AI, the Slush Pile, and the Chances of Getting an Agent/Publisher

It should come as no surprise that there are Large Language Models that have been trained to evaluate manuscripts (LLM’s are what most of us are referring to when we talk of AI in this age of Chat GPT). More than merely examining grammar, these models are tuned for genre. Thus an agent can ask “Does this manuscript meet all the requirements of a good cozy mystery?” and get a detailed answer. I’m guessing the analysis would identify both the elements that are satisfied and those that are deficient.

For many authors, LLMs are the evil that should be put back into Pandora’s box and never considered again. There are a swath of applications where I agree with this sentiment. But there are at least an equal number of applications where I disagree. Asking a LLM to provide feedback on your manuscript is one of them. Publishers doing the same is another. I think this is a positive development for genre authors and at worst a neutral one for literary works.

Here’s why: submitted manuscripts in a slush pile likely get at best a cursory look. I’m willing to bet that just one thing that seems off in a submission letter is enough to see the corresponding manuscript routed directly to the trash bin. If someone told me that a large number of submissions never get read, or don’t get read past the first few pages, I would not be surprised. Similarly, I wonder how many really good works got missed because they didn’t nail the opening or the outline.

LLMs fix that. Now an agent/publisher can pass every single submission through an automated, detailed analysis. They can look at a short report and see that the opening needs work but after that, there’s a well-structured story. Even non-genre works can get a reading on the author’s command of grammar. Sure, an outlier like Naked Lunch is still screwed, but I submit that even without AI something in a similar vein would never get read in this day. That work is screwed, AI or no AI.

For authors, this means that the odds your manuscript will be “read” if even by an LLM, will soon reach 100%. Instead of a human reading your work (or more likely dismissing it after a precious few pages), the person will be reading a detailed evaluation of the structure and calibre of your work. Hit a sufficient number of positive points, and they’re going to take the time to sit down and read the whole thing, knowing that at least what they’re looking at is worth the effort.

That has to be a good thing.