Most reading advice is about how to read more. Almost none of it is about how to decide what to read next, which is strange, because that decision determines the value of every hour that follows it.
Reading a mediocre book quickly is worse than reading a good book slowly. The selection problem is upstream of everything else.
Why the default method fails
The default method is drift. You finish something, you look at your shelf or your library app, and you pick whatever appeals in that moment.
This fails for a specific reason: the version of you choosing is the version that just finished a book. If that book was excellent, you will reach for something similar and get diminishing returns from a mode you have just exhausted. If it was a slog, you will reach for something undemanding as relief, which is often the right call but tends to become a habit.
Either way the decision is being made by a temporary emotional state rather than by anything about what you actually want from reading.
Recommendation algorithms make this worse rather than better, since they are explicitly built on similarity. Optimising for what resembles what you have already read is a good way to narrow steadily without noticing.
The question that works better
Instead of asking what you want to read, ask what you want reading to do for you over the next few weeks.
The answers are usually one of a small set. You want to learn something specific. You want to be absorbed and escape. You want to be challenged and pushed. You want company, meaning a voice you enjoy spending time with. You want to finish something, because you have not completed a book in a while and the momentum matters more than the choice.
These are genuinely different jobs and they call for different books. The mistake almost everyone makes is choosing a book for one job while wanting another. A demanding literary novel picked up when you needed absorption will fail at both, and you will blame the book.
Running more than one at once
The single change that improves this most is keeping several books going simultaneously, chosen to serve different jobs.
This gets treated as a lack of discipline, which I think is exactly backwards. You do not read the same way at every hour of the day. Attention at nine in the morning is a different resource from attention at eleven at night, and matching the book to the available attention is straightforward optimisation rather than a character flaw.
A workable arrangement: something demanding for whenever your attention is genuinely sharp, something absorbing for evenings, something you can read in fragments for gaps and commutes.
The failure mode is running too many, at which point none of them accumulates enough momentum to stay alive and they all quietly stall. Three is manageable. Six is a graveyard.
The list problem
Almost everyone who reads seriously maintains some list of books they intend to get to, and almost everyone’s list has become a source of low-grade guilt rather than a useful tool.
The reason is that lists accumulate without any mechanism for removal. A book goes on because it sounded interesting once, and then stays for years while the interest that put it there evaporates.
What helps is treating the list as a queue with a limit rather than an archive. When something new goes on, something old comes off. This forces you to notice that you no longer want to read most of what is on there, which is the useful information the list was hiding.
It also converts the list from a record of obligations into a record of current intentions, which is a very different object to look at.
Where recommendations actually help
Not from strangers, and not from lists of books everyone should read.
The recommendations that work are from a person who knows what you have read and liked, and who tells you specifically why they think this one is for you. That last part is the whole thing. A recommendation with reasoning attached lets you evaluate it. A bare title does not.
The other reliable source is following a writer backwards. If a book worked for you, the acknowledgements, the influences, the writers the author keeps citing are a far better lead than any algorithm, because they represent an actual intellectual lineage rather than a purchase pattern.
On reading things that are not for you
A deliberate practice worth having: occasionally read something well outside what you would normally choose, selected specifically because it is outside.
Not as self-improvement. As calibration. It is very easy to develop confident opinions about a genre you have not read in a decade, and the only correction is going back into it.
The frequency that seems to work is roughly one in ten. Frequent enough to keep the map updated. Rare enough not to feel like homework.
The practical version
When you finish something, wait a day before choosing the next one. The temporary state passes and the choice improves.
Then ask which job you need filled rather than which book you want. Check whether you already have something on hand that fits, before adding anything new. And if what you actually need is momentum rather than quality, pick something short and easy without apology, because a finished short book restarts a stalled reading habit better than a great long one that stays unfinished.
What I have not solved
Rereading.
I know that rereading is often more valuable than reading something new, particularly for books that landed hard the first time and which you have since changed enough to meet differently. I know this and I still almost never do it, because the pull of the unread is stronger than the argument for the reread.
If you have found a way to make rereading a genuine habit rather than a good intention, I would like to know what it looks like in practice, because every system I have tried for it has quietly collapsed within a month.