Where can i find real OnlyFans chat conversations?

S

Singularity1

Newbie

Sep 2, 2026

26

13

• Oct 4, 2026

• #1

The participant names can be anonymized. I am thinking about going into AI OFM, so i need it for research purposes.
Google and ChatGPT weren't much help. The best I could find is the kassyy-chatter-v2 dataset on huggingface, but it is very small, and it is from the same chatter. I need a much bigger sample size.

 

T

tobzy33

Newbie

Aug 27, 2026

11

2

• Oct 4, 2026

• #2

Singularity1 said:

The participant names can be anonymized. I am thinking about going into AI OFM, so i need it for research purposes.
Google and ChatGPT weren't much help. The best I could find is the kassyy-chatter-v2 dataset on huggingface, but it is very small, and it is from the same chatter. I need a much bigger sample size.

Click to expand...

The dataset you found is fine for learning structure, but real conversion have slang drift, callbacks and pacing shifts across days that's what's missing and why a bigger sample of the same chatter won't fix it

 

Ksatriaseo

Junior Member

Sep 8, 2020

128

29

• Oct 5, 2026

• #3

Real customer chats are usually private, so I wouldn’t use leaked or scraped conversations even with names removed.

Better option is to study public creator FAQs, captions, comments, and consented training examples, then build your own chat library from real conversations only after getting clear permission.

 

G

Garlick

Registered Member

Jun 12, 2026

73

23

• Oct 5, 2026

• #4

Yeah, finding a big clean dataset for that is probably the hard part. Most real OFM chats won’t be public in bulk.

I’d probably mix whatever public datasets you can find with synthetic conversations and use that mainly to study patterns, not copy exact chats.

 

J

Ji Xao

Newbie

Sep 8, 2026

15

2

• Oct 6, 2026

• #5

have you looked at datasets outside the OF niche? general customer support and sales chat datasets can still teach a loot about conversational flow.

 

opentab

Newbie

Sep 29, 2026

17

5

• Oct 6, 2026

• #6

Tag your own synthetic set by fan stage (new sub, went quiet after a PPV, haggler) and test it on one small live account, because a static dataset can't show what actually gets a fan to reply