Last November, I walked into Powell’s City of Books in Portland with a vague idea that I wanted something to read on a long flight. I didn’t have a specific book in mind. I didn’t have a genre preference. I just wanted something good.

The person behind the counter, whose name tag said “Marianne,” asked me three questions. What was the last book I loved? What was I in the mood for, something heavy or something light? And was I okay with being surprised?

I said: Stoner by John Williams. Medium weight. And yes, surprise me.

She handed me a copy of An Artist of the Floating World by Kazuo Ishiguro. I’d never read it. She said: “If you liked Stoner’s quiet desperation, you’ll like what Ishiguro does with memory and regret. It’s a different register, but the emotional territory is similar.”

She was right. It was exactly what I wanted. And no algorithm on earth would have made that connection.

The recommendation machine

We live in an age of algorithmic recommendations. Amazon tells you what to read based on your purchase history. Goodreads surfaces books based on your ratings. TikTok’s algorithm can turn an obscure novel into a bestseller overnight if enough people film themselves crying over it.

These systems are powerful. They’re also limited in ways that matter.

Algorithmic recommendations work through pattern matching. They look at what you’ve bought or liked and find other things that people with similar profiles have bought or liked. This produces useful results much of the time. If you enjoyed one thriller by a certain author, you’ll probably enjoy another thriller by a similar author, and the algorithm is good at making that connection.

What algorithms can’t do is understand why you liked something. They can see the surface patterns: genre, author, subject matter, price point. They can’t see the deeper currents: that what you loved about that memoir wasn’t the subject matter but the prose rhythm, that the reason you liked that novel wasn’t the plot but the way the author handled silence between characters, that you’re in a specific emotional place right now and need a book that meets you there.

A good bookseller can see those things. Not because they’re mind readers, but because they’re trained observers of reading habits, and because they’ve read widely enough to make connections that cross the boundaries algorithms use to sort books.

What a bookseller actually does

I’ve spent time in independent bookstores across the country talking to the people who work there, trying to understand what they know that computers don’t. The conversations have been some of the most interesting I’ve had about books.

At Books Are Magic in Brooklyn, a bookseller named James told me that his job is “translation.” A customer comes in and describes what they want using vague, approximate language. “Something like that book about the woman in the lighthouse, but funnier.” His job is to decode that description, which might not correspond to any actual book, into a recommendation that satisfies the underlying desire.

At The Elliott Bay Book Company in Seattle, a staff member described her approach as “reading the reader.” She watches how people move through the store. Do they head straight for a specific section, or do they wander? Do they pick up books and read the first page, or do they flip to the back cover? These behaviors tell her something about what kind of reader she’s dealing with, and that informs her recommendations.

At Parnassus Books in Nashville (the store Ann Patchett co-founded), the staff maintains what they call institutional memory, a collective knowledge of which books work for which kinds of readers. A customer who loved one book five years ago comes back, and the bookseller remembers that and builds on it. It’s a running conversation that unfolds over years, and it produces recommendations of startling accuracy.

None of this can be replicated by an algorithm. Not because the technology isn’t sophisticated enough, but because the underlying skill is human in a way that resists quantification. It’s empathy applied to reading. It’s pattern recognition that includes emotional and aesthetic dimensions, not just behavioral ones.

The problem with “readers also bought”

Amazon’s recommendation engine is, by most accounts, the most sophisticated in the retail world. It processes enormous amounts of data and produces suggestions that are often relevant. But “relevant” and “inspired” are different things.

The structural limitation of collaborative filtering (the technique behind “readers also bought”) is that it pushes you toward the popular center. It recommends books that lots of people buy, because that’s the data it has the most of. The books it’s worst at recommending are the unusual ones, the ones that don’t fit neatly into an established category, the ones purchased by a small number of readers with eclectic tastes.

These are precisely the books that a good bookseller excels at recommending. The odd, the unexpected, the book that came out three years ago from a small press and never got much attention but is perfect for this particular reader at this particular moment.

There’s also the feedback loop problem. Algorithmic recommendations tend to narrow your reading over time. You buy a mystery, so the algorithm recommends more mysteries. You click on those, so it recommends even more mysteries. Before long, your entire recommendation feed is mysteries, and the algorithm has no reason to suggest the memoir or the science book or the poetry collection that might open a new door for you.

A bookseller does the opposite. A good one will deliberately push you outside your comfort zone. Not aggressively, not presumptuously, but gently. “I know you usually read thrillers, but have you tried this? It’s a novel about a family in Lagos, and it has the same propulsive quality you like in thrillers, but it does something different with it.”

That kind of recommendation is an act of generosity. It requires the bookseller to know both the book and the reader well enough to make a connection that isn’t obvious. When it works, it’s one of the great pleasures of being a reader.

The shelf talker as art form

One of the things I love about independent bookstores is the shelf talker: that little handwritten card tucked next to a book with a staff recommendation. They’re a minor art form, and the best ones do in three sentences what a marketing department spends thousands of dollars trying to do.

I’ve been collecting favorites for years. A few that stuck with me:

“I’ve pressed this book into the hands of eleven people this month and I’m not stopping.” That was at a store in San Francisco, next to a copy of Piranesi by Susanna Clarke.

“This will make you want to call your mother. Or maybe never call her again. Depends on your mother.” That was at a store in Chicago, next to a memoir I won’t name because the description is the whole recommendation.

“If you think you don’t like poetry, read page 42. Just page 42. Then we’ll talk.” I read page 42. We talked. I bought the book.

These tiny reviews work because they’re personal. They’re one human being saying to another: I read this, and it did something to me, and I think it might do something to you too. No algorithm produces that. No publisher’s marketing copy achieves it. It’s the bookseller’s unique contribution to the reading ecosystem, and it’s irreplaceable.

At ScrollWorks, we rely heavily on these bookseller recommendations. When Elena Marsh’s Still Waters was first published, the marketing budget was modest. What built its readership was booksellers who read the advance copy, connected with it, and started recommending it face-to-face. That organic process of one reader telling another is still the most powerful marketing force in publishing, and booksellers are its primary agents.

Curation versus aggregation

The fundamental difference between a bookstore and Amazon is the difference between curation and aggregation.

Amazon aggregates. It has everything. That’s its value proposition: whatever you want, it’s there. The catalog is infinite, and the search and recommendation tools help you navigate it. This is genuinely useful. If you know what you’re looking for, Amazon is hard to beat.

A bookstore curates. It has a finite number of books, chosen by people with taste and judgment. The constraint is the value. When you walk into an independent bookstore, someone has already done the work of filtering the half-million titles published this year down to the few thousand on the shelves. Every book in the store is there because a human being decided it was worth stocking.

This is a radically different shopping experience. In an aggregated environment, the burden of choice is on you. You have to know what you want, or at least be able to describe it to a search engine. In a curated environment, the burden of choice has been partially lifted. You can browse. You can discover. You can pick up something you’ve never heard of and trust that it’s there for a reason.

The curation model depends on the expertise of the curators, which brings us back to booksellers. The quality of an independent bookstore is, to a large extent, the quality of its staff. Stores that hire readers, that encourage their staff to read widely and develop opinions, produce better curation. Stores that treat bookselling as unskilled retail labor produce worse curation. The difference is enormous.

The bookstore as social infrastructure

There’s another dimension to independent bookstores that has nothing to do with buying books, and it’s worth naming.

Bookstores are gathering places. They host readings, book clubs, community events, children’s story hours. They’re places where strangers can start conversations about ideas. They’re anchors in neighborhoods, the kind of “third place” (not home, not work) that sociologists keep telling us we need more of.

When a neighborhood loses its independent bookstore, it loses more than a place to buy books. It loses a piece of social infrastructure that’s hard to replace. Amazon can deliver a book to your door in 24 hours. It can’t host a conversation about that book with your neighbors.

We’ve done author events at independent stores across the country, and the energy in those rooms is different from anything you get at a chain or online. When James Whitfield read from Echoes of Iron at Literati Bookstore in Ann Arbor, the Q&A lasted an hour. People didn’t want to leave. They wanted to talk about history, about storytelling, about the research process. That conversation happened because a physical space existed where it could happen, staffed by people who cared enough to organize it.

The economics of hand-selling

Let’s talk money, because none of this works if it doesn’t pencil out.

Independent bookstores have had a complicated economic trajectory. They declined sharply in the late 1990s and 2000s under pressure from chains and Amazon. Then, against most predictions, they started growing again around 2009. The American Booksellers Association reported increases in membership for over a decade. By the early 2020s, there were more independent bookstores in the U.S. than there had been in years.

This recovery happened partly because indie stores figured out what they could do that Amazon couldn’t. They leaned into curation, community, and the in-store experience. They became less like retail outlets and more like cultural institutions.

But the margins are still thin. A typical independent bookstore operates on a net margin of one to three percent. That’s a business where everything has to go right just to break even. A bad holiday season, a rent increase, a construction project that blocks foot traffic for three months: any of these can be fatal.

The booksellers who work in these stores are poorly paid by almost any standard. They do the job because they love books and they love connecting people with books. This isn’t a sustainable model if we want talented people to keep doing this work, and the industry needs to grapple with that honestly.

What readers can do

If you value what independent bookstores provide, and I hope this piece has made a case for why you should, here are some concrete things you can do.

Buy from them. Even if the book is a dollar cheaper on Amazon. Even if you have to wait a day or two for a special order. The money you spend at an independent bookstore stays in your community in a way that money spent on Amazon does not.

Use Bookshop.org for online orders. It’s not perfect, but it routes a portion of each sale to independent bookstores and provides a meaningful revenue stream for stores that don’t have their own online presence.

Ask for recommendations. Walk in without a plan and say: “What’s good?” You’ll be surprised how much a bookseller can do with that question. Let them guide you somewhere unexpected. Take the risk. The worst case is you read a book you didn’t love. The best case is you discover something that changes you.

Attend events. Show up for readings and signings, even for authors you haven’t heard of. These events are often the most interesting precisely when the author is relatively unknown, because the conversation tends to be more intimate and less performative.

Tell people about your bookstore. Recommend it the way you’d recommend a good restaurant. Social media is fine for this. So is just telling a friend. The stores that thrive are the ones that their communities actively support and talk about.

The human element

I keep coming back to Marianne at Powell’s, handing me that Ishiguro novel. What she did in that thirty-second exchange was process information that no algorithm has access to: my body language, my tone of voice, the specific way I described what I was looking for, and, probably, something intuitive that she couldn’t have articulated herself. The recommendation she made wasn’t just accurate. It was generous. It said: I see you as a reader, and here’s something I think you’ll love.

That’s what a bookseller knows that algorithms don’t. Not more about books (Amazon knows more about books in aggregate than any human being ever could). But more about reading, about what it means to put a specific book in front of a specific person at a specific moment. That’s a human skill, and it’s one worth preserving.

We publish books because we believe they matter. Booksellers put those books into the hands of the people who need them. Without that last step, the whole chain is incomplete.

Written by Sarah Chen, Senior Editor at ScrollWorks Media.