Publishing software provider Trilogy has launched Manuscript AI, a tool that uses artificial intelligence to evaluate unsolicited manuscripts and helps analyze slush piles for commissioning editors and literary agents.
According to Trilogy, the system—which can run as a standalone software-as-a-service platform or integrate with Trilogy's existing Title Management product—analyzes manuscripts in minutes and generates a report that focuses on four key areas:
- Sales Potential, which estimates commercial performance on a 0-100 scale by comparing the manuscript's content against actual sales figures from bestsellers in the genre over the past 20 years
- Predicted Rating, which forecasts reader satisfaction on a 1-5 star scale by analyzing the text against published books with known Amazon and Goodreads ratings
- Genre Fit, which measures if the manuscript's content matches its stated genre (e.g., flagging a romance where no love interest appears until 60% through the book)
- Style Fit, which analyzes writing mechanics including sentence structure, dialogue distribution, emotional tone, and pace, comparing these against successful books in the genre (e.g. a thriller with long, flowery sentences would score poorly because its style wouldn't match how most bestselling thrillers are written)
The tool then gives an overall Manuscript Score on a 100-point scale intended to help editors prioritize which submissions are worthy of further review. Explanatory text in the system's interface note that "each manuscript is different, and we recommend adding a human touch to assessments and not blindly using these scores alone."
"Our focus is basically the slush pile—the 94% to 97% of manuscripts that never get published," Alex Dare, managing director of Trilogy Group, told PW. "That's a waste of intellectual property and a waste of potential. An average commissioning editor can process a couple of manuscripts a day. We can do the same job in a couple of minutes."
Trilogy trained the system on approximately 2.7 million copyright-cleared and public domain books stored in a database in Switzerland. The company said it follows data integrity protocols and does not retain manuscripts unless explicitly given permission to use them for training purposes.
"We look at the genre and we won't try to come up with an answer unless we've got a minimum of 10,000 books in that genre," Dare said. The system covers major categories including memoir, nonfiction, drama, romance, and science fiction.
Trilogy said it sells Manuscript AI licenses primarily to medium and large publishers, but is exploring offering distinct licenses for the tool. "We have been talking to associations about being distributors," Dare said, noting that one association with 4,000 members—2,000 of whom are small publishers—is evaluating the service.
Manuscript AI is entering an increasingly competitive AI-driven manuscript assessment marketplace, which includes products specifically aimed at helping publishers deal with slush, such as Libraro and Storywise. But Dare says that overall, Trilogy can afford to take a long-term view on the application of AI technology.
"We're fine. We have income. We make our money, so we can experiment with this," he said. "I think this will take a couple of years as this market evolves.”
