Many people copy/paste output from ChatGPT to their websites so careless that you can spot the
regenerate response
by searching for it in Google.
@Ammon mentioned the source of this finding being last week’s Pubcon preso by Jennifer Slegg (author of The SEMpost) and also shared in on her Twitter TL.
This thread is for collecting some of the hilarious examples 
Yup, the source was Jennifer Slegg (author of The SEMpost) who shared it at PubCon, and then reshared it via Twitter too.
Here’s a really cool advanced Google query to find these examples
for german speaking area (by Paul Schreiner), you can adapt accordingly
intext:"regenerate response" site:*.de OR site:*.at OR site:*.ch -intitle:chatgpt -intitle:AI
e.g. to only include .gov and .edu sites in the US in such a search
intext:"regenerate response" site:*.gov OR site:*.edu -intitle:chatgpt -intitle:AI
Reveal umhb.edu as “offender” again… Why become a teacher a UMHB
When someone is pretending to be tech-savvy enough to even talk meaningfully about Python, yet is so incredibly technically inept as to leave default text in… You have to see that as a massive demerit on credibility for the whole site, right?
Incidentally, the reason this kind of stuff is worth noticing is that that simple error, of leaving the “regenerate response” text in there turns it all into clearly identifiable training data. By taking other sentences from any of these examples, you can then often find other AI generated texts that hadn’t left the obvious, but were easy to spot once you had the patterns to look for. This is, after all, how machine learning itself works in most cases - a set of prequalified training data than you then let the machines spot the patterns within, to detect further types of the same in the wild (the texts without “Regenerate response” but talking about the same keywords or concepts, will often use some of the exact same sentences in them).
