The LaunchGraded Standard, in full.
9 dimensions. 51 checks. 100 points. Every weight, severity rule and verdict band is published here, so a score can always be traced back to the rules that produced it.
What is the LaunchGraded Score?
The LaunchGraded Score is a number from 0 to 100 that shows how ready a public website is to launch. It is the sum of 51 fixed checks grouped into 9 dimensions: messaging, conversion, trust, ux & mobile, seo, performance, accessibility, technical, discoverability. Every check has a published point value, and the values add up to 100. Most checks either earn their points or do not. A few scaled checks earn a share of their points from a Lighthouse category score and pass at 90 % or above. Every check, weight and severity rule is listed on this page, so any score can be traced back to the rules that produced it. The score describes the public page you submit, as a first-time visitor and Google's Lighthouse would see it. It does not predict traffic, rankings or revenue. The current rubric is LaunchGraded Standard 2026.2.
Does AI decide the score?
No. Every check is a fixed rule over facts a browser and Google's Lighthouse can observe on the page, such as whether a title exists, whether links resolve, whether text is legible on a phone and how fast the page paints. No language model judges your site, so nobody's opinion, including ours, moves the number. The same page always gets the same score, and a rescan after a fix changes the score only by what you fixed. AI appears in exactly two labelled places, both downstream of the score. One is the optional overview paragraph on the paid LaunchFix report. The other is the launch story written for a LaunchDay order. Both read the finished results and neither can change a point, a severity or a verdict. That is why the score can be repeated, compared across scans and trusted as a measurement rather than an impression.
What do the four verdicts mean?
The score maps to one of four verdicts: 90 or more is Launch Ready, 80 or more is Almost Ready, 60 or more is Needs Work, and below 60 is Not Ready. The verdict is the headline, because a single number hides what matters. Launch Ready means the site clears the LaunchGraded readiness bar. Almost Ready means good shape, with high-impact issues to fix before launch. Needs Work means several meaningful gaps are worth closing first, and Not Ready means critical issues should be fixed before you send real traffic. One rule can hold a verdict back: any failing launch-blocker check, such as a broken link gate, caps the verdict at Almost Ready whatever the score says. A 96 with one blocker reads Almost Ready with one blocker. The number stays an honest sum, and only the verdict waits for the fix. Certification needs 90 or more and zero blockers.
What does the LaunchGraded Score not measure?
The score does not measure copy quality, design taste, business viability, traffic, rankings or revenue. Judging whether a headline is persuasive would need a language model, and a model's opinion is not repeatable, so we only check that a headline exists, is visible and is short enough to parse, not whether it is good. Whether a design looks modern is subjective, while whether text is legible on a phone is not, so we check the second. We do not read your analytics or predict outcomes, and we never look behind a login: only the public page you submit is scanned, so onboarding, dashboards and checkout flows are out of scope. Those limits keep every check repeatable and keep the score honest about what it covers. A high score means your public launch experience is solid. It is not a verdict on product-market fit or on whether the business behind the site will succeed.
Will I get the same score if I scan again?
Yes, if the page has not changed. Scores come from fixed rules rather than a model, so scanning the same page twice gives the same answer, and when the number moves, the difference is real. The rubric is versioned. This page describes LaunchGraded Standard 2026.2, and a change to any weight or rule becomes a new version instead of a silent edit. Before and after comparisons are only drawn between scans on the same version, so an improvement in your score is never an artefact of us changing the rules. Scores from earlier versions stay valid and carry their own label. A scan stores what was observed on the page, while the titles, severity labels and fix advice you read are generated from that record, so wording can improve without changing anyone's score. Rescan as often as you like after a fix.
Principles
- Rule-based. Every check is a fixed rule over what a browser and Google's Lighthouse can observe on your public page. No language model decides anything in the score.
- Deterministic. Same page, same answer. Rescan without changing anything and you get the same score; change something and the difference is real.
- Versioned. This is LaunchGraded Standard 2026.2. A change to any weight or rule is a new version, and before/after comparisons are only ever drawn between scans on the same version.
- Facts and presentation are separate. A scan stores what was observed. The titles, severity labels and fix advice you read are generated from that record, so copy can improve without changing anyone's score.
9 dimensions, weighted by how often they sink a launch.
- 01 · 13 pts
Messaging
Can someone tell what you're selling in five seconds?
- 02 · 12 pts
Conversion
Is there one obvious next step, and a price behind it?
- 03 · 11 pts
Trust
Do visitors see proof, a human, and the legal basics?
- 04 · 7 pts
UX & Mobile
Does it work with a thumb on a phone?
- 05 · 14 pts
SEO
Can Google index it, and does the link look right when shared?
- 06 · 9 pts
Performance
Does the page load fast enough that people stay?
- 07 · 7 pts
Accessibility
Can people using assistive tech get through it?
- 08 · 14 pts
Technical
HTTPS, favicon, a real 404, nothing silently breaking.
- 09 · 13 pts
Discoverability
Can search engines and AI assistants tell what you are from structured data?
All 51 checks.
“Blocker” marks a check whose failure caps the verdict regardless of points. “Scaled” checks award a share of their points from the Lighthouse category score and pass at 90 % or above. Effort is our estimate for a typical fix. What each check looks for is described in plain words; the exact selectors and thresholds aren't published, so the score can't be gamed without actually fixing the thing. Each check below has its own link (right-click its id to copy) so a specific rule can be cited directly. Lighthouse-sourced checks follow Google's own Lighthouse audits; the performance checks follow Core Web Vitals thresholds; the accessibility checks are grounded in WCAG 2.1.
Messaging13 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Visitors can tell what you do from the first line#msg.headline The headline is the first — and sometimes only — thing a visitor reads. If it's missing, buried below the fold, or too long to parse, they leave before understanding what you do. | P0 | 4 | Page | 15 min |
Your headline is backed up by a one-liner#msg.subheadline A one-liner under the headline removes the ambiguity about who this is for and why it matters, right when attention is highest. | P2 | 3 | Page | 15 min |
Visitors can see the product (screenshot or demo)#msg.product_shown Visitors can't picture what they're getting from text alone. Without a screenshot or demo they're asked to trust a description of something they've never seen. | P2 | 3 | Page | 2 h |
Your browser tab and search title say what you are#msg.title_tag The <title> is what shows in the browser tab, in search results and in most link previews. A missing or generic one ("Home") wastes the most-seen line you have. | P3 | 1 | Lighthouse + Page | 5 min |
The page has enough copy to explain what you do#content.word_count A page with almost no text gives visitors and search engines nothing to work with beyond a headline — there's no room to cover objections, use cases or how it works. | P3 | 1 | Page | 1 h |
Your copy is easy to read#content.readability Copy that's either too dense (long sentences, jargon) or oddly simplistic reads as unpolished. A quick Flesch Reading Ease check over your body text flags the extremes — it's a formula, not a style opinion. | P3 | 1 | Page | 1 h |
Conversion12 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
There's one obvious next step#conv.primary_cta A missing or vague call to action leaves even a convinced visitor unsure what to do. Undecided visitors leave rather than hunt for the next step. | P0 | 8 | Page | 30 min |
Visitors can find out what it costs (or how to talk to you)#conv.pricing_path Hidden pricing with no path to a conversation makes visitors do extra work to find out whether they can afford or qualify for this. Many won't bother. | P2 | 3 | Page | 30 min |
Not-ready visitors have a low-commitment way to stay in touch#conv.secondary_optin Most first visitors aren't ready to sign up or buy. Without a lighter option — updates, a waitlist, a newsletter — you lose everyone who was close but not convinced. | P3 | 1 | Page | 30 min |
Trust11 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
You show evidence other people use this#trust.social_proof A first-time visitor has no reason to believe you yet. Without logos, testimonials, review counts or usage numbers, they have only your word that this works. | P1 | 4 | Page | 2 h |
Privacy policy and terms are linked#trust.legal_links A site that asks for an email but links no privacy policy reads as unfinished, and some payment and ad platforms require both pages before they'll approve you. | P2 | 2 | Page | 30 min |
Visitors can reach a human#trust.contact_path A visible way to reach you is the cheapest trust signal there is. No contact link, email or social handle makes the product feel unattended. | P2 | 2 | Page | 15 min |
Your content shows a recent date#content.freshness A visible published or updated date lets readers judge whether content-style pages still reflect current information — undated or stale content reads as abandoned. | P3 | 1 | Page | 15 min |
Your content shows who wrote it#trust.author_byline An unattributed article reads as anonymous, generic content. A named author with a bio gives readers someone to trust the claims came from. | P2 | 2 | Page | 15 min |
UX & Mobile7 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Your page adapts to phones#ux.viewport Most first visits from a shared link are on a phone. A page with no responsive viewport renders as a shrunken desktop site and loses those visitors before they read a word. | P0 | 3 | Page + Lighthouse | 5 min |
Buttons and links are tappable on a phone#ux.tap_targets Links and buttons that are too small or too close together cause mis-taps on a phone, and a mis-tapped call to action is a lost conversion. | P2 | 2 | Lighthouse | 1 h |
Text is legible on a phone#ux.font_size Text under about 12 px forces phone visitors to pinch and zoom, and most simply leave instead. | P2 | 2 | Page | 30 min |
SEO14 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Google is allowed to index your homepage#seo.indexable Your page tells search engines to stay away. Every link, mention and launch post will point at a page Google won't list. | P0 | 3 | Lighthouse + Page | 5 min |
Search results show a description you wrote#seo.meta_description Without a meta description, search engines and link previews pick a random sentence from the page — usually not the one you'd choose. | P2 | 2 | Page + Lighthouse | 5 min |
Your link looks right when shared#seo.social_preview A launch spreads by links. Without Open Graph title, description and image, your link shows up on social and in chat as a bare URL or a broken card. | P2 | 2 | Page | 30 min |
Search engines know the one true URL#seo.canonical Without a canonical tag, www/non-www, trailing slashes and tracking parameters can split your page into several URLs that compete with each other in search. | P2 | 2 | Page + Lighthouse | 15 min |
Links say where they go#seo.link_text "Click here" and "Learn more" links tell neither visitors nor search engines what's behind them, and JavaScript-only anchors can't be crawled at all. | P3 | 1 | Lighthouse | 30 min |
You have a sitemap search engines can find#seo.sitemap_present A sitemap tells search engines every URL you want indexed instead of relying on them to discover pages by crawling links, which is slower and can miss pages with no internal links pointing at them. | P2 | 2 | Extra request | 15 min |
The URLs in your sitemap actually resolve#seo.sitemap_reachable_urls A sitemap listing dead or redirected URLs wastes the crawl budget search engines give your site and signals a stale or auto-generated file nobody checks. | P3 | 1 | Extra request | 15 min |
robots.txt points crawlers at your sitemap#seo.robots_sitemap_link robots.txt is the first file most crawlers check. A Sitemap: line there gets your sitemap discovered immediately instead of waiting on a manual submission or a lucky crawl. | P3 | 1 | Extra request | 5 min |
Performance9 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Slow pages lose visitors before the content renders, and the drop-off compounds with every extra second on a phone connection. | P1 | 6 | Lighthouse | 1 day |
The main content appears within 2.5 seconds#perf.lcp Largest Contentful Paint is when the visitor sees your hero. Past 2.5 seconds on mobile, a noticeable share of them assume the page is broken. This is a live lab measurement, so it can read a few hundred ms differently between two scans of an unchanged page. | P2 | 2 | Lighthouse | 2 h |
The page doesn't jump around while loading#perf.cls Layout shift is when a button moves as the visitor reaches for it. It reads as jank and causes mis-taps on the very elements you want tapped. | P3 | 1 | Lighthouse | 1 h |
Accessibility7 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Accessibility failures shut out visitors using screen readers or keyboards, and the same issues — unlabelled controls, poor contrast — trip up everyone on a phone in sunlight. | P2 | 3 | Lighthouse | 1 day |
Images have text alternatives#a11y.image_alt Without alt text, screen-reader users hear "image" and search engines see nothing — your product screenshot becomes invisible to both. | P3 | 1 | Lighthouse | 30 min |
Form fields have labels#a11y.form_labels An input with only a placeholder has no name once someone starts typing, and no name at all for a screen reader — the email capture you rely on is the field most often affected. | P3 | 1 | Lighthouse | 15 min |
The page declares its language#a11y.lang Without a lang attribute, screen readers guess the pronunciation rules and browsers may offer to translate a page that's already in the visitor's language. | P3 | 1 | Lighthouse | 5 min |
Text is readable against its background#a11y.contrast Low-contrast text — light grey on white is the classic — is hard to read for everyone and unreadable for many, especially on a phone outdoors. | P3 | 1 | Lighthouse | 1 h |
Technical14 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Your site is served securely#tech.https Browsers show a "Not secure" warning when a page is served over HTTP or mixes HTTPS and HTTP content. For a first-time visitor that reads as "don't type your email here". | P0 | 2 | Page + Lighthouse | 30 min |
Your site sends baseline security headers#tech.security_headers Strict-Transport-Security, X-Content-Type-Options and a framing defence (Content-Security-Policy or X-Frame-Options) are cheap, standard headers that close off common browser-level attacks. Their absence is one of the first things a technical visitor checks. | P2 | 2 | Extra request | 1 h |
Every resource loads over HTTPS#tech.mixed_content An HTTPS page that loads even one image, script or stylesheet over http:// gets flagged as mixed content — browsers may block the resource outright or strip the page's "secure" padlock. | P3 | 1 | Page | 30 min |
Nothing is silently breaking in the browser#tech.console_errors JavaScript errors on load usually mean something a visitor expects — a form, a menu, an embed — isn't working, and you won't hear about it from them. | P2 | 2 | Lighthouse | 1 h |
You have a favicon#tech.favicon A blank tab icon is a small thing that signals "unfinished" in every tab, bookmark and share card. | P3 | 1 | Page | 15 min |
You have a recognizable brand mark#tech.brand_mark A site with no logo anywhere — not in the header, not in your Organization schema, not even a proper home-screen icon — reads as unfinished, and gives visitors nothing to recognize you by on a second visit or a share. | P3 | 1 | Page | 30 min |
The plain http:// address redirects to https://#tech.http_redirects_https Most hosts redirect http:// to https:// automatically, but a server that answers on plain HTTP without redirecting serves an insecure, unencrypted copy of the page to anyone who types the bare domain or an old http:// link. | P3 | 1 | Extra request | 15 min |
www and the bare domain don't both serve live content#tech.host_canonicalization If both www.example.com and example.com load the same page with no redirect between them, search engines see two copies of every page, and links pointing at each host build up authority separately instead of together. | P3 | 1 | Extra request | 15 min |
A wrong URL gets a proper 404#tech.not_found If a wrong URL returns 200 with a blank or homepage-lookalike page, search engines index junk URLs and visitors who mistype get no signal to correct it. | P3 | 1 | Extra request | 30 min |
Your headings follow a clean hierarchy#content.heading_hierarchy Exactly one <h1> and no skipped heading levels is how screen readers and search engines build an outline of the page — jumping from an <h1> straight to an <h3> (or having none, or several) breaks that outline. | P2 | 2 | Page | 30 min |
No broken links on the page#gate.broken_links A dead link on launch day is the fastest way to look unfinished. If it's in the nav or footer, it's on every page. | P0 | gate | Page + Extra request | 30 min |
Discoverability13 pts
| Check | If it fails | Points | Source | Effort |
|---|---|---|---|---|
Your page has structured data search engines can parse#struct.schema_present Structured data (JSON-LD) is how search engines and AI assistants understand what your page is about beyond the visible text — without it you're relying entirely on crawled prose. | P2 | 2 | Page | 1 h |
Your structured data declares what you are#struct.organization_or_product Organization, Product, SoftwareApplication and WebSite are the schema types search engines and AI assistants rely on most to build knowledge panels and answer "what is this" queries — without one of them, structured data on the page has nothing to attach to. | P2 | 3 | Page | 1 h |
Your structured data has its required fields filled in#struct.required_fields A schema block missing its commonly-required fields — a Product with no offers, an Organization with no name — is often ignored by search engines even though it technically validates as JSON-LD. | P2 | 2 | Page | 30 min |
Your structured data is clean, not malformed or conflicting#struct.no_duplicate_schema Malformed JSON silently drops your structured data with no visible error, and two blocks of the same type that disagree (different names for what should be the same Organization) leave search engines unsure which one to trust. | P2 | 2 | Page | 30 min |
You use breadcrumb or Q&A markup where it applies#struct.breadcrumb_or_faq BreadcrumbList markup can earn a breadcrumb trail in search results, and QAPage markup makes a genuine Q&A page eligible for a rich result — free visual real estate in the SERP. (Legacy FAQPage markup lost this eligibility in 2026.) | P3 | 1 | Page | 1 h |
AI crawlers can access your site#geo.ai_crawlers_allowed AI assistants (ChatGPT, Perplexity, Claude) can only cite or summarize pages their crawlers can reach. Blocking GPTBot, ClaudeBot and similar crawlers in robots.txt keeps your site out of those answers — sometimes a deliberate choice, but worth confirming it's intentional. | P2 | 2 | Extra request | 15 min |
You publish an llms.txt#geo.llms_txt_present An llms.txt gives AI assistants a curated, plain-text summary of your site instead of making them parse rendered HTML. It's informational only — Google's own AI-optimization guidance says it doesn't affect Search visibility — but it's a cheap, emerging convention for the AI-crawler audience specifically. | P3 | 1 | Extra request | 1 h |
How findings are ranked.
Severity
| Severity | Rule | On the report |
|---|---|---|
| P0 · Launch blocker | Fails a check marked as a launch blocker, or the broken-links gate | Must fix before launch |
| P1 · High impact | Fails a check worth 4 points or more | Strongly recommended before launch |
| P2 · Improvement | Fails a check worth 2–3 points | Worth fixing soon |
| P3 · Polish | Fails a 1-point check, or a scaled check between 70 % and 90 % | Nice-to-have |
Impact and effort
Impact is the points at stake: high at 6 or more, medium at 3–5, low at 2 or fewer. A blocker is always high impact.
Effort is a fixed estimate per check — 5 min · 15 min · 30 min · 1 h · 2 h · 1 day, for a founder or developer who knows the site.
Fix first is the top five failing checks ordered by severity, then impact, then least effort. Quick wins are failing checks that take an hour or less, ordered by impact.
Severity labels: P0 Launch blocker · P1 High impact · P2 Improvement · P3 Polish.
Four verdicts. One capping rule.
| Score | Verdict | What it means |
|---|---|---|
| 90–100 | Launch Ready | Your site clears the LaunchGraded readiness bar. |
| 80–89 | Almost Ready | Good shape — fix the remaining high-impact issues before launch. |
| 60–79 | Needs Work | Several meaningful gaps are worth closing first. |
| 0–59 | Not Ready | Fix the critical issues before sending real traffic. |
What's revealed for free, exactly.
The free scan always shows the full, fair score, the verdict and all 9 dimension scores. Individual findings are revealed by a fixed rule, not a sales decision — the same rule runs on every scan, so what you see for free never depends on how bad your score is.
- If something's failing, the worst one is explained in full: why it matters, how to fix it, the technical detail.
- Every other check appears as a locked row with its dimension, pass/fail and severity visible, so counts like “8 issues worth fixing” are verifiable.
The rule is decided once per scan and fixed for this version of the Standard. What the LaunchFix adds.
What we deliberately don't score.
- Copy quality and tone
- Judging whether your headline is persuasive would need a language model, and a model's opinion isn't repeatable. We check that a headline exists, is visible and is short enough to parse — not whether it's good.
- Business viability
- A high score means your public launch experience is solid — it isn't a verdict on product-market fit or whether the business behind it will succeed.
- Traffic, rankings, revenue
- We don't read your analytics and we don't predict outcomes.
- Anything behind a login
- Only the public page you submit. Onboarding, dashboards and checkout flows are out of scope.
- Design taste
- Whether it looks modern is subjective. Whether text is legible on a phone is not — that one we check.
How the Standard is tested, and how it changes.
Calibration
Before a version is locked we run it over three groups of real sites: launch-ready indie sites, clearly unfinished ones, and live sites with real traction. The version passes when the good group averages 80 or more, the weak group 55 or less, and no live, performing site lands in Not Ready. Any surprising result has to be explainable check by check — a detection miss means we fix the check, not the site.
LaunchGraded Standard 2026.2 is being calibrated now; the aggregate results are published here once it is locked.
Version history
- LaunchGraded Standard 2026.2, 9 dimensions, 51 checks, verdicts, severity and the blocker cap. Replaces letter grades. Adds SEO, accessibility and technical dimensions from Lighthouse.
- v1 (2026), 10 checks, letter grade, 70-point cap on broken links. Scores from v1 stay valid and are shown with their own label; they are not compared against v2 results.