I just wish we could have room for nuance on this. No one should be trusting any automated detection tool blindly. There are also certain patterns of results that are likely meaningful and worth being part of trying to not let AI slop infiltrate artistic and writing spaces.

The eye test and some directly plagiarized lines make it abundantly clear, but it's also a case where every published piece from them in the last year flags at 100% on Pangram.

It's super fucked up how all the major public allegations and news stories have targeted black and POC authors. I'm actually working right now on a piece about two white American authors who appear to be scamming dozens of lit mags with AI work.

Thanks for sharing all this. I've been looking for more proven examples of false positives on Pangram; this certainly appears to be one given the source, though I looks like on a very short excerpt; I don't believe any of his broad published work gets falsely flagged at least.

Nah, you got challenged for posting questionably true statements and making overly broad claims out of specific truths, and for not responding to any of the reasonable critiques and responses and just doublign down.

Putting out an open call again for scan results showing provable false positives from Pangram, especially on creative writing. In extensive searching and now thousands of my own tests, the closest I have found is one instance of two sentences from a provably 2018-written story being falsely flagged.

This is NOT me saying it does not have any false positives. But as publishing world increasingly uses it as part of a process of trying to fight back against AI slop, it seems vital to understand real cases where it fails.

There are lots of vague claims out there ("it flagged a pre-AI novel" but without naming the novel, showing the results, etc.). I really WANT to find concrete, proven examples.

Of note—especially given the massive undue combination of suspicion/allegation/publicity currently falling on POC writers, I'm currently working on a piece about two white American writers who appear to be scamming dozens of literary mags over the last year with AI.

The evidence is overwhelming on multiple levels, from eye test to directly plagiarized phrases; it is also backed by every single piece from each of them over the last year flagging at 100% on Pangram, which again, if anyone can show ANYTHING like that happening as a false positive, I want to see.

The evidence is overwhelming on multiple levels, from eye test to directly plagiarized phrases; it is also backed by every single piece from each of them over the last year flagging at 100% on Pangram, which again, if anyone can show ANYTHING like that happening as a false positive, I want to see.

Of note—especially given the massive undue combination of suspicion/allegation/publicity currently falling on POC writers, I'm currently working on a piece about two white American writers who appear to be scamming dozens of literary mags over the last year with AI.

There are lots of vague claims out there ("it flagged a pre-AI novel" but without naming the novel, showing the results, etc.). I really WANT to find concrete, proven examples.

This is NOT me saying it does not have any false positives. But as publishing world increasingly uses it as part of a process of trying to fight back against AI slop, it seems vital to understand real cases where it fails.

I am not saying I have that confidence. I am saying a claim was made here that is provably false, and that the only thing anyone shared (you) pointing to it is TRANSPARENTLY SLOP, every sentence, without the need for any AI detector, just literally any experience having to review AI output.

Putting out an open call again for scan results showing provable false positives from Pangram, especially on creative writing. In extensive searching and now thousands of my own tests, the closest I have found is one instance of two sentences from a provably 2018-written story being falsely flagged.

I do not waste time dealing with transparently AI slop writing... I am 100% confident by eye that no one wrote that page, and as I've shown here, the claim it is making is immediately provably false.