It came up in discussion in This Week in Testing, I’d love to know how you would define it?
(And I plan to add good definitions will be added to our Glossary)
2 Likes
I have never even heard of “vibe testing”. I would guess it is whether or not a person feels like the developers have tested the acceptance criteria appropriately.
2 Likes
When it comes to automated tests I would say it’s similar to Vibe coding: you tell the AI what you want to test/what the test should do and let them figure out the rest.
The problem compared to vibe coding is that, aside from error messages, you won’t have any clear indications that it works or not. You still need to be able to know what the code does to be sure it tests everything correctly.
If you want to use the term without automation I think it’s almost the same as exploratory testing 
. At most you ask the AI what/how you should test a feature and only do that without any more questions.
1 Like
Wow… Is this a thing now?
1 Like
The topic came up in the most recent This Week in Testing episode (also available on most podcast players)
If that’s what it is (and I believe that is the case), I would define it as “an unprofessional, ineffective and pointless software-related activity masquerading as testing while being neither testing nor checking”.
It’s got nothing in common with exploratory testing, which is commonly defined as simultaneous test design, test execution and learning. There is no test design because AI has no cognitive function and cannot design anything even though it can create. It also cannot learn, so I very much doubt whether it can take the result from one test and use it to create another test. This iteration is at the heart of exploratory testing, and to be honest most human testers aren’t any good at it either.
You would hope not, but most testers cannot be persuaded to abandon their search for increasingly bad ways to do testing. We know how to do testing well - we’ve known for decades, but it’s difficult and people are lazy. We’re almost at the point where testers don’t even want to have to push the button to start the machine that does all the testing for them.
1 Like
Hello,
I saw an excellent demo of “Vibe-testing” (or prompt engineering) at StarEast 2025 conference last month (Orlando, Florida). It was given by Melissa Benua (you can see her talked on YouTube: https://youtu.be/6SA_B06NQdI?si=yaML6nyqEe0u3fNQ) and I was very impressed.
She used the tool “Cursor” to open a web page on one side, and using the LMM on the right side asking question about what test coverage the code has, what’s missing. It was giving very good answers on unit test, integration test, system test and UI test. BTW, you can do almost the same with GitHub and Copilot… we’ve tried internally.
She ask the LLM different test strategies to implement based on different key persons (Jason Arbon, Andrew Knight); asked to implement a couple of new tests. The LLM proposed some new unit tests. She revised them, looks good. Asked to implement in the code but asked to validate the PR first. Asked the LLM to push everything in production if the test passed, which it did.
Basically, it definitively speed up test implementation and you can trust the LMM but you need to verify.
At the conference I saw others asking LLM on specific test on a UI module, asked to write “something” with playwright, which it did. Verified! Looks good! Asked the LLM to add everything in the code, which it did (create PR, push in production if test pass). Worked perfectly.
“Vibe-testing” is really a great new tool to help in testing. It doesn’t replace good testing practices and other type of test like exploratory testing, or testing integration with HW. It also has many limitations.
You can read an hands-on article about “Vibe-Coding” at the following link. They obtained 20% productivity improvement: (Nobody Codes Here Anymore - Alex Ghiculescu's Newsletter).
At our company we are putting in place different POC with “Vibe-testing” to see how far this new tool can improve our testing.
2 Likes
I believe the way we define vibe coding from a development perspective is the same way we can define vibe testing. It is probably about using LLM tools and prompts to write scripts and manage other QA tasks, like documentation, etc.