Most recommendations are useless in a specific and fixable way: they describe the recommender's emotional response rather than the work. Beautiful, moving, incredible. A reader learns that somebody enjoyed something and nothing about whether they would. The fix is to describe the experience of reading instead, which takes the same number of words.
This applies to anybody writing about work, whether in a review, a shop listing, a post, or a message to a friend. It is also what makes a recommendation trustworthy over time, which is the thing that determines whether anybody acts on the next one.
Describe the experience
What is it actually like to read? Slow and quiet, or dense and fast. Mostly conversation, or mostly images. Funny in a dry way, or not funny at all. These are checkable properties, and a reader can match them against what they want without needing to share the recommender's taste.
Concrete detail carries this best. One specific thing that happens, one image, one exchange, tells a reader more about the register of a work than any amount of general praise, which is the specificity argument that runs through the piece on what makes an explicit scene work.
Say who should skip it
Naming the readers who will not enjoy something is the single most effective way to make a recommendation credible. It demonstrates that the recommender has thought about fit rather than simply enthusiasm, and it saves the exact people who would otherwise have been disappointed.
It also costs nothing. Almost every work has an audience it does not suit, and saying so plainly does not reduce the number of people who read it. It increases the number who read it and were glad they did.
Comparisons need a shared reference
Saying a work is like another work is efficient and only functions if the reader knows the reference. Comparing something obscure to something else obscure is a common failure that reads as knowledgeable and communicates nothing.
The stronger form names the specific quality being compared: like this other work in its pacing, not in its subject. That gives a reader something usable even if they do not know the reference, and it forces the recommender to identify what they actually mean.
Avoid the plot summary
Summarising events is the default and it is nearly useless, because plot is the least distinctive thing about most work and because a summary flattens tone entirely. Two books with identical summaries can be completely different reading experiences.
A single sentence of situation is enough, followed by what the work is like. That ratio, one line of what and several of how, is what separates a recommendation from a listing.
Declare the stake
If you know the creator, worked on it, or received it free, say so. It costs a clause and it removes the possibility of somebody discovering it later, which damages trust in everything else you have ever recommended.
This is the transparency point made about rankings in the piece on what best-of lists measure. A stake does not disqualify a recommendation; concealing one does.
Recommendations in a shop listing
The same rules apply when describing your own work, with the difference that saying who should skip it feels commercially risky. It is not: it reduces refunds, complaints and mismatched readers, which is the signalling argument in the piece on erotic and pornographic modes.
Being specific about register, length and content also improves discovery, since it puts the words readers use into the description, which is the point made in the piece on how readers search.
Length and commitment are information
Readers are deciding how much of their time to hand over, and almost no recommendation says how much is being asked. Whether something is forty pages or four hundred, finished or ongoing, and whether an ongoing work updates reliably are all decisive facts for somebody choosing between several options, and all of them are usually omitted.
Stating them takes a clause and removes a whole category of frustration. An unfinished serial recommended without that context is the commonest complaint readers make about recommendations, and it is entirely avoidable by anybody willing to spend six extra words.
Writing something a reader can use
Describe what reading it is like, name who should skip it, compare only on shared references and state what quality you mean, keep the summary to a line, and declare any stake. Five habits, and they turn enthusiasm into information.
Corrections and suggestions are welcome through the letters page.