HTML Page Auditor (Lighthouse-style)

The sample page. The box opens with a 16-line page that has a title, a viewport tag and one <h1>, but no meta description, an image with no alt and a search box with no label. 85 of the 94 checks that run apply to it. Those three gaps are the 3 failures, each worth 3 points, alongside 25 warnings. SEO earns 22 of 28 points, so 22 ÷ 28 × 100 = 78.6, shown as 79. Accessibility earns 20.5 of 29 (71), and the overall score is the average of the seven rings, (79 + 71 + 85 + 84 + 90 + 94 + 85) ÷ 7 = 84.

Fixing the three failures. Add <meta name="description" content="Acme widgets are built to last.">, give the image alt="The Acme widget" and add <label for="q">Search</label>. SEO becomes 26 ÷ 29 × 100 = 90: the description adds a 1-point width check, which it passes. Accessibility becomes 27.5 ÷ 30 × 100 = 92, and the overall score rises from 84 to 89 with 0 failures.

The common mistake: counting title characters instead of pixels. Google cuts titles by width, not length, so the auditor estimates the width at 20 px and passes it at 600 px or less. WIDGET WAREHOUSE MEGA SALE: NEW MODELS IN STOCK NOW is only 51 characters but about 639 px, a warning, while the 55-character little lists of lilies: illinois field notes, all filed is about 440 px and passes. The sample’s 30-character title is about 315 px.

This tool runs the kind of checks Google’s Lighthouse runs, but on the parts of a page that live in the markup itself. You give it the HTML; it parses that HTML into an inert document with the browser’s built-in DOMParser — no scripts run, no images or stylesheets are fetched, nothing is rendered — and then it inspects the structure. From that it scores SEO (29 checks: a title, a meta description, a viewport tag, a language, exactly one h1, a canonical link, and is it actually indexable rather than noindex?), Accessibility (31: do images have alt text, do form fields have labels, do buttons and links have an accessible name, are heading levels sane, is zoom left enabled?), Best Practices (19: a doctype and a charset, deprecated tags, insecure http:// resources, inline event handlers), a structural Performance pass (20), Security (19: Subresource Integrity, iframe sandboxing, insecure forms, CSP, risky JavaScript sinks), Privacy (11: third-party origins, known trackers, third-party fonts and embeds) and Mobile (7). That is 136 checks; each page runs only those that can apply to it.

It is honest about its limits: it cannot produce a timed performance score from pasted text, and it cannot reach out and audit an arbitrary live URL. Measuring Largest Contentful Paint, Cumulative Layout Shift, and Total Blocking Time requires loading the page in a real browser. And a page running in your browser is forbidden by the same-origin policy from reading another website’s HTML, so “type any URL” isn’t possible without a server proxy — which would break the promise that nothing you paste ever leaves your device. So you paste the source (or fetch a page on this same domain), and the Performance ring scores the structural signals that genuinely live in the markup: render-blocking scripts in the head, images missing width and height (a classic layout-shift cause), the number of render-blocking stylesheets, and overall document weight. For pages on this site, Load & measure shift then loads the page in a sandboxed frame and captures the real CLS, LCP, page weight, request count, blocking time, and the exact elements that shifted.

Google released Lighthouse as an open-source project in 2016, first to audit progressive web apps, and built it into the Chrome DevTools Audits panel in 2017. Its 0–100 scores and the green, orange and red bands used here (90 and above, 50 to 89, below 50) made page quality something a team could track like a test result.

In May 2020 Google named three Core Web Vitals — Largest Contentful Paint, First Input Delay and Cumulative Layout Shift — with “good” thresholds of 2.5 seconds, 100 ms and 0.1, and made them part of search ranking in the 2021 page experience update. Interaction to Next Paint replaced First Input Delay in March 2024. The 4.5:1 contrast rule comes from WCAG 2.0, published by the W3C in December 2008.

Every category is a weighted bundle of individual audits. A pass earns the audit’s full weight, a warning earns half, a failure earns nothing, and an audit that doesn’t apply — the image-alt check on a page with no images, say — is dropped from the maths entirely so it can’t unfairly drag a score down. The audit list sorts failures and warnings to the top, each with what’s wrong and a fix you can copy, so the report becomes a punch-list you can work straight down. Fix an item, edit the HTML, and the scores update. Show Work lists every check with the value it measured, so you can see exactly why a ring is not green.

The HTML you paste is parsed in your browser and never uploaded. For the background on what each check means, read How Web Page Auditing Works.

Related tools: Meta Tag & Open Graph Generator, Schema / JSON-LD Generator, and robots.txt Generator.