How to Tell If That Post of Advice Is AI Bullshit
Right, I wasn't going to write more on this, but every time I block an obvious AI-driven blog, five more clutter up the tags. So this is my current (April 2024) advice on how to spot AI posts passing themselves off as useful writing advice.
No Personality - Look up a long-running writing blog, you'll notice most people try to make their posts engaging and coming from a personal perspective. We do this because we're writers and, well, we want to convey a sense of ourselves to our readers. A lot of AI posts are straight-forward - no sense of an actual person writing them, no variation in tone or text.
No Examples - No attempts to show how pieces of advice would work in a story, or cite a work where you could see it in action. An AI post might tell you to describe a person by highlighting two or three features, and that's great, but it's hard to figure out how that works without an example.
Short, Unhelpful Definitions - A lot of what I've seen amount to two or three-sentence listicles. 'When you want to write foreshadowing, include a hint of what you want foreshadowed in an earlier chapter.' Cool beans, could've figured that out myself.
SEO/AI Prompt Language Included - I've seen way too many posts start with "this post is about..." or "now we will discuss..." or "in this post we will..." in every single blog. This language is meant to catch a search engine or is ChatGPT reframing the prompt question. It's not a natural way of writing a post for the average tumblr user.
Oddly Clinical Language - Right, I'm calling out that post that tried to give advice on writing gay characters that called us "homosexuals" the entire time. That's a generative machine trying to stay within certain parameters, not an actual person who knows that's not a word you'd use unless you were trying to be insulting or dunking on your own gay ass in the funniest way possible.
Too Perfect - Most generative AI does not make mistakes (this is how many a student gets caught trying to use it to cheat). You can find ways to make it sound more natural and have it make mistakes, but that takes time and effort, and neither of those are really a factor in these posts. They also tend to have really polished graphics and use the same format every time.
Maximized Tags (That Are Pointless) - Anyone who uses more than 10 one-word tags is a cop. Okay, fine, I'm joking, but there's a minimal amount of tags that are actually useful when promoting a post. More tags are not going to get a post noticed by the algorithm, there is no algorithm. Not everyone has to use their tags to make snarky comments, but if your tags look like a spambot, I'm gonna assume you're a spambot.
No Reblogs From The Rest of Writblr - I'm always finding new Writblr folks who have been around for awhile, but every real person I've seen reblogs posts from other people. We've all got other stuff to do, I'm writing this blog to help others and so are they, the whole point of tumblr is to pass along something you think is great.
While you'll probably see some variation in the future - as people get wise to obviously generated text, they'll try to make it look less generated - but overall, there's still going to be tells to when something is fake.
I don't have any real advice for what to do about this (other than block those blogs, which is what I do). Like most AI bullshit, I suspect most of these blogs are just another grift, attempting to build large follower counts to leverage or sell something to in the future. They may progress past these tattletale features, but I'm still going to block them when I see them. I don't see any value in writing advice compiled from the work of better writers who put the effort in when I can just go find those writers myself.
dededos
areudirty
janainaart
theartofmadeline
titsay
