Websiteprompts.ai alternatives: industry or art

Websiteprompts.ai publishes more than one hundred website prompts, free, and its FAQ is explicit that they are free for personal or commercial projects. The catalogue is organised by trade: Esthetician. Towing. Junk Removal. Painter. Recording Studio. Farm. Appliance Repair. The prompts are aimed primarily at Lovable and Bolt and work anywhere a model can read a paragraph.

That naming convention is the most informative thing about the library, and it is worth reading carefully. The unit is an industry, not a look.

The verdict before the detail: if the question is what should go on a towing company's website, an industry prompt answers it well and costs nothing. If the question is why does the result look like every other site generated this week, the answer sits one layer up, and that layer is Thanor.

What an industry name settles#

More than people expect, and less than they need.

Name a trade and a competent model already knows the content requirements. Given towing, it will produce a service area, a phone number high on the page, a list of vehicle types, hours, and a form. Given esthetician, it produces treatments, a price list, a booking link and before and after imagery.

That is genuine value and it is the hardest part to fake. A prompt written by somebody who has actually built sites for trades carries information gathering that a generic brief does not.

What the trade name settles:

• Which sections exist at all.
• What each section is called.
• The rough priority of the calls to action.
• The vocabulary and the tone.

What it does not settle:

• The type scale, and what the display size resolves to at 1440px.
• The palette, as values rather than as warm and trustworthy.
• The motion: which curve, at which duration, on which elements.
• The order those sections actually run in, and which one carries the weight.

The second list is what a visitor reads as design. Two towing sites with identical content and different answers to the second list are two different businesses to anyone who lands on them.

The library also offers a style filter, with options along the lines of light, minimal and dark. That is a useful sort order and it is still a pointer rather than a value. Minimal is agreed on as an idea and disagreed on at every number inside it: the spacing rhythm, the weight of the type, how much air sits above the fold, whether anything moves at all.

The alternatives, side by side#

Names on the left. The third column is the argument.

One column that does not empty, and it is not an accident of how this table was written.

Every product above is built to apply a decision faithfully. None is built to make one, because a tool with a strong house style is a tool nobody can point at a client brand. That is the correct product decision, and the whole market has taken it.

The consequence is a gap in the workflow rather than a defect in the software. Somebody supplies the opinion, or a model supplies the median of everything it has read, which on the web is a centred hero, three feature cards, a soft fade upward on scroll and a weightless sans at one of four sizes.

That is why so many pages rhyme, whichever library the prompt came from.

Free is not the axis#

The instinct is to compare free against paid, and it is the wrong axis. Plenty of free prompts are better than plenty of paid ones, and price predicts almost nothing about depth.

The axis that predicts the output is precision. One line against a full specification. Not length, precision: a long prompt full of adjectives is still a short prompt in the only sense that matters.

Run the test yourself on any prompt, free or paid. Search it for:

1. A hex value. Not deep navy, an actual #0B0B0F.
2. A ratio or a size. Not large headlines, a stated scale resolving to a stated size at a stated width.
3. A duration in milliseconds, attached to a named curve.
4. A stated section order, with a reason each section sits where it does.

Most prompts in circulation, at any price, contain none of the four. The ones that contain all four produce pages that do not look like each other, because the model has stopped guessing.

That is the same test we hold ourselves to, and it is the right one: a paid library has to clear the free layer by a visible margin or it has no reason to exist. The margin is not the count of prompts. It is whether the four items above are present on every single one.

The number nobody prints#

There is a structural reason the precise layer is scarce, and it is visible in the search index.

Across roughly 110,000 indexed URLs belonging to Framer, Webflow, Lovable, Replit, Cursor and Windsurf combined:

• easing appears in three URLs.
• cubic-bezier appears in zero.

Framer runs a design dictionary several hundred entries deep, with pages for easing, bezier-curve, spring-animation, kerning, leading and baseline. Every entry defines the term properly. Every entry stops before the number.

A builder cannot publish the specification, because the specification substitutes for the editor. Print cubic-bezier(0.22, 1, 0.36, 1) at 620ms beside #0B0B0F and a 1.25 type ratio resolving to 72px on desktop, and the reader reaches the result in any tool on the market without opening the canvas being sold.

So the layer has to come from somewhere that is not a builder. It is its own category of product, with its own unit: decisions, written at a precision a machine executes without interpreting.

Prompt libraries are the natural home for it. Most of them stop at the content layer because the content layer is what people ask for, and the art direction layer is the one they cannot name until they have shipped four sites that rhyme.

The tell is easy to spot once you look for it:

• A prompt that reaches the motion layer says entrance at 620ms on cubic-bezier(0.22, 1, 0.36, 1), staggered by 80ms down the column.
• A prompt that stops short says smooth scroll animations, which covers a 200ms linear fade and a 900ms overshoot equally well and pins down neither.

Which one to use#

Three answers for three situations.

1. You need to know what goes on the page. Take the free industry prompts. They encode real information gathering about a trade, they cost nothing, and the commercial terms are stated openly. The free layer here is enormous and worth using.

1. You need parts. Take the free component registries. A component and a decision solve different halves of the job, and picking one while needing the other loses a weekend.

1. You need the page to look like somebody decided. This is the layer above both, and it is where the industry prompt stops on purpose. A towing prompt that hard coded one type scale and one palette would be a worse towing prompt, because it would only suit the third of towing companies whose brand happened to match.

The verdict: use the free prompt for the content and bring the art direction separately. The two stack cleanly and they do not compete. An industry prompt tells the model what a towing company needs on the page. A specification tells it the palette in hex, the type scale at a stated ratio, the curve and the duration, and the order the sections run in. Paste both and the same builder produces a site that is correct for the trade and unmistakably not the fourth one generated this week.

That stack is also the cheap answer, because the expensive half is free and the half you pay for is written once and reused on every project after it.

Browse the Thanor library and read one full specification before your next build.