| Web Devs Positive About AI Tools |
| Written by Janet Swift | |||
| Wednesday, 28 May 2025 | |||
|
Like it or not, AI is having a revolutionary impact on web development. Maybe we haven't all succumbed to vibe programming, but almost all of us are using AI tools in our everyday routine and the results of the recent survey from Devographics reveals.
To discover more about which AI tools are being used and web developers' reaction to them the State of Web Dev AI survey was conducted earlier this year. It comes from Devographics, the team led by Sacha Grief which regularly reports on the State Of JavaScript and of CSS. This survey was created to identify upcoming trends in terms of using AI for web development, in order to help developers make better technological choices. Questions were devised to probe which tools are the most used and the best regarded and also to reveal developer's biggest pain points when it comes to actually using AI to develop web apps. Respondents were primarily a mix of respondents from past Devographics surveys and social media traffic. It attracted over 4,000 responses. The report notes that: Although the survey was open to all developers regardless of whether they use AI or not, the topic of the survey by itself is likely to have biased the respondent set towards developers who do have an interest in AI. A key result from the survey is shown in this graphic:
If you are familiar with how the results of Devgraphics surveys are presented you'll know that here we are looking at a chart that sums up Experience and Sentiment of different fundamental AI models ordered in terms of usage. In the report in can be viewed either by Experience or by Sentiment (as here) and you can click on the names of the Models to the left to find out more about them. Experience has three categories - Used it (Blue); Heard of it (Light blue); Never heard of it (Beige) and Sentiment has three categories - Positive (Green), Neutral (Colorless); Negative (Red). As you can see OpenAI's ChatGPT is overwhelmingly blue for experience and either green or colorless for sentiment. Commenting on this the report states: It should be no surprise that OpenAI's ChatGPT still benefits from its first-mover advantage and large mindshare, making it both the most-used (91.2%) and most-loved (53.1%) model provider. While a small minority (8%) expressed negativity to ChatGPT In fact there were no "Never heard of it" responses. A close second in terms of sentiment was Anthropic's Claude with 47% positive votes followed in third place by DeepSeek with 32% positive votes. While 7% had never heard of Claude, less than 1% had never heard of DeepSeek. The models with the most negative responses were Microsoft Copilot (29%) and Google Gemini (21%) both of which had been used by 64% of respondents. At the bottom of the table, Phind had only been used by 7%. with 84% never having heard of it. Most respondents (73%) were neutral towards it but 17% were negative. In response to a question about other models, 25% of respondents were using Grok and 22% Qwen. GitHub Copilot was nominated by 5% - is this a confusion with Microsoft Copilot as otherwise the percentage would surely be higher - and Codeium by 4%. This serves to highlight a problem faced by the survey - the way in which AI products change their names - click on Codeium to find out about it and you'll find details for Windsurfer, its new name. Amazon Q (formerly Amazon Whisperer) is also listed, with 3%. One product that cropped up in this list was Cursor and it tops the list for IDEs Experience and Sentiment. Quoting the report: Cursor has a dominant awareness lead, with 82.2% of respondents having used it or heard of it, versus just 54.1% for runner-up Zed. When looking at freeform comments left about Cursor, it seems like the main issue is actually its price, indicating that the market might have room for a cheaper alternative. The other two products included in the table of AI-powered text editors are Windsurf Editor and Void, with 66% and 89% of respondents never having heard of them respectively. In answer to the question about what other IDEs were in use VS Code topped the list with 41%. This reflects its status as the most popular IDE and since the date of its survey it has added so many AI-powered features that it will undoubtedly be included as a AI tool at the next iteration of the survey. GitHub Copilot was also on the list of other IDEs with 19%, as was JetBrain AI with 7%. They were also in the top three of the list of seven Coding Assistants:
GitHub Copilot stands out in terms of awareness - 71% of respondents had used it and all the rest had heard of it. JetBrains AI is the most negatively regarded, 23%, but notice that sentiment was expressed mainly by those who had heard of it or never heard of it. Only 4% of respondents had both actually used it and were negative about it. Looking at Code Generation tools only a minority of respondents (31%) has used any products in this category. The reports notes: With Vercel's might behind it, v0 has quickly established itself as the leader in this budding new industry sector. But StackBlitz' Bolt is not far behind, and certainly worth keeping an eye on. Even so only 23% of respondents had used v0 and only 20% were positive towards it. Bolt had been used by 13% with 15% being positive about it. So what do web devs use AI for? Overwhelmingly for Code Generation, even though not yet by using tools specific to the task!
Most of us aren't quite vibe coding just yet, with a majority of respondents (69%) generating less than 25% of their code through AI – and only a small minority (8%) generating more than 75% of it. Learning and research comes next with 66%. Then there a tasks related to documentation including Image generation at 38%. So what is holding us back from greater use of AI-powered tools?
In the case of Models, the overwhelming response among 1,354 respondents is Hallucination and inaccuracies, a well-publicized drawback, but notice that is is only cited by 46%. The second and third points for Models are Context and memory limitations and Poor generated code quality are 16% and 12% respectively. When it comes to IDEs, the top pain point is Context and memory limitations which is noted by 15% of 550 respondents When it comes to Code Generation, Poor generated code quality is selected by 53%, but the number of respondents to this is only 164. Hallucination and inaccuracies are a problem for 9% of this group. As AI-powered tools advance, something they are doing at a rapid pace, all these pain points will undoubtedly ease.
More InformationRelated ArticlesGitHub Copilot Provides Productivity Boost Developers Wary Of The AI Tools They Use Developers Like Code Assistants Even When They Are Incorrect GitHub Sees Exponential Rise In AI CSS Ecosystem In the Spotlight To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on Twitter, Facebook or Linkedin. 
Comments
or email your comment to: comments@i-programmer.info
|
|||
| Last Updated ( Wednesday, 28 May 2025 ) |
