Start a job

Our standard

Every site and landing page we build is held to written standards for the site, its design and its copy, and an audit judges your client’s site against the site standard. This page sets out the numbers the work must meet and where each one comes from, so you can weigh it up before sending a job.

AI models build the work. Automated tests check what software can measure, and a separate AI reviewer that didn’t build it checks the rest against the written standards. Our process sets out which does what at each stage of a job.

The site

The numbers a page has to meet before it comes to you.

Speed is measured with Lighthouse’s mobile test, which simulates a slow phone: a 150 ms round trip, a 1.6 Mbps download and a processor slowed four times. In the first four rows below, our builds have to meet a stricter line than the official one, or come with a written reason for any page that misses it.

CheckOfficial lineOur builds
Lower is better
Main content shown (LCP)2.5 s2 s
Layout shift while loading (CLS)0.10.05
Time blocked from responding (TBT)200 ms150 ms
Higher is better
Lighthouse performance score9095
Tap target size24 px44 px
Contrast of normal text4.5:14.5:1

The official lines are Google’s, apart from the last two, which are WCAG’s. An audit judges your client’s site against them, and gives the Lighthouse score only for context. So a site that meets Google’s line but misses ours isn’t reported as failing. Each finding names the rule it breaks, and the report ends with every rule it cites and where the rule comes from.

Every page is built to WCAG 2.2 AA, the current W3C Recommendation. Automated tools find only part of what goes wrong: when the UK Government Digital Service built 142 barriers into a test page in 2018, the best tool caught 40% of them. So the automated test comes first, then an AI review in a real browser (below). This isn’t a formal conformance audit, and we don’t call a site conformant.

Search rules come from Google’s own documentation. Each page meant for search has its own title, description and canonical address, and structured data describes only what’s visible on the page. A business’s own site never marks up its own star rating, because Google recommends review markup only for sites that review other businesses.

Testimonials are only ever ones you supply in writing, because fake reviews break Australian Consumer Law.

How a build is checked

  • Speed test (Lighthouse) of every page against the first four rows of “Our builds”.
  • Accessibility test (axe) of every page, which includes contrast: zero violations, apart from any you agree to in writing.
  • Page tests for valid HTML, broken links, and, on every page meant for search, a title, description and canonical address that no other page shares.
  • AI review in a real browser, including keyboard use and tap targets against 44 px. Our process lists the rest.

The design

A client’s brand sets the look. These rules set how it works.

Body text is 18 px by default, in lines of about 45 to 75 characters, the range Robert Bringhurst gives in The Elements of Typographic Style. Every size comes from one scale and every gap from another, and a page uses at most two typefaces.

Our design checklist names the looks that fail on sight, because they mark a page as a template: a centred headline over three icon cards, a pill-shaped badge above the headline, gradient text, and rows of identical rounded cards with soft shadows.

Photos are your client’s own, of their own work. If they have none, the page is designed to work without them and the handover says so. Stock photos are used only when you ask for them, and each one’s licence is listed. There are no carousels, and movement is toned down for visitors whose device asks for less of it.

Your client’s brand sets the fonts and colours, and the accessibility rules still apply to them. If a supplied colour fails the contrast minimum, we tell you where before the build, along with the nearest colours that pass. The failing colour is used only if you confirm it in writing, and the test results then show the failure.

How a build is checked

  • Accessibility test (axe) of every page, including contrast.
  • Screenshot test (Playwright) of every page in three browser engines at four widths.
  • AI review of the screenshots against the design standard and the checklist’s list of templated looks.

The copy

Much of the copy standard comes from research on how people spot machine-written text.

Copy is in Australian English, plain and specific, and every number in it has a source. The standard starts from what people who use AI heavily notice when they pick out machine-written text: stock vocabulary, set sentence shapes and a safe, predictable line of thought. A test fails any page that uses the stock words and shapes on its list, or more than one em dash.

Instead of AI detectors, which studies have found neither accurate nor reliable, the AI review asks whether each sentence could sit unchanged on a competitor’s site.

How a build is checked

  • Spelling test (cspell) of every page against the Australian English dictionary.
  • Copy test of every page for the listed stock words and sentence shapes, and for more than one em dash.
  • AI review against the copy standard’s checklist, which includes the competitor test above and a source for every number.

Common advice we don’t repeat

“Titles should be 50 to 60 characters.”
Google sets no length limit. It shortens long titles to fit the screen, so the useful words go first.
“53% of mobile visitors leave after 3 seconds.”
The figure comes from a 2016 Google benchmark that published no method. It never appears in our reports.
“FAQ markup gets you rich results.”
Google stopped showing FAQ rich results on 7 May 2026, so we don’t sell the markup as a search feature.
“A better Lighthouse score ranks higher.”
The score summarises lab measurements. Google’s page experience signals come from real visitors, and Google says even good Core Web Vitals don’t guarantee top rankings.
“White space raises comprehension by 20%.”
We could find no study behind the figure, so it isn’t in our standard.

Sources

The ones behind this page, each read on 9 November 2026. The full standards cite more.

Start with one audit

The audit is the smallest job: one site, every finding with its evidence, and a fix list ranked by what matters most to the business, as a PDF under your agency’s name. It’s a fair way to judge the work before you send anything bigger. The sample audit shows what you’d get.

A$295. No GST is charged, because we’re not registered for it.

Start with an auditor write to hello@blindstamp.com.au