Editors now use algorithms to sort through book pitches

Every year, Paula Munier receives approximately 10,000 unsolicited queries. That's not a rounding error or a metaphor for "a lot" it's the actual count from an agent who has been behind since her very first week in the business. The pile never shrinks. The reading never ends. And somewhere in that relentless flow, good manuscripts arrive on the wrong day, opening pages stall where the story finds its footing by chapter three, and query letters miss while the prose sings.

This is the structural reality of editorial workflow at scale: the volume of submissions has outpaced the industry's ability to evaluate them effectively. Around 16 million unsolicited manuscripts are submitted to agents annually, according to research cited by literary historian Laura B. McGrath, with roughly 94-97 percent never reaching publication. The commonly cited industry benchmark suggests that just 1-2 percent of submitted manuscripts are accepted. But that statistic, while accurate as a rule of thumb, only tells part of the story. Behind that number lies a much larger challenge: are publishers rejecting bad manuscripts, or missing valuable ones?

The answer has driven a quiet revolution in how editors approach the slush pile not through hope or hunch, but through systems. Systematic triage. Structured analysis. And now, increasingly, AI-assisted evaluation that promises to do in minutes what used to take an editor an entire morning.

The Two-Question Problem

Froukje, writing for EditBook.ai's editorial blog, describes the core cognitive trap that undermines most slush pile reading: editors conflate two distinct questions. "The goal of a first-pass read isn't to decide whether you love a book," she writes. "It's to decide whether a book is worth the time it would take to find out. Those are different questions, and conflating them is part of what makes submission reading so exhausting."

This distinction matters because loving a book requires reading it fully. But determining worthiness for a full read doesn't. A structural analysis can evaluate whether the narrative arc is coherent, whether pacing holds across chapters or stalls in the middle, whether the narrator's voice is consistent or shifts in ways that suggest the author hasn't found their footing yet. None of that tells an editor whether they'll love the book. But it tells them whether the manuscript is ready for a full read and that's a useful filter before committing the time.

The problem is that traditional slush pile reading happens under conditions designed to defeat careful evaluation. Editors are tired. Fridays arrive. A manuscript arrives at the wrong moment and gets a three-minute read instead of a thirty-minute one. The system, such as it is, starts making decisions that have more to do with attention than with quality.

"Good manuscripts arrive on the wrong day. Opening pages are slow, but the story finds its footing by chapter three. The query letter is awkward, but the prose is assured. I don't think there's a way to eliminate that problem entirely reading takes time, and time is the one thing an editorial department is always short of." EditBook.ai editorial analysis

Systematic triage doesn't eliminate the problem. But it makes the first pass more consistent, so fewer manuscripts fall through the gaps for the wrong reasons.

The Signal vs. Noise Architecture

From an operational standpoint, slush pile management is a signal detection problem. The signal manuscripts with commercial or literary potential is often buried, not absent. The challenge is finding it under layers of noise: submissions that are not viable, queries that miss their mark, manuscripts that arrive without research into whether they fit the publisher's list.

WriteStats, in their analysis of publisher slush pile management strategies, identifies several structural reasons why good manuscripts get rejected for non-quality reasons. Market misalignment ranks high: a manuscript may be well-written but fail to align with current genre demand, publisher imprint strategy, or seasonal acquisition priorities. A strong dystopian novel may struggle not because of any deficiency but because the market has moved on.

The evaluation process itself introduces additional noise. Gut instinct, time constraints, and individual editorial preferences all influence decisions made quickly sometimes within minutes. High-potential manuscripts can easily be overlooked not because they're weak but because they arrived at the wrong moment, or because the query letter failed to convey their strengths.

Laura B. McGrath, the literary historian and data scientist who has been researching the profession of literary agents, crystallized the dynamic in remarks at a Book Industry Study Group panel: "While the writer might see the slush pile as an access problem, for an agent it is a volume problem. There is simply no way to begin to address that volume." Her framing shifts the perspective from what authors experience the frustration of sending work into a void to what agents actually face: an industrial-scale processing challenge with human-scale time constraints.

The Human-Machine Partnership

In 2026, trade publishers use AI at the acquisitions stage for three specific tasks: first-read triage on submissions, verification of comp titles in query letters, and market-fit scoring against backlist performance. Purpose-built tools Trilogy Manuscript AI, Storywise, Familiar launched in 2025-2026 specifically for high-volume slush triage.

The landscape is evolving rapidly from where it stood in early 2025, when Jane Friedman's industry snapshot observed that AI in publisher submissions was "lots of experimenting but no implementation yet." Fifteen months later, that framing no longer holds. At least one major institutional product has launched.

Publishers Weekly reported in January 2026 that Trilogy, a publishing software provider, launched Manuscript AI a tool that evaluates unsolicited manuscripts and helps analyze slush piles for commissioning editors and literary agents. The system analyzes manuscripts in minutes and generates a report focusing on four key areas:

• Sales Potential: 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: Forecasts reader satisfaction on a 1-5 star scale by analyzing the text against published books with known Amazon and Goodreads ratings
• Genre Fit: Measures whether the manuscript's content matches its stated genre, flagging issues like a romance where no love interest appears until 60 percent through the book
• Style Fit: Analyzes writing mechanics including sentence structure, dialogue distribution, emotional tone, and pace, comparing these against successful books in the genre

The tool then provides 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 notes 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. 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." Alex Dare, managing director of Trilogy Group, via Publishers Weekly

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.

What AI Can and Cannot Evaluate

The Manuscript AI for Trade Publisher Acquisitions: 2026 Playbook from ManuscriptReport.com identifies a clear pattern in how AI is being integrated into publisher workflows: the integration that's working pairs AI for first-draft analysis with human editorial judgment for the acquisition decision. General-purpose large language models like ChatGPT and Claude remain unreliable for comp-title verification work due to documented hallucination and stale-data failure modes.

The clearest evidence that pre-acquisition AI detection isn't yet operational at Big Five scale is the Shy Girl cancellation. Hachette's Wildfire imprint canceled debut novelist Mia Ballard's book in March 2026 after Reddit-driven analysis flagged AI-pattern text post-publication. The book had launched in the UK in November 2025; Ballard claimed a freelance editor had added AI text without her knowledge. The intake processes didn't catch it. Readers did.

This case illustrates both the promise and the current limitations of AI in editorial workflow. AI tools can process vast quantities of submissions and flag structural issues, market positioning, and style consistency. But the opening the moment when a reader decides whether to keep going, whether the world of the book is one they want to spend time in remains fundamentally human.

As EditBook.ai's Froukje observes: "The one thing analysis doesn't help with is the opening. Not structurally you can measure sentence complexity and readability but in terms of whether you want to keep going. Whether the world of the book is one you want to spend time in. That's a feeling, and it's yours. No analysis replaces it."

This is where systematic triage shifts from efficiency play to quality assurance. What analysis provides is more confidence about the manuscripts being set aside. When a structural report shows fragmented pacing, an inconsistent narrative voice, and no clear market position, an editor can pass with more assurance that they're not missing something. And when a report shows a well-constructed manuscript with a clear voice and a legible premise even if the opening pages haven't grabbed the editor yet they know it's worth going further before deciding.

The 10,000-Query Challenge: A Practitioner's System

Paula Munier's experience as an agent offers a window into how human practitioners develop their own systematic approaches when technology hasn't yet arrived to help. Her slush pile numbers 10,000 unsolicited queries per year create a volume problem that demands structure.

"Only 1 in 200 queries is well-written enough, well-conceived enough, and well-targeted enough to prompt me to ask to see more material," she notes. The reason is often apparent: many writers send one-size-fits-all queries, set up mail merges that include every agent in Literary Marketplace, and demonstrate no knowledge of the specific agent they're approaching. "The salutation 'Dear Paula Munier' is a dead giveaway," she observes.

Her system for beating the odds involves research and relationship-building that transforms unsolicited queries into something closer to solicited material. Writers who attend conferences, participate in online and offline pitch contests, or engage with agents on social media gain the ability to reference previous contact in their subject lines. "Requested material from Bouchercon" or "Nice meeting you at the Boston Book Festival" catches an agent's attention in ways that cold queries cannot.

This isn't about favoritism or insider access it's about the same signal-vs.-noise dynamic that drives AI adoption. When an agent can reference a previous encounter with a writer, the query provides additional context that helps evaluate fit, voice, and potential. The query itself becomes more informative because the writer has done research.

The Systematic Triage Framework in Practice

Based on documented editorial practices and emerging AI tools, a systematic triage framework for managing high-volume submissions typically operates across three stages:

The framework isn't about replacing human judgment at any stage. It's about ensuring that human judgment operates on the manuscripts most likely to reward it. The first pass filters for structural readiness. The second pass evaluates craft. The third pass the one that actually requires reading happens only when the preliminary signals suggest the manuscript has potential.

What This Means for SubmitArticle Readers

For readers researching editorial workflows, submission systems, and the intersection of human and AI evaluation in publishing, the systematic triage approach offers several practical insights.

First, the volume problem is real and structural. Whether you're an editor managing submissions, an author navigating the query process, or a platform designer building submission infrastructure, understanding that evaluation operates under time constraints shapes every decision about workflow design.

Second, AI tools are entering the editorial pipeline but haven't replaced human judgment. The current generation of purpose-built tools like Trilogy Manuscript AI can analyze structure, genre fit, style, and market potential. But they cannot evaluate whether a reader will want to spend time in a book's world that remains a human decision.

Third, query letters and submission research remain high-leverage points for authors. The data suggests that well-targeted, researched queries perform significantly better than mass submissions. This isn't about gaming a system it's about meeting editors where they are in a workflow designed to handle volume.

Fourth, the distinction between "worth reading" and "will love" matters for anyone building or refining submission systems. Conflating these questions leads to exhaustion and inconsistent evaluation. Systematic triage separates them, allowing editors to apply the right evaluation method at each stage.

The Road Ahead: Integration and Accountability

Regina Brooks, president of AALA and founder and president of Serendipity Literary Agency, has experimented with AI-driven submissions tools and believes they could eventually eliminate the query letter as we know it. Her reasoning: AI can evaluate writing styles, compare submitted manuscripts to successful published books, and scan work for weaknesses like pacing issues or overused tropes. The tools may help writers produce better submissions and eliminate the struggle to craft the perfect query.

But experimentation hasn't yet become implementation at scale. The Shy Girl case demonstrates that current intake processes don't reliably catch AI-generated content. Readers either through algorithmic platforms like Reddit's analysis tools or through post-publication review remain a final quality checkpoint that publishers can't yet fully automate.

McGrath's framing offers a useful orientation for the industry as it navigates this transition: "I'm not an AI enthusiast. I'm also not a doomsayer. I like to think about AI less as artificial intelligence, which suggests a replacement of human labor, and instead something closer to augmented intelligence, which is a way of extending our abilities and a way of extending our reach."

This is the direction the systematic triage approach seems to be heading: not AI replacing editors, but AI handling the volume problem so that human editorial judgment can focus on the manuscripts that deserve it. The slush pile remains vast. The time remains limited. But the gap between signal and noise is narrowing, one structured evaluation at a time.

Where to Read Further

For readers wanting to explore these themes more deeply, the following sources provide direct access to the practitioners, data, and frameworks discussed in this article:

• EditBook.ai's analysis of triage without losing hidden gems Froukje's detailed breakdown of the structural evaluation approach to first-pass reading
• WriteStats' publisher slush pile management guide comprehensive coverage of signal-vs.-noise dynamics and hidden gem identification
• Publishers Weekly's report on Trilogy Manuscript AI the full picture of how AI manuscript assessment tools are being built and deployed
• Jane Friedman's industry snapshot on AI in the slush pile the February 2025 baseline that shows how quickly the landscape is shifting
• ManuscriptReport.com's 2026 AI acquisitions playbook practical guidance on how trade publishers are actually using AI at the acquisition stage
• Career Authors' query game guide Paula Munier's practitioner perspective on navigating the slush pile from the author's side