Question 1

How does SlopCheckr calculate the slop score?

Accepted Answer

In two stages, and they are not equally reproducible. First, extraction: we fetch your public URL and read the document — headings, button and link labels, paragraphs, CSS class-name frequencies, declared colours and font families. Some paths additionally render the page in headless Chromium. A set of detectors then searches that document for specific CSS and utility-class signatures: the gradient function used for glow backdrops, the symmetrical three-column card grid, the fully rounded pill button, the blur behind glass panels, and a fixed dictionary of marketing clichés. The exact strings each detector matches are published at /methodology. That stage is deterministic — the same page returns the same measurements. Second, scoring: each detector that fires adds a fixed penalty, and the headline score is clamp(4, 96, floor(6 + penalty × 0.82)). A language model then writes the critique, the copy rewrites and the buyer jury; that part varies between runs. The score runs 0 (nothing detected) to 100 (everything fires). Because the detectors match strings in your markup rather than rendered pixels, a site built without those utility-class names can carry the same pattern and still score clean — that limitation and the rest are published in full at /methodology.

Question 2

How long does a scan take?

Accepted Answer

We are not going to quote a number we have not measured. A scan launches a headless browser against your URL, so most of the wall-clock time is your page loading, and a run that includes the model pass takes longer than one that does not. We have not instrumented this end to end yet. When we have, the median and the spread will appear here with the sample size and the date, the same as every other figure on this site.

Question 3

Why do AI-generated websites convert poorly?

Accepted Answer

We do not have a bounce-rate figure we can stand behind, so we are not going to quote one. The mechanism is what matters: when a visitor recognises the pattern — the dark purple glow, the centred "Empower your workflow" headline, the three identical icon cards — they stop asking whether your product solves their problem and start asking whether your company is real. That is a much harder question to answer in the first viewport.

Question 4

There are free, open-source slop scanners. Why pay for this?

Accepted Answer

Because they answer a different question, and for some people the free answer is the right one. A pattern scanner fingerprints your CSS against a fixed catalogue of tells — default font stacks, gradient hero text, glassmorphism, pill buttons — and hands you a number. That is genuinely useful, it is fast, and if a score and a list of triggered pattern names is what you need, use one. What a fixed catalogue cannot do is look at your page. SlopCheckr renders it, screenshots it, and scores the two hundred largest elements on six weighted signals — area, contrast, position, typography, isolation and colour rarity — to show where attention actually lands rather than where you intended it to. It then puts the page in front of four buyer profiles (a staff engineer, a finance buyer, a design critic, and someone who arrived from an ad and gives you five seconds) and reports what each one would do and what stopped them. That last part needs a language model reading your page, which is exactly what a tool promising deterministic, zero-cost scanning has ruled out for itself. A scanner tells you your CSS looks generated. We tell you what a buyer thought and where their eye went.

Question 5

My design is deliberately minimal. Will it be flagged as slop?

Accepted Answer

Possibly, and you should know why before you trust the number. The detectors match strings and computed styles, not intent: a restrained page that happens to use Inter, a centred hero and a rounded call-to-action will accumulate penalties even when every one of those choices was deliberate. The inverse is also true — a site built with the same patterns under different utility-class names can score clean. That is a real limitation of string-and-style matching rather than a bug, and it is published in full at /methodology alongside the exact strings each detector looks for. Treat the score as a fingerprint of how conventional the page reads, not a verdict on the people who made it. The written critique, the attention map and the buyer jury are where the judgement actually lives.

Question 6

Does SlopCheckr need access to our private GitHub repository?

Accepted Answer

No. SlopCheckr works purely through read-only public DOM extraction, exactly as a visitor or a browser sees your site. You do not need to install an npm package, provide API keys, or grant repository access.

Question 7

Can I use the recommendations with Cursor, Claude Code, or v0?

Accepted Answer

Yes. Every audit generates prompts formatted for Cursor, Claude Code, v0.dev, and Bolt.new. Copy the generated prompt and paste it into your editor to perform the refactor.

Question 8

What is the difference between the free tier and Pro?

Accepted Answer

The free tier gives you 3 audits per month with the global slop and craft scores, the five-dimension breakdown, and the top 2 detected findings. Pro ($15/mo, or $9/mo billed yearly) covers up to 150 audits per month and unlocks all detected code diffs, line-by-line Tailwind patches, the predicted salience map, the blind swap test, and white-label PDF exports.

Question 9

Can I integrate SlopCheckr into GitHub Actions or call it from the API?

Accepted Answer

Yes. There is a public v1 API with Bearer token authentication and pre-built GitHub Action workflows. You can configure pull-request checks that audit staging preview deployments (Vercel, Netlify, Cloudflare) and fail the PR above a slop threshold you set. The interactive documentation is at /docs.

Question 10

Can agencies use SlopCheckr for client audits?

Accepted Answer

Yes. That is what the Agency plan is for: run audits across a prospect list and export branded teardown dossiers you can put in front of a client before you pitch a redesign. There are also 10-audit packs ($29 one-time) if you only need a handful.

Field	Flagged	Standard
`visualOriginality`	Radial purple and cyan blur orbs	A solid ground with structural borders that carry the hierarchy
`typographyDiscipline`	Default unkerned Inter or Roboto at display sizes	A display face with optical tracking, paired with tabular mono for data
`copyOriginality`	Superlatives with nothing behind them — empower, seamless, all-in-one	Concrete mechanics: latency figures, protocols, the specific pain removed
`layoutAsymmetry`	Three identical cards in a row, repeated down the page	Asymmetric column layouts with density where the content warrants it
`brandVoice`	Hyper-rounded gradient pill CTAs	Components whose shape and weight match the rest of the interface

You built it with AI. Buyers can tell.

Put your landing page in front of four simulated buyers

Staff infrastructure architect

How the scan works.

The page is rendered and photographed

The measurements are scored

The fixes are written out

What the report scores

Audit localhost:3000 before you launch.

Predictable pricing.
Zero hidden fees.

Just need a one-off audit for a single website?

Need programmatic API access or automated CI gates?

Questions

Audit your frontend before your buyers do.