SitePrompts alternatives: when 230 words is not a brief

A SitePrompts prompt runs about 230 words, in four labelled parts: an opening brief, a layout section, a design tokens section carrying hex values and a type scale, and a list of constraints.

The tiers are the same length. Premium prompts average 228 words and free ones average 229, so the paid decision is not a decision about depth. The whole library comes to roughly 5,475 words across 24 prompts, about eleven double-spaced pages, which is the short end of the eleven libraries ranked by how long their prompts run.

Two things follow, and they point in opposite directions. Part of that 230 is the most useful forty words most prompts never carry. The rest of the page is about what the other blocks would have said.

The part a short brief gets right#

Prohibition is the highest-leverage instruction in a website prompt, and their constraint blocks use it properly.

• A checkout prompt bans pre-checked upsell boxes and a countdown timer.
• A portfolio prompt bans a hero section, a stock photograph and skill bars.

That is worth more than three paragraphs of adjectives, because a model will reach for exactly those defaults and a sentence forbidding them costs nothing to write. Their pricing page also carries a proof pair: the same fictional brief through the same builder, once as one line and once as a structured prompt, with a screenshot of each. The second is visibly better, and showing that rather than asserting it is the right way to sell this category.

Ten of their prompts are readable in full with no account and no card. Read one before you read the rest of this page, because everything below is easier to judge with a real specimen in front of you.

What you are choosing between#

What 230 words hands back to the model#

A brief is not judged on length. It is judged on which decisions it takes off the model's desk. Here is what a 230-word brief typically leaves on it, and what arrives instead.

1. The type scale. With no ratio and no breakpoint behaviour, you get the default ramp, which is a display size that looks fine at 1440px and collapses into the body copy on a phone.
2. The motion. "Smooth animations" is not an instruction. Without a named curve and a duration in milliseconds, entrances land at the library default, and 300ms on a structural move is the timing that reads as cheap.
3. The section order. Absent an order, the model produces hero, features, testimonials, pricing, call to action, in that sequence, every single time.
4. The never list. Nothing forbidden means a pure mid-grey palette, Inter as the display face and a stock gradient, which is the clearest tell in generated design.

None of those four are exotic. They are the decisions a designer makes in the first hour, and they are the ones that separate a page that survives a client presentation from a page that gets rebuilt.

The nine blocks, itemised#

This is what a Thanor brief contains, block by block, and it is the part worth stealing whoever you buy from.

1. Concept. One sentence naming what the page is arguing, so every later choice has something to be right about.
2. Art direction. Named colour values, at least four, written as #0F1317 and never as "deep charcoal".
3. Typography. A display face and a text face by name, a scale with its ratio, and the size the display resolves to at phone width. Inter and Space Grotesk are banned outright, because they are the default reach of every generated design of this period.
4. Layout and rhythm. A spacing step, a grid, a maximum measure, and the vertical rhythm between sections stated as a number.
5. Motion. Named curves with real values, durations in milliseconds, structural moves held between 600ms and 1200ms because slower reads as expensive, and a prefers-reduced-motion fallback that actually changes something.
6. Sections. Every section in order, with what it contains and which one carries the weight.
7. Assets. Where images and video come from, and what to do when they are not supplied.
8. The never list. The defaults the model must not reach for, named.
9. Acceptance and self-check. A list the output is measured against before anybody calls it done.

A brief with nine blocks and exact values runs past 1,200 words because the values take that long to write, not because length is the goal.

Where short is the right answer#

This is the part a comparison page usually skips, and skipping it would make everything above less credible.

• A single section does not need 9,000 words. Rewriting one pricing table is a constraint list and a token block, and that is it.
• A fix is not a build. If the model already produced something close, the shortest instruction that names the delta wins.
• Their newest prompt runs 3,257 words, with 42 numbered sections, real CSS blocks, a stagger table, eleven named breakpoints and a responsive acceptance checklist. It is free, and it is a better argument for structured briefs than anything on this page.

Length is a symptom. The thing to look for is whether the prompt ever names a number.

The free layer, which is most of the market#

• Their own ten, in full, no account.
• Several sites publishing a hundred or more, free. Four of the five results on the front page for the obvious search give the product away.
• The builders themselves. Bolt ships web design guidelines and visual edits, Lovable extended design systems to every paid plan, v0 runs a free design mode. All of them enforce consistency with tokens you already have, and none of them authors the tokens.

So which one#

• Want a short, clean, four-part brief with a genuinely good constraint block, and ten free specimens to judge it by? SitePrompts is that, and the free tier means you never have to guess.
• Want the free layer? It is real, it is large, and for a component or a single section it is usually enough.
• Want every block filled in, with a named curve at a real duration, a type scale that resolves at phone width and a section order that was chosen rather than defaulted? That is what Thanor writes, and it is one brief you paste whole.

Read a Thanor brief next to the free ones above and the difference takes about ten minutes to see.