Which AI Tools Do People Use for Different Kinds of Apps?
The Claude tool family appears in 61.7% of recorded developer-tool stacks and 36.7% of social-project stacks. Compare four categories in VCL Research.
The tools people report using differ with the kind of product they build. In VCL’s latest published comparison, the Claude.ai / Claude Code family appeared on 100 of 162 Developer Tools projects with recorded tool data: 61.7%. For Social projects, it appeared on 61 of 166, or 36.7%.
An overall tools ranking hides that difference. Breaking the sample into categories gives writers and builders a more specific answer to “what are people using?”
*Analysis updated October 23, 2026. Dataset v9; measurement cutoff October 1, 2026 at 00:00 UTC.*
Download the chart · Download the source CSV
Four categories, different tool mixes
The table shows the share of projects with recorded tool data within each category. Related products are grouped: a Claude.ai / Claude Code selection does not establish which product was used, or that both were used. The same applies to ChatGPT / Codex.
| Project category | Projects with tool data | Claude.ai / Claude Code | ChatGPT / Codex | Cursor | Google AI Studio |
|---|---|---|---|---|---|
| Productivity | 334 | 52.7% | 46.1% | 32.9% | 14.7% |
| Games | 200 | 45.5% | 42.0% | 35.5% | 20.5% |
| Developer Tools | 162 | 61.7% | 46.3% | 35.8% | 19.1% |
| Social | 166 | 36.7% | 38.0% | 30.7% | 12.0% |
The Claude family has the highest recorded share among these four families in Productivity, Games and Developer Tools. In Social, ChatGPT / Codex is slightly ahead: 63 projects versus 61. That two-project difference should not be presented as a decisive lead.
What the comparison can tell us
These four categories contain 849 distinct eligible projects; 660 have recorded tool data, giving 77.7% coverage. Categories overlap, and a project can name several tool families. Neither the category sample sizes nor the percentages should be added together.
The choices are builder- or submitter-reported. They do not measure subscription market share, time spent in each tool, or which assistant wrote most of the code. The comparison also does not establish that one tool causes better results for a particular category.
Its useful contribution is narrower: a description of the recorded stacks behind different kinds of public VCL projects. The overall tools article shows the broader ranking and the combinations that appear together.
A statistic to cite
In VCL Research’s dataset v9, the Claude.ai / Claude Code family appeared in 61.7% of Developer Tools projects with recorded tool data (100/162), compared with 36.7% of Social projects (61/166).
Data and source. State of Vibe Coding 2026, Dataset v9: AI tools by project category, cutoff October 1, 2026 at 00:00 UTC. Read the methodology and full report. When sharing the chart, please credit VCL Research and link to this article or the versioned figure.