You sent the draft and asked what they thought. They wrote back that it was clear, thoughtful, and "in a good place." You still do not know if the design will fail in production. You asked for a compliment with extra steps. They were polite. Polite is not a review.
A review hunts defects in something already built. It is the one shape where a long answer is usually a good answer. An opinion is a lean plus a reason. A decision is a pick. If you do not name the shape, people default to the cheapest thing that still feels like a reply: kindness.
Kindness is the attractor
Most readers do not want to hurt you. "Thoughts?" gives them an exit that preserves the relationship and skips the work. They praise the parts that are fine and stay quiet on the part that is wrong, because naming the wrong part is a bigger social risk than you think it is.
The reply mirrors the ask. If the ask was unbounded, the reply will be too — warm, vague, useless for changing the work.
Say the unsocial thing: you want what is broken. You want it first. You can take a no.
Point at the work, and freeze the rest
A review without an artifact is a vibe check. Point at the brief, the spec, the options already on the page. Say what is not in scope — "the deadline is not movable; tell me what in the design fails that date" — or they will review the premise you already decided, which feels helpful and wastes the turn.
Do not also ask them to pick. A decision and a review in one letter get you a compliment plus a hedge. Split them. Review first, decide second, or decide first and ask for defects in the chosen path.
What you put in the briefing
The thing to review, in enough detail that they do not need the session. The questions you already closed. The assumption that, if wrong, would make the review moot. A note in your own words: "I want the holes, not a polish pass."
Approve that note. If the assistant wrote "I'd love your thoughts," change it. That sentence is how you get the compliment.
They may have no AI. They still need a page they can mark up in a reply. They do not need a share link to a chat where you and a model already agreed the draft was great.
How you know it worked
You receive a list. Some of it stings. You can act. That is a successful review.
You receive "nice work." That is a successful compliment. It is a failed review. Send one more message, shorter, naming the shape, if the work still needs the holes. Do not send the same "thoughts?" again and hope they become braver. They answered you the first time.