Getting there...
UPDATE 7: More features and updates have been implemented, like the option in the AI Humanizer to select the editing style (Light, Balanced or Agressive). Giving the tool now 4 stars as I see may improvements and bug fixes. Keep up the good work!
- - -
UPDATE 6: I like the changes and fixes but found a new issue in my history. Some humanized text gets mixed up with other text, screenshot: ela.d.pr/i/KxSziw
Maybe take a look, thanks.
- - -
UPDATE 5: Thanks for the fixes and adjustments. The result page looks now better and one can see the original and humanized text. Will the limit of 750 words be lifted at one point?
Suggestion, for an easier workflow: Could there be a button (after one uses the Detect AI Generated Text) on the result page to humanize the text right away? This spares an extra click to get to the AI Humanizer page first and click the Eliminate AI.
PS: Pumping up my rating to 3 tacos as we're making progress here. :)
- - -
UPDATE 4: I have run another test with an AI generated text from Claude trying to humanize it. First it shows the word count is 838 while in reality its 876 (confirmed in two different apps)! Anyway, when trying to humanize the text - Eliminate AI - then nothing happens. Is this another glitch / issue you working at right now? Please confirm.
- - -
UPDATE 3: Thanks for getting back. What you describe sounds very complicated and unnecessary from my point of view. The process should be:
1. AI Detector - Detect AI Generated Text (Original Text - which might be generated by AI)
2. AI Humanizer - Make original text more human
3. AI Compare - Comparison of Original Text with Humanized Text
If that could be in one workflow, that would be perfect. Because what you describe or what the functionality is right now this is not possible. As mentioned in the screencast the is only a comparison of the humanized text and some extra AI generated version but not the original text. You see this in the video!
One other point - Are you planning to have an API so one can use Activpievces or Boost.space (Make) for automation?
PS: I am not able at the moment to take / make calls as I am travelling remote with limited internet speed / access. ;)
PS2: Changing my rating to 2 tacos as I see you are trying to fix and improve polygraf and listing to feedback. :)
- - -
UPDATE 2: I think you don't understand what I meant earlier. For some reason, is Polygraf changing the ORIGINAL text when I click trough to comparison after humanizing the text. I made a recording so you better understand what I experience: ela.d.pr/v/H1EMjc (Copy this link) Please look into this and get back to me. Cheers.
- - -
UPDATE: Thanks for looking into this. There is still something weird with the results and when clicking trough to the comparison text. See my screenshots 1. ela.d.pr/i/yEGNAl and 2. ela.d.pr/i/nPgjxL Not sure why but suddenly the original text has a better human score than before. đ¤ Actually, it looks like that the text gets altered again even with me not doing anything.
- - -
Hi, I have bought the deal and had two documents checked. Both of them identified the scanner as AI generated, so I used the humanizer. After completion, I have used the compare tool with the original text, but it shows me the same scores for both text. Also, the text is identical (original text) but now has better scores in terms of sounding human. Can you check on that or let me know that you work currently on updates which fixes this so I can test this again? Using the same name here as in Polygraf if you want to check the account.
Yagub_PolygrafAI
Edited Jun 18, 2025Hi Elias,
Thanks for the depth analysis, this is much appreciated.
Word increase: During the coming 2 weeks, we are planning on adding unlimited (max token input) words for humanizer.
Suggestion: I like your suggestion, we will look into it.
History comparison: Thanks noticing this. It is not mixing data, rather browser based clash of functionalities - which should be fixed today.
ASK: Since we have updated and given new functionalities - can you fix the title of your review?
Cheers,
YR
---------------------------------
You are my favorite QA person right now!
- the issue you had was due to some sort of a adblocker - disabling a pop-up, which has been resolved and deployed yesterday.
- we deployed an updated model yesterday for the humanizer too - expect to see in the near future the ability to control the heat (how much do you want to humanize?). Currently it is set to humanize at 30-60% range, depending on the topic to bypass spam filters and social media bots.
- I appreciate you taking screen recordings, it made the browser related issues much easier for us to eliminate.
Thank you.
Previous update:
I/We understand what you are talking about, we have designed the model as the following:
Step 1: AI generated content -> Step 2: AI Generated content (action is complete)
If you press compare, you are triggering the system edit functions
You can see that it opens a new project, and then pressing compare one more time it humanizes the content (Step 3) as it is a new project now, and places next to the (Step 2).
We have a different compare logic right now, but for the next week, the roadmap includes:
Original vs. Step 2(humanized) vs. Step 3 (humanized) comparison
The current comparison is Step 2 vs. Step 3
There is always a learning curve for everything, and that is why we welcome these and other feedbacks. We have also offered to see your comments in a call before (see your email) but we received no response.
Polygraf is built for insights generation, we have 17 different model running behind the scene each handling one specific task, therefor requires a different approach.
____________________
Elias, the grammar score is not an AI score. Your Cognitive flow is at 72%, which is already great. Sometimes, humanizing a human generated or already humanized content creates more failures.
Additionally, the update is still pending. We are testing it before it is pushed live, it should be live within the hour.
Update:
Hi Elias
You found a browser related bug, we are on it and will be fixed within 30 minutes, appreciate you bringing it to our attention.
As a thank you we are granting you a 100,000 advanced tokens.
Thank you for your support and hope you can update your score after the bug is fixed.
âââââ
Hi Elias,
This is a user error or a browser caching error.
The humanizer did its job correctly, however on the compare page (either you are inputting or your browser is blocking the new input) the compared data is the same one.
Since some part of these computations are handled on browser level, it is recommended to use chrome, and not to have ad blockers.
Iâd be happy to look into it with you on a zoom call. Can you send a meeting inquiry on our website please?
Also as a thank you to your feedback we are adding 10,000 advanced tokens to your account which will unlock some of the advanced features too.