HomeDashboardDeveloper AI trust

Analysis 07 of 10Data

What do developers trust about AI tools?

Developers use AI tools daily but rarely trust the output, and small companies buy more freely than large ones. How should a sales team pitch them?

The short answer 50.6% of professional developers use AI tools daily, but only 4.7% of those daily users highly trust the accuracy of what the tools produce. Buying power sits in small companies and with managers: 30.1% of developers at companies under 20 people influenced a substantial purchase, against 11.0% at companies of 10,000 or more. A pitch should lead with how a tool checks its own output, aimed at smaller teams and their managers.

  • Public data, not my business
  • Self-selected survey audience
Source
Stack Overflow Developer Survey 2025, Stack Overflow
Period
Fielded May 29 to Jun 23, 2025
Licence
Open Database License (ODbL) 1.0
Charts
5, each with its data table and CSV

Key figures

  • 50.6%Use AI tools dailyof 26,045 professional devs who answered
  • 4.7%Highly trust AI accuracyamong 13,019 daily AI users
  • 19.9%Influenced a substantial purchaseof 31,309 professional devs
  • 34.4%Fully remoteof 28,286 professional devs

The charts

Chart 1Lead chart · breakdown

Daily AI use does not mean daily trust.

Do the developers who use AI most trust its output?

Among 13,019 daily AI users, only 4.7% highly trust the accuracy of the output; 43.2% somewhat trust it. Daily users trust the output more than any other group, but most still land on somewhat trust or worse.
Download CSV

Stack Overflow Developer Survey 2025 · public data, not my business

Trust by use: 5 rows. trust_gap in tech.json: professional developers who answered both questions, percent within each use group.
How often they use AIRespondentsHighly distrust (%)Somewhat distrust (%)Neither trust nor distrust (%)Somewhat trust (%)Highly trust (%)
Daily13,0197.719.924.543.24.7
Weekly4,48013.734.324.626.31
Monthly / infrequent3,27627.53819.214.60.7
No, plan to soon1,16515.538.326.218.51.5
No, don't plan to3,77662.424.99.33.30.1

Chart 2Ranking

Small companies buy more freely than large ones.

Where do developers influence a substantial purchase?

Share of professional developers who influenced a substantial purchase, by company size. 30.1% at companies under 20 people, 11.0% at 10,000 or more.
Download CSV

Stack Overflow Developer Survey 2025 · public data, not my business

Buying power by size: 8 rows. purchase_influence in tech.json, role group All.
Company sizeInfluenced a substantial purchase (%)Respondents
Just me / freelance33.51,097
<2030.14,828
20-9921.85,882
100-49916.65,239
500-99914.91,985
1,000-4,99914.23,349
5,000-9,99912.91,241
10,000+114,024

Chart 3Breakdown by segment

Managers and architects carry the most buying influence at every size.

Which roles hold that buying influence?

The same measure split by role group. Managers and architects are highest in every company size; at companies under 20 people, 52.9% of them influenced a substantial purchase.
Download CSV

Stack Overflow Developer Survey 2025 · public data, not my business

Buying power by role: 8 rows. purchase_influence in tech.json by role group; respondent counts per cell are in the source JSON.
Company sizeManager / Architect (%)Developer (%)Data / AI / DevOps (%)Other (%)
Just me / freelance39.231.437.338.6
<2052.925.330.549.9
20-9942.617.522.939.3
100-49935.112.621.526.6
500-99931.412.317.914.4
1,000-4,99935.810.812.517.6
5,000-9,99930.89.31417.5
10,000+21.98.413.713.5

Chart 4Ranking

JavaScript, HTML/CSS and SQL lead.

What do professional developers actually use?

JavaScript, HTML/CSS and SQL remain the most common languages in production, out of 24,751 professional devs who answered.
Download CSV

Stack Overflow Developer Survey 2025 · public data, not my business

Languages: 12 rows. top_languages in tech.json (multi-select, so shares add to more than 100%).
LanguageShare of professional developers (%)Respondents
JavaScript68.917,046
HTML/CSS63.115,616
SQL61.415,207
Python54.913,596
Bash/Shell (all shells)48.912,103
TypeScript48.812,083
C#29.97,407
Java29.77,346
PowerShell23.15,721
C++21.85,398
C19.24,742
PHP19.14,724

Chart 5Distribution

Fully remote is still the single biggest arrangement.

Where do developers work?

34.4% of 28,286 professional devs work fully remote; only 16.0% are fully in-person.
Download CSV

Stack Overflow Developer Survey 2025 · public data, not my business

Work arrangement: 5 rows. remote_split in tech.json.
ArrangementShare of professional developers (%)Respondents
Remote34.49,722
Hybrid, leans in-person19.95,622
Hybrid, leans flexible17.64,966
In-person164,536
Fully flexible12.23,440

01Findings

What the charts say.

  1. Only 4.7% of 13,019 daily AI users highly trust the output; most somewhat trust it or less.

    Chart 1

  2. 30.1% of developers at companies under 20 people influenced a substantial purchase, against 11.0% at 10,000 or more.

    Chart 2

  3. Managers and architects have the highest purchase influence of the four role groups at every company size.

    Chart 3

  4. JavaScript (68.9%), HTML/CSS (63.1%) and SQL (61.4%) are the most used languages, and 34.4% work fully remote.

    Chart 4 Chart 5

02Method

How the figures were built.

Steps

  1. Kept the 15 columns used in this analysis out of 172 raw survey columns.
  2. Filtered to respondents who chose 'I am a developer by profession' for most KPIs and charts; the published table keeps all respondent types so it can be queried either way.
  3. Dropped rows with 'NA' (skipped question) from the denominator of each specific chart or KPI, and state n for each.
  4. Split multi-select columns (languages, databases, platforms) on ';' and kept them as raw ';'-joined strings in the table; unnested only inside the analysis queries.
  5. Grouped the 15 DevType values into 4 role groups (Manager/Architect, Developer, Data/AI/DevOps, Other) for the purchase-influence-by-role view.

49,191 responses, 37,467 professional developers used for most charts; each chart states its own base. Every derived figure on this page (changes, indexes, averages, counts) is computed from the published rows, and the table under each chart says how.

Checks

  • Skipped answers (NA) are dropped from each chart's denominator, and each chart states its n.
  • Each table on this page is the published chart data in tech.json, shown row for row.

The queries

The query runs on the table dev_survey (49,191 responses). Open it in Live SQL, which runs DuckDB in your browser on the same public tables; the query loads ready to run.

Trust in AI accuracy by use frequency

Open in Live SQL 
SELECT ai_select AS ai_use_frequency, ai_acc AS trust_in_ai_accuracy, COUNT(*) n,
       ROUND(100.0*COUNT(*)/SUM(COUNT(*)) OVER (PARTITION BY ai_select), 1) pct_within_use_group
FROM dev_survey
WHERE main_branch = 'I am a developer by profession' AND ai_select <> 'NA' AND ai_acc <> 'NA'
GROUP BY 1,2
ORDER BY 1,2

Top languages among professional devs

Open in Live SQL 
SELECT lang, COUNT(*) n FROM (SELECT unnest(str_split(languages_used, ';')) lang FROM dev_survey WHERE main_branch = 'I am a developer by profession' AND languages_used <> 'NA') GROUP BY 1 ORDER BY n DESC LIMIT 10

Purchase influence rate by org size

Open in Live SQL 
SELECT org_size, COUNT(*) n, ROUND(100.0*SUM(CASE WHEN purchase_influence LIKE 'Yes, I influenced the purchase of a substantial%' THEN 1 ELSE 0 END)/COUNT(*), 1) pct_substantial_influence FROM dev_survey WHERE main_branch = 'I am a developer by profession' AND org_size <> 'NA' AND purchase_influence <> 'NA' GROUP BY 1 ORDER BY pct_substantial_influence DESC

Remote work split among professional devs

Open in Live SQL 
SELECT remote_work, COUNT(*) n FROM dev_survey WHERE main_branch = 'I am a developer by profession' AND remote_work <> 'NA' GROUP BY 1 ORDER BY n DESC

Caveats

  • Respondents are Stack Overflow's own audience, self-selected, and skew toward developers who already engage with the site.

Releases and revisions

  • The 2025 survey is a fixed release; it is not revised.

03Sources

Sources and licence.

Licence: Open Database License (ODbL) 1.0. Public data, not my business.

Download the chart data

Each CSV is made in your browser from the rows in that chart's table.

All ten analyses on the Dashboard

04Contact

Want this kind of breakdown on your own data?

Open to e-commerce, data and operations roles. Los Angeles or remote.

zacharybhenley@gmail.com Book a 30-minute call