Anyone tried Grubby.ai humanising content

I'm finding a tool for my university project and this thing seems to work, I only have 300 words to try so i'm not sure yet.

It claim that it can evade Turnitin, GPTzero, etc. which from my limited test seems to be true.

Might not be a bad idea to use it for content writing too.
But are they really legit?

The youtube videos promoting them are mainly their own creation.
Can't tell apart from the Reddit reviews either.

Anybody has genuine experience and tried them? For Academia especially

Ok, because it's hard to test with just 300 free words input.

i tried the paid subscription and they don't refund even after i point out the failures which is evident only after you pay for it to test long articles.

They point you to this

Subscriptions on Grubby are non-refundable once processed, as outlined in our Terms of Service (section 7.5). That said, we do have a Turnitin Bypass Guarantee that may apply to your case — but it's specific to Turnitin detections, not other detectors.

Click to expand...

There's no Turnitin detector on its website for you to test against in the first place.

The problem is when you put long text, the context are so off and sounds so bad.
There's no way you can use that unless you are happy with poor grades.

So naturally you will edit some specific parts manually and test it again.
You can't test it that way nor can you test on Turnitin. You won't see this problem until you bought their sub for more text input.

So you can only test your cowork product with GPTzero which flag it as partially AI written

Meaning, you can only use their full written content and unable to make amendment to their garbled outputs.
Which make them not usable for Academia which they claimed to be for sure.

As for how AI proof they really are?

I put the content onto Claude and it could identify instantly.

Don't use this version. It looks like it was run through a paraphrasing or "humaniser" tool, and that has done real damage:


1. Your interview quotes have been changed. Quotes must match the appendix word for word, and the marker can check them.
Misquoting is an accuracy and integrity problem, much worse than any grammar slip.

2. Meanings have been reversed or broken:

3. It reads worse. "Also" appears in almost every sentence, and the flow is flatter. That pulls the grade down toward B+, not up.


What to do:
■ Go back to your previous version, the one you sent at 21:03.
■ Keep any of my proofread fixes you've already made up to 4.1, but revert the rewordings I listed last time.
■ For 4.2 onward, apply only mechanical fixes: spelling, plurals, tense, "the", full stops.
■ Don't run it through any rewriting tool. Your own voice with correct grammar is the safest version and the strongest one.

Click to expand...

I won't be so pissed if not for their false advertsing, refund policy is very unethical.

I would like to test it on a bigger sample first before I believe in its results. AI detectors do not have consistency either, so for academic purposes, I will put emphasis more on original writing and citation guidelines.

Will check this out.

Actually you can trigger any AI chat to stop writing like AI by telling it to write like a normal human, not so professional. When AI gets too professional, you know it's AI written content.

AI always tend to funnel towards certain ways or talking or describing things or just words in general.
like load-bearing is one of the biggest examples.

anything that claims to humanize text, i honestly stay far away from.
I have threw in harry potter and novels in "GPTZero, Originality.ai" etc and they gave me high chance of AI...
yeah, cuz AI was trained off that text.

For my own text i write, i run it thru some workflows that detects stuff including but not limited to Negative parallelism, composure verb, Over-explained subtext, over-states its own theme etc.
It's not perfect, but at the end of the day, neither are AI Detectors.

xReminisce said:

AI always tend to funnel towards certain ways or talking or describing things or just words in general.
like load-bearing is one of the biggest examples.

anything that claims to humanize text, i honestly stay far away from.
I have threw in harry potter and novels in "GPTZero, Originality.ai" etc and they gave me high chance of AI...
yeah, cuz AI was trained off that text.

For my own text i write, i run it thru some workflows that detects stuff including but not limited to Negative parallelism, composure verb, Over-explained subtext, over-states its own theme etc.
It's not perfect, but at the end of the day, neither are AI Detectors.

Click to expand...

You can pretty easily detect if it's AI content if they have dashes instead of complete sentences

How was he today - in other news he was good.

Click to expand...

There are other tools out there you can also do to make your content look as human like as possible.

What's the point of going to uni if you're going to cheat.

You realize you're just moving yourself right towards a reality where you're obsolete, right?

If you can cheat a university, who do you think is going to hire you?

Humanity really astounds me.

Instead of spending your time trying to become valuable to the world, you're looking at how to cheat. Good luck..

TomBelfort said:

What's the point of going to uni if you're going to cheat.

You realize you're just moving yourself right towards a reality where you're obsolete, right?

If you can cheat a university, who do you think is going to hire you?

Humanity really astounds me.

Instead of spending your time trying to become valuable to the world, you're looking at how to cheat. Good luck..

Click to expand...

AI is tool. If you don't want to use it in the future, then you will be beat out by your competition.

While I agree with what you said, unfortunately the era of AI is here and we must adapt to it.

Albert Wesker said:

AI is tool. If you don't want to use it in the future, then you will be beat out by your competition.

While I agree with what you said, unfortunately the era of AI is here and we must adapt to it.

Click to expand...

Oh? It's the future is it? I'll be beaten by my competition?

Tell me more about this AI thing.

Gosh, I need to learn how to use a humanizer AI app so i can beat my my competition.

Here I am thinking 4 billion tokens a day to run my business and software factories. Can you help me? Can you teach me about AI?




And for fun, I ran this again now





My usage has actually gone UP. I'm averaging 5.0 billion tokens daily now.

Albert Wesker said:

You can pretty easily detect if it's AI content if they have dashes instead of complete sentences

There are other tools out there you can also do to make your content look as human like as possible.

Click to expand...

thats what the workflow is for...........................
to detect those patterns...

Asking AI not to write like AI is not working. Advanced detectors are based on token prediction and statistical patterns. Detectors like ZeroGPT basically have another underlying LLM doing the calculations on chunks of text. And any LLM does generate next token based on statistical patterns. You could try generating on lower confidence tokens but it would be trash.
Google just released synthid.com, you can check there. But in terms of SEO, AI detection is not as bad as you'd think. Even hand written articles are mostly polished with AI, so Google would not allow themselves to punish this content.

xReminisce said:

AI always tend to funnel towards certain ways or talking or describing things or just words in general.
like load-bearing is one of the biggest examples.

anything that claims to humanize text, i honestly stay far away from.
I have threw in harry potter and novels in "GPTZero, Originality.ai" etc and they gave me high chance of AI...
yeah, cuz AI was trained off that text.

For my own text i write, i run it thru some workflows that detects stuff including but not limited to Negative parallelism, composure verb, Over-explained subtext, over-states its own theme etc.
It's not perfect, but at the end of the day, neither are AI Detectors.

Click to expand...

Do you also use grammarly to do this? That is an interesting concept!

Nikolaos said:

Do you also use grammarly to do this? That is an interesting concept!

Click to expand...

No, i do not

I would be careful with these tools since most of the reviews look like their own promotion, and for university work it is safer to write it yourself and use AI only for ideas.

I’ve tested a few AI humanizers and I’m not a big fan of using them just to beat AI detectors. They can make the text sound a bit different but sometimes they also mess with the meaning and make good content worse. For academic work especially I’d focus on writing it properly in your own words and following the university rules rather than relying on a detector bypass tool.

The irony is even completely human written content is detected as AI and again writer has to put extra efforts in proving these AI dector tools wrong. For university assignments, I will still advise to get it done yourself and try to avoid AI writing style and pattern as much as possible.