The Query System, Literature's Stochastic Toll Road

Four months ago, I published “Yes, It Costs $25,000 to Query Literary Agents”. Some people upset themselves by not taking the time to understand what I was saying, instead accusing me of setting up some kind of grift in which I would sell a cheaper way to get read by literary agents. If you’re looking to buy such a thing from me, I apologize for the lack of wares. I don’t have one, because there isn’t one. To be clear:

• I did not prove that it is impossible to query for free and get read. Some people have inherited connections or fantastic sob stories and will not have to spend a dime.

• I never established that everyone who spends $25,000 gets read and signed by a literary agent. On the contrary, I have said that spending $25,000 guarantees nothing. You can spend $100,000 on the query process and remain unread. The stochastic toll road is… stochastic.

• I did not advise—and do not advise—securing an agent through direct payment. If someone claiming to be a literary agent offers to read your work for a fee, you are probably being scammed. Reputable literary agents—the only kind you want, because they’re the ones who can get editors to read—care about your connections and platform, not your money.

The $25,000 figure doesn’t come from direct costs, but from the costs most people will incur in order to get the social status or leverage necessary to get fairly read by a literary agent. To be specific, the observed market rate of a 1% chance at a fair read by a literary agent is about $250. A conference that offers a 3% chance will typically settle around $750. A referral editor who introduces 20% of their clients to literary agents will typically charge $0.05-0.10/word more than an ordinary freelance developmental editor; as the read rate goes up, so does the price. MFA programs tend to be outliers, some being far more expensive than their read rates would suggest, and others being fully funded for those who get in. Platform buying has similar economics; inflating a follower count in a way that a literary agent won’t detect (they have software to vet numbers, and AI is advanced enough to teach them how to use it) is not cheap.

Lucky breaks, of course, sometimes happen. Some people attend one inexpensive conference, find a lost wallet, return it to its owner who happens to be one of the 25 literary agents who control half the award-eligible slots, and get signed that way. I wouldn’t advise planning one’s life on the basis of such windfalls, but they are not physically impossible. Market rates, of course, change. Fair reads by literary agents are not traded on stock exchanges, so we must estimate by observing people’s choices, and the data is thin enough that resolvable figures will always be 6–12 months out of date. In other words, the cost of querying may have gone down or up.

The system being assessed is opaque and adversarial by design. It was built to manage a low-trust society. You must query if you are not trusted to be able to write. And it was built to conceal how everything works, because publishing’s prestige requires it. The promise made to the public is that, whenever a writer with even 20 percent of my level of talent is spotted, agents and editors fight like cats in heat to connect the author’s work with readers. It ain’t true, and maybe it shouldn’t be true, because publishing is a business and has every right to make investment decisions on financial bases, and it might be the case that the search costs for serious talent exceed any commercial value in finding it. What infuriates me is the constant flagrant lying. See, I suspect that even if these people wanted to discover talent outside of closed networks, they’d struggle due to the sheer volume of submissions, and in this light it’s hard for me to fault them for having a system that doesn’t work, when I cannot prove I would be able to design a better one. The crime is not their failure to make text matter, but the coverup.

Two dozen people decide what the world reads. The evidence runs against their claims of superior taste and competence, but even if it did not, routing the bulk of literature through such a small number of people would create an untenable bottleneck. It is absurd. Publishing is an appeals court where, even if the judges were wise and fair, one would have to wait two hundred years for the docket to clear, at which point the question of one’s innocence becomes strictly historical.

And yet people exist who claim querying works. Online, in places where their agents and editors can see them being grateful, they insist that they have been served well. Some will claim they’ve spent less than $25,000 on querying, and since that figure’s an average, they’re not all lying. They’ll say they have “wonderful” agents, and will feel that way until they die on submission (which does not happen to those who secure the few agents with real clout) or get $5,000 “book deals” with no promotion and, following atrocious sales results, a need to change one’s name and probably facial bone structure. (Was Clavicular onto something? Better get that hammer.) A small percentage will, even though deals worth taking are usually reserved for people who haven’t had to rely on query letters since the Civil War, do well enough in self-publicity to maintain standing with their agents and editors, albeit at the expense of thousands of unpaid hours marketing, resulting in the loss of the day job they were wrong to believe they no longer needed. It’s not surprising that some people still believe in traditional publishing or the query process. Vegas knows this type well. It’s called “I have a system.”

Spoiler: You probably don’t. Poker is a feasibly profitable grind, but it is a grind, and the lifestyle of competitive play is brutal. Blackjack is winnable with flawless execution, but only a few hundred players can sustain an edge in noisy casino conditions, and those people tend to be on lists for immediate ejection. Roulette wheels these days are replaced long before they acquire tendencies. Slots? No, you don’t have a system. The machine generates millions of random numbers per second. You have selective memory; we all do. Casinos hire people who understand mathematics and psychology well enough to make you feel like you are winning when you are slowly losing. And even if you did have a system, you’d burn your edge on hotels and food.

A common topic of discussion in publishing is whether literary agents should bring back reading fees. The zero-dollar sticker price of querying—I’ve already discussed why the real cost burden is higher—often represents well the quality of service queriers will get. Alas, I don’t think reading fees will solve much. The payment doesn’t prevent rejection unread; it might mean you are charged to be rejected unread. Agents are not going to let you sit there and watch as they earn or do not earn that reading fee.

If fair reads were easy to get and cheap, everyone would try to get one, and publishing would not work if that were the case. The heuristic that acquisitions editors and literary agents use is the so-called “80 percent rule.” If you’re not accepting 80% of what you read, you’re wasting time that your competitors aren’t, so you’re not reading selectively enough. And this 80% figure matches what insiders get. Eighty percent of what is actually read will be accepted; the other 20 percent is held off for development. If you can convince these people you’re one of them, they will want you to win, and you’ll either get an enthusiastic acceptance or a detailed rejection with diagnostics and next steps. The rest of us are expected to be grateful to get a biased, dismissive skim seven months late. Will direct payment change that? I doubt it. The market rate observed in the indirect channels shows how willing some people are to pay to get read by literary agents. I suspect that bringing back reading fees would result in higher prices for the same shitty quality of service.

Currently, some people sink tens of thousands of dollars into conferences and referral editors and get nothing to show for it—for everyone who queries for free and gets signed, there are several who spend five or six figures and don’t. It seems like it would be an improvement to replace this stochastic toll road with a consistent price. Perhaps one could spend $7,500 for a guarantee that an agent will read a manuscript and recommend it highly to senior editors at top publishing houses; perhaps there could even be a menu in which authors decide with their credit cards which favors agents call in and how much begging and badgering they will do to get pages read in the right places. This seems like a great idea… until one realizes we have no way of predicting where the market rate settles. The inefficiencies of today’s indirect-payment system have costs, but efficient markets can result in higher prices, rather than lower ones, as an asset becomes more attractive to capital. The reason publishing is so expensive is not only that everything is inefficient (although that is part of it) but that people exist for whom it is worth $25,000—or $100,000, or $500,000—to be able to say they are published or famous or bestselling authors. You and I just are not in that world. Arguably, the root issue is rich people having too much money; if we could solve that problem, it would fix far more in our world than just publishing. This one seems to be out of reach at this time.

If literary agents handed out fair reads to everyone who wanted one, they wouldn’t be able to guarantee the 80 percent chance of acceptance once a manuscript is read. It would probably drop into the single digits. Readers might prefer this, if selectivity based on text (i.e., what is promised to readers) were actually delivered, but the people who currently have careers in publishing would view the change as a rug pull—a replacement of the system they paid to be able to control by one that doesn’t follow the rules they agreed, long ago when they got established, to play by. Text and its quality, from the perspective of people who work in or rely on traditional publishing, are variables that no one really wants on a balance sheet. Text is impossible to de-risk. Two printers in 1631 had an off-by-one-word error in “thou shalt (not) commit adultery” and it ruined their lives and we’re still talking about it. If readers could be convinced to buy covers and author photos and blank pages, publishing would prefer it.

So how does one fix all this? I don’t know. Reading fees would replace the current system of indirect payment for access and social proof with a fee-for-service model that would be more honest, but I don’t know if it would be more functional. I suspect there would be agents offering $500 reads to send back very kind, very detailed rejections written by Claude Opus 4.8, as well as agents offering $50,000 reads because they can actually get a job done… as well as agents offering $50,000 reads who would not get the job done. You’d have to spend the money to find out which you had bought. The current embarrassing mess would be replaced by some other embarrassing mess. I don’t see improvement, because I don’t know how to restore trust to traditional publishing when it has gone without it for so long. Writers don’t trust agents to read. I’ve shown statistically that they shouldn’t. The query system exists because agents don’t trust writers to be able to read; I haven’t seen their slush piles, so I can’t evaluate whether this is the right call.

In the late 2010s, publishing’s inside discussion was how long it would take for writers to start seeking agents-for-agents. Everyone knew the system was so dysfunctional no elegant solutions existed, but a new level of indirection can always be added. What no one could agree on was how one would go about securing these second-degree agents. The reason you could, as recently as the mid-1990s, show up at a literary agent’s office and be confident you’d walk out represented was that most people didn’t know that literary agents existed, because the Internet was in its infancy. The informational barrier made the system work. The view within publishing was that it would take less than six months for the unwanted querying masses to discover these second-degree agents, making the new system as overtraveled and dysfunctional as the old one. And so no one deliberately created a second-order agent system, but we get one through the reliable execution of self-interest. The ex-agents offering manuscript assessments for $0.15 per word, who will at least have you believe they keep in contact with the Manhattan sausage-makers, have become, to be tediously precise about it, that new class of intermediate intermediary whose existence keeps the myth of public access alive.

Then came November 2022. At the end of that month, ChatGPT was released. It was rudimentary by today’s standard; the LLM’s competence at literary assessment was laughably bad. Nearly four years later, AI does not write at a literary level, and there is no reason to believe that will change—AI is often creative and it is often competent, but its competence is never creative and its creativity is never competent. And yet its skill at literary assessment has improved; it is not excellent, but I will show later this summer that Claude Opus is no worse than a leading literary agent. A level of reading for quality that was once reserved for a few hundred people with generational connections can now be attained for about $3.50 in tokens. Of course, it still can’t offer a book deal. Yet.

We don’t know what the future looks like. Publishers (or algorithms) will inevitably discover how much better—and cheaper—LLMs are than the two dozen humans trusted to decide what the world reads, two dozen humans who save most of the slots for their friends. This is predictable, ordinary economics. What I cannot say is whether this will leave us—and literature—better off. In a capitalist system it is inevitable to be ruled by people whose interests are opposed to one’s own. I don’t find it likely, but we could find ourselves missing the era in which all it took to become a published author was to hire the right referral editor on Reedsy. It’s too early to say what AI curation means for literature. We don’t have it yet, and the only reason we even want it is that the apparent alternative, what we have now, is so expensive, inefficient, duplicitous, and tasteless.

The only advice I can give is never to pay a toll one cannot afford.