Switch the
defaults off
One default costs little. Seven together cost a lot, because together they are what a template looks like. This is the inspector’s real arithmetic, on one hero section.
Each switch is the finding a scan would record, at its real weight. Alone, each default costs a few points; together they cost most of the rating, because together they are what a template looks like.
Supercharge your workflow with AI
Unlock the power of seamless automation for modern teams.
Every mark,
traced
The rating is never a verdict on its own. Each point lost is a finding you can read and check.
- WhereThe page, the section and the text or code it found, quoted from your site.
- WhyWhat makes it read as stock, in a sentence.
- What it is worthThe rating scored again without it everywhere it appears, never estimated.
- The fixAs a prompt for your AI builder, which tells it not to invent customers or numbers.
+1
Testimonials anyone could write
The testimonials section of a stock test page
- “Flowly has completely changed how our team works. I would recommend it to anyone! Sarah Johnson Product Manager”
Link each quote to the person or their company, or ask them for a detail with a number or a named feature. Never invent one; remove a quote you cannot back.
+1
Stock avatar faces
The testimonials section of a stock test page
<img src="https://i.pravatar.cc/150?img=1" class="h-10 w-10 rounded-full">
Show the real person's photo with their permission, or drop the avatar and keep the name, role and company.
+1
Stock phrases
The hero section of a stock test page
- “Say goodbye to”
Say the specific thing instead: who it is for, what it does, and one fact that proves it.
Five ratings
Higher is better. A site is rated section by section, then page by page, with the home page counting twice and up to ten points off when every page repeats one template.
65 things
it checks
It reads what your site serves: the HTML, the stylesheets and, for apps drawn in the browser, their compiled code. A fixed checklist, so the same pages always get the same rating. No model guesses at anything.
The squares show how much each one counts. Most count little alone; together they are what a template looks like.
The copy
Stock phrases, the overused words, contrasts nobody asked for, claims with no number or name in them.
- Stock phrases
- Overused words
- "Not just X, it's Y"
- Vague claims
- Emoji as bullets
- Round, unsourced numbers
- Stock calls to action
- "Fast. Simple. Secure."
- Heavy em-dash use
- Title Case Headings
- An arrow on every button
The proof
Testimonials nobody could check, initials in circles, overlapping faces beside five stars, a hero that never shows the product.
- Testimonials anyone could write
- "Trusted by thousands", no names
- No product in the hero
- Initials in circles instead of people
- Overlapping faces with a star rating
Left unfinished
Lorem ipsum, placeholder pictures, menus of dead links, the starter title. These hold the rating down.
- Template text left in
- Placeholder images
- Stock avatar faces
- Links that go nowhere
- Starter title or description
- Chat assistant leftovers
The look
Gradients clipped into headlines, glows, frosted cards, one word of the headline in another colour, sparkle and rocket icons, fake window dots, the stock palette.
- Gradient headline text
- Blurred glow blobs
- Coloured glow shadows
- Default purple button
- Indigo-to-pink gradients
- Frosted glass cards
- Coloured stripe on cards
- Icons in tinted squares
- Effects in every section
- Stock colour palette
- Unthemed component kit
- Only the default fonts
- Template-trend typeface
- Template glow tokens
- One radius on everything
- Dark page, grey body text
- This year's default palette
- One word of the headline styled apart
- A soft shadow under every card
- A pulsing status dot
- Monospace labels outside code
- Thin rules everywhere
- Sparkle, rocket and brain icons
- Fake window buttons
- Squeezed letter-spacing on headlines
- A glow that follows the pointer
The layout
The badge over the headline, identical card rows, a floating pill menu, tiny spaced capitals, every section centred, the template order.
- Badge pill above the headline
- Identical card grid
- All-caps label over every heading
- Centred hero
- Numbered 1 · 2 · 3 steps
- FAQ accordion
- Three tiers, "Most popular" in the middle
- The template section order
- Fade-in on every section
- Floating pill navigation
- Decorative 01 / 02 / 03 numbers
- Tiny letter-spaced capitals
- Every card lifts on hover
- Every section centred
- A whole sentence as the headline
- An endless scrolling strip
- A bento grid of rounded tiles
Before
you ask
- Does it tell me if AI made my site?
- No, and nothing can do that reliably. It rates how much of a site is stock: the things that appear unchanged on thousands of others. A hand-built page can rate low, and a page made with an AI builder can rate high once it is made your own.
- I use Tailwind and shadcn/ui
- That is fine. The kit is not marked down; its untouched defaults are. Set your own colours, corner radius and type on top of it and it rates like anything else.
- My site is a React app
- If the page is empty until JavaScript runs, Slop reads the app’s compiled code and rebuilds the pages from it, without running it. The report says when it did, because sections are then approximate.
- What is kept?
- Nothing. The rating is worked out while you wait and lives on your screen. Download it if you want to keep it.
- Can my coding agent use it?
- Yes. Paste the fix prompt into your AI builder, or have an agent call
POST https://preflight-api.agentmarket.lol/api/slop/checkwith the site’s address and read the streamed result. - Why ten pages?
- Ten are enough to see whether a site repeats one template everywhere, and they keep an inspection under half a minute on the free tier this runs on.