were closed and deleted later drop out, so older months lose some questions over time and recent months have had less time to be deleted. The trend is still unmistakable, but treat single-month values as approximate.
• Admired: in the survey’s technology section, the share of people who used a technology this year and want to keep using it next year.
• Desired: the share of all respondents who want to use a technology next year.

The survey is promoted mainly on Stack Overflow itself; the survey team notes that 77.4% of respondents have a Stack Overflow account (Community data). So the sample leans toward people who already know the site, which makes the drop in usage even more striking.

How big is Stack Overflow?

1. 24,244,920 questions on Stack Overflow, from the live Stack Exchange API (Stack Exchange API).
2. 36,100,748 answers on Stack Overflow (Stack Exchange API).
3. 31,683,764 registered users (Stack Exchange API).
4. 12,396,905 questions have an accepted answer (Stack Exchange API).
5. 7,212,277 questions are counted as unanswered (Stack Exchange API).
6. 68.6% of developers used Stack Overflow for technology content in the past year, second only to GitHub at 69.5% (Community data).
7. 77.4% of survey respondents have a Stack Overflow account (Community data).

By our arithmetic on the API counts, about 51% of all questions ever asked have an accepted answer (12.40 of 24.24 in millions), and Stack Overflow averages roughly 1.5 answers per question.

Source: Stack Exchange API /2.3/info for stackoverflow, fetched October 30, 2026.

How many questions are asked on Stack Overflow now?

The archive is huge, but very few new questions are being added. We asked the Stack Exchange API how many questions created in each September since 2019 still exist today:

8. 136,849 questions asked in September 2019 (API, Sep 2019).
9. 141,728 questions asked in September 2020, the peak in this series (API, Sep 2020).
10. 119,729 questions asked in September 2021 (API, Sep 2021).
11. 103,417 questions asked in September 2022, the last September before ChatGPT launched (API, Sep 2022).
12. 52,685 questions asked in September 2023 (API, Sep 2023).
13. 24,730 questions asked in September 2024 (API, Sep 2024).
14. 5,072 questions asked in September 2025 (API, Sep 2025).
15. 969 questions asked in September 2026 (API, Sep 2026).

Source: Stack Exchange API /2.3/questions with filter=total, one call per September (UTC), fetched October 30, 2026. Deleted questions are excluded.

By our arithmetic, September 2026 had under 1% of the questions of September 2020 (969 against 141,728), and about a fifth of September 2025. Between September 2022 and September 2023, the first September after ChatGPT launched, the count roughly halved. Part of the 2025 and 2026 drop may be deletion timing and moderation changes, but no deletion effect explains a fall of this size.

How do developers find answers in 2026?

16. 82.7% use online search tools as one of their first three moves to find an answer (Community data).
17. 69.9% ask an AI agent (Community data).
18. 8.0% post in a public online forum, the kind of place Stack Overflow used to own (Community data).
19. 65.1% usually find answers without posting anything (Community data).
20. 44.7% say AI tools answer enough of their questions, a reason they participate less in communities (Community data).
21. 24.0% use Stack Overflow as one of their most-used sources for work questions, behind official docs at 30.5% (Knowledge data).
22. 58.9% learned to code in the past year from technical documentation, and 52.6% from AI codegen tools (Community data).
23. 33.8% learned to code from Stack Overflow or Stack Exchange in the past year (Community data).
24. 11.1% used Stack Overflow or Stack Exchange to learn AI programming (Community data).

Source: Stack Overflow Developer Survey 2026, Community data (“Which actions do you turn to first…? Select your top 3”), fetched October 30, 2026.

How has AI changed Stack Overflow use?

25. 64.1% visit Stack Overflow less for simple questions (Community data).
26. 28.9% ask fewer questions (Community data).
27. 26.5% contribute less (Community data).
28. 21.8% report no change (Community data).
29. 15.9% visit more for verification, for example source attribution (Community data).
30. 5.7% visit more because AI-generated answers lead them there (Community data).
31. 1.2% contribute more (Community data).
32. 12.2% search Stack Overflow to check an AI answer before using it (AI data).

The survey team’s own summary is that Stack Overflow is moving “from ask to verify”. People still come, but increasingly to check something an AI tool told them, not to ask the question in the first place.

How many developers use AI tools?

33. 83% use at least one AI tool or agent at work (AI data).
34. 65.9% use AI coding assistants or coding agents (AI data).
35. 62.5% use general-purpose AI chat tools (AI data).
36. 26.2% use AI agents or automated workflows (AI data).
37. 17.8% use internal AI tools built by their company (AI data).
38. 17.2% do not use AI tools at work (AI data).
39. 77% of AI users use AI daily overall (AI data).
40. 73.0% of coding assistant users use them every day, and 65.6% of chat tool users do (AI data).
41. 30.9% of daily AI users spend four hours or more a day with AI, and 24.1% spend two to four hours (AI data).
42. 40.5% use three to five AI tools in a typical week (AI data).

Source: Stack Overflow Developer Survey 2026, AI data (n = 17,464), fetched October 30, 2026.

Which AI models and tools do developers use?

43. 74.6% used Anthropic models for work in the past year (Technology data).
44. 74.3% used OpenAI models (Technology data).
45. 54.5% used Google AI models (Technology data).
46. 26.4% used Microsoft models (Technology data).
47. 18.9% used DeepSeek, 9.8% Mistral AI and 7.6% xAI (Technology data).
48. 60.5% of Anthropic users want to keep using it, against 41.5% for OpenAI (Technology data).
49. 65.5% of coding agent users use Claude Code, 58.7% GitHub Copilot, 29.5% OpenAI Codex and 25.5% Cursor (AI data).
50. 30.2% of agent orchestration users use LangChain (AI data).
51. 38.9% of AI automation users use Make, 32.7% n8n and 18.5% Zapier (AI data).
52. 38.3% of no-code agent users use Lovable (AI data).
53. 33.7% did not switch AI tools in the past year; the top reason for switching was better output quality, at 25.1% (AI data).

Source: Stack Overflow Developer Survey 2026, Technology data (LLM vendors, n = 11,895), fetched October 30, 2026.

Do developers trust AI answers?

54. 62% of developers are favorable toward AI tools overall: 31.6% very favorable and 30.4% somewhat favorable (AI data).
55. 87% have some level of trust in AI output, up from 31% last year, per the survey team’s highlights (2026 survey highlights).
56. 48.0% trust AI when they can easily verify the output (AI data).
57. 6.6% trust it for many tasks including important work decisions (AI data).
58. 12.3% do not trust it for most work tasks (AI data).
59. 76.5% validate AI output by running it locally, and 38.0% read official docs (AI data).
60. 9.6% use AI answers as-is (AI data).
61. 16.0% say AI handles complex tasks very well (AI data).
62. 69.3% say “plausible but wrong code” cost their team the most time, and 33.9% lost time to outdated APIs (Technology data).
63. 41.3% do not avoid AI at work or school; the biggest reason for avoiding it is concern over losing skills, at 17.1% (AI data).

What makes an answer trustworthy?

This is the section that matters most for brands. Developers are telling us what an answer needs before they act on it, and the same signals decide which sources AI engines lean on.

64. 79% consider source attribution important or very important when deciding whether to trust an AI-generated technical answer (Knowledge data).
65. 52.1% say source attribution is very important, and another 26.5% say it is important (Knowledge data).
66. 93% of respondents think trusting AI answers comes down to citing sources, in the survey team’s framing (2026 survey highlights).
67. 72.6% rely on information that is recent or clearly up to date (Knowledge data).
68. 60.0% want it to come from an official or trusted source (Knowledge data).
69. 37.6% want sources, citations, links or supporting evidence (Knowledge data).
70. 59.1% trust answers that are well-organized and logical (Community data).
71. 48.0% trust content with many upvotes, likes or views (Community data).
72. 39.8% trust information with many citations they can trace back (Community data).

Source: Stack Overflow Developer Survey 2026, Knowledge data (n = 12,996, select up to 5), fetched October 30, 2026.

Which platforms do developers use?

73. 69.5% used GitHub for technology content in the past year (Community data).
74. 58.4% used YouTube (Community data).
75. 53.8% used Reddit (Community data).
76. 37.8% used LinkedIn (Community data).

Source: Stack Overflow Developer Survey 2026, Community data (n = 10,741), fetched October 30, 2026.

Reddit at 53.8% is worth noting: it is one of the sources AI engines cite most, which our Reddit statistics cover in detail.

How do developers buy software?

77. 59.5% influence software tool choices by giving feedback or requirements (Work data).
78. 48.2% evaluate, recommend or shortlist tools (Work data).
79. 17.7% make the final purchase decision (Work data).
80. 2.3% would recommend a technology mainly for its AI features (Work data).
81. 38.1% say their organization sets limits on AI usage or spending (Knowledge data).
82. 29.7% say AI use is mostly optional and up to individuals, while 24.0% have approved tools and guidance (AI data).
83. 10.5% of responses are freelance or self-employed, up from 3.9% (2026 survey highlights).
84. 33% say they are not happy at work this year (2026 survey highlights).

About the 2026 survey

85. 30,903 valid responses from 169 countries are used in the 2026 results (Methodology).
86. 39,266 total rows were recorded between June 23 and August 5, 2026 (Methodology).

Why Stack Overflow’s decline matters for AI visibility

Stack Overflow is the clearest case study of what AI answers do to a source site. Developers still need the answers, but they get them from ChatGPT, Claude or a coding agent, and only some of them click through to check. That shift has three consequences for any brand that sells to technical buyers:

1. Your buyers research in AI tools first. If 69.9% of developers ask an AI agent as a first move and 48.2% shortlist software, the shortlist increasingly starts inside an AI answer. Our AI search statistics and ChatGPT statistics show how big that audience has become.
2. Citations decide trust. 79% of developers say source attribution matters. Recent and official are also the top two trust signals for work information in the survey, so those are the pages you want AI engines such as Perplexity, ChatGPT search and Google’s AI answers to cite. Our GEO statistics show that adding sources and statistics is one of the strongest levers on AI visibility.
3. Community sources still feed the models. Stack Overflow, GitHub and Reddit content is what models learned from and what AI search still retrieves, much like Wikipedia. Being discussed accurately in those places matters even when clicks fall.

You can’t see any of this in Google Analytics. To know whether ChatGPT, Perplexity, Claude or Gemini mentions your product when a developer asks for a recommendation, and which sources it cites, you need to track the answers themselves. Mentionova runs a set of your buyers’ prompts across those engines, shows which sources each one cites and flags answers that get your product wrong. For a wider shortlist, see our best AI citation tracking tools and best LLM brand monitoring tools, and our AI crawler statistics for the bot side of the story.

Stats ledger: what each source measured

Source: tools8020 stats ledger for this page. Every row links to the primary page or API we fetched on October 30, 2026.

What to do with this data

1. Stop using Stack Overflow traffic as a proxy for developer demand. The questions moved to AI tools; they did not go away.
2. Write for verification. Developers check AI answers against official docs and tests. Make your docs, pricing and changelog pages the ones they land on to verify.
3. Keep pages current. Recency was the top trust signal at 72.6%. Date your pages and update them.
4. Cite your own numbers. If attribution decides trust, give AI engines something to attribute: specific, sourced claims on pages you control.
5. Be present where models learn. Accurate answers on Stack Overflow, GitHub and Reddit still shape what models say about your category.
6. Track AI answers directly. Use an AI visibility tool such as Mentionova to see how ChatGPT, Perplexity, Claude and Gemini describe your product and which sources they cite.

Frequently asked questions

How many questions are on Stack Overflow?

24,244,920 questions and 36,100,748 answers, according to the Stack Exchange API on October 30, 2026. 12,396,905 questions have an accepted answer.

How many users does Stack Overflow have?

31,683,764 registered users, per the Stack Exchange API on October 30, 2026. In the 2026 Developer Survey, 77.4% of respondents had an account and 68.6% used Stack Overflow for technology content in the past year.

Is Stack Overflow dying?

As a place to ask questions, it has shrunk dramatically. The API counts 141,728 surviving questions from September 2020 and 969 from September 2026. As a reference, it is still widely used: 68.6% of developers used it in the past year, and 15.9% now visit more to verify AI answers.

How many questions are asked on Stack Overflow per month now?

The API counts 969 non-deleted questions created in September 2026 and 5,072 in September 2025. Deleted questions are not counted, so the true number asked was somewhat higher.

Why did Stack Overflow traffic drop?

Developers get answers from AI tools. In the 2026 survey, 64.1% say they visit Stack Overflow less for simple questions because of AI, 28.9% ask fewer questions and 44.7% say AI tools answer enough of their questions.

What percentage of developers use AI?

83% use at least one AI tool at work, according to the 2026 Developer Survey. 65.9% use coding assistants or agents and 62.5% use general-purpose chat tools.

Which AI coding tool do developers use most?

Claude Code, used by 65.5% of coding agent users, ahead of GitHub Copilot at 58.7%, OpenAI Codex at 29.5% and Cursor at 25.5%.

Do developers trust AI-generated code?

Conditionally. 48.0% trust AI when they can easily verify the output, but only 6.6% trust it with important decisions. 76.5% run AI code locally before using it, and 69.3% say plausible but wrong code cost them the most time.

How many people took the 2026 Stack Overflow Developer Survey?

The results use 30,903 valid responses from 169 countries, out of 39,266 rows recorded between June 23 and August 5, 2026.

Does source attribution matter for AI answers?

Yes. 79% of developers say source attribution is important or very important when deciding whether to trust an AI-generated technical answer, and 52.1% say it is very important.

How can a brand see what AI tools say about it to developers?

Run the prompts your buyers use in ChatGPT, Perplexity, Claude and Gemini and record whether you are mentioned and which sources are cited. An AI visibility tracker such as Mentionova does this on a schedule and shows the cited sources for each answer.

How we compiled these statistics

We fetched 7 pages from the Stack Overflow Developer Survey 2026 site and 9 Stack Exchange API endpoints on October 30, 2026 and recorded 106 data points in docs/content/stats-ledgers/stack-overflow-statistics.json. Every numbered figure above maps to a ledger row with a verbatim quote from the page or API response we fetched. Survey figures use the survey’s own published percentages and sample sizes. Question counts are non-deleted questions from the public API, so they understate how many were originally asked, most of all for older months. Ratios and year-on-year comparisons are our own arithmetic and are labelled as such. We did not use third-party traffic estimators.

Sources

• Stack Overflow Developer Survey 2026: highlights
• Stack Overflow Developer Survey 2026: AI, full data
• Stack Overflow Developer Survey 2026: Community, full data
• Stack Overflow Developer Survey 2026: Knowledge, full data
• Stack Overflow Developer Survey 2026: Technology, full data
• Stack Overflow Developer Survey 2026: Work, full data
• Stack Overflow Developer Survey 2026: Methodology
• Stack Exchange API: /info for stackoverflow