—
—
—
—
Open Graph preview · image measured, not assumed
See your link before you post it
Renders your card for LinkedIn, X, Slack and WhatsApp — then downloads the image and decodes its true dimensions, because declaring 1200×630 and serving 400×400 is the most common way a preview breaks.
Try
Fetching page…
0/100
How your link looks when shared
—
Want to know when this breaks?
These checks pass today. They break quietly — a deploy drops your og:image,
a migration adds noindex, a rewrite makes your content invisible
to AI crawlers. We'd re-scan weekly and email you only when something regresses.
This doesn't exist yet. We're checking whether it's worth building. Leave your email and we'll tell you if it ships — no other mail, ever.
What gets checked
81 checks across six categories. Most of them require fetching something, decoding something, or evaluating a rule — the ones you could do by reading your own source are the least interesting part.
Social & Open Graph
We fetch your og:image and decode its real dimensions from the file header, then tell you which platforms will crop it. A page can declare 1200×630 and serve 400×400 — every tag-reading tool calls that a pass.
SEO basics
Titles, headings, canonical targets, sitemap contents. We also fetch a sample of your own internal links to find the dead ones, and probe the www/non-www variant to catch two indexable copies of one site.
Launch readiness
Certificate expiry and chain trust read straight off the TLS handshake. Plus redirect chains, real 404 status codes, cookies set before consent, and placeholder copy still sitting in the live page.
AI visibility
Whether your content survives without JavaScript, and a full robots.txt rule engine that answers can GPTBot fetch this exact URL — group specificity, longest-match precedence and wildcards included.
Performance & delivery
Measured from the actual response: whether compression is on, whether anything is cacheable, render-blocking scripts, DOM size, and images with no dimensions that will shift your layout as they load.
Accessibility & trust
Unlabelled form fields, links with no accessible name, mixed content, forms posting over plain HTTP, and whether a stranger can work out who is behind the site at all.
Questions
Why does AI visibility matter?
Most AI crawlers don't execute JavaScript. If your page renders entirely client-side, assistants see a blank document — so they can't summarise or cite you, no matter how good your content is. Traditional SEO tools don't test for this.
Do you store the sites I scan?
No. Results are held in memory for 60 seconds so a repeat scan of the same URL is instant, then discarded. Nothing is written to disk and there's no database.
Why does my score differ from other tools?
A passed check is 1, a warning 0.5, a failure 0. Categories are then averaged with weights rather than equally: AI visibility counts 1.5, SEO and launch readiness 1.25, social 1, and performance and accessibility 0.75 each. A perfect social score can't hide a broken AI visibility score.
We also measure some things other tools infer. If your og:image tag says 1200×630 but the file is 400×400, we download it and score the file, not the tag.
What does the projected score mean?
It's the real score recomputed with the listed fixes applied — the same arithmetic, with those checks flipped to passing. It isn't an estimate or a growth-hack number, and it deliberately only counts the items shown, not every warning in the report.
Does scanning affect my site?
A scan makes roughly a dozen requests: the page, robots.txt, sitemap.xml, llms.txt, your og:image, your favicon, your manifest, a deliberate 404 probe, and the www/non-www variant. It then fetches up to eight of your own internal links to find broken ones.
Those link requests are sequential, never parallel — a burst of simultaneous requests to someone else's server is indistinguishable from a small attack. We identify ourselves as LaunchCheckBot, follow at most five redirects, cap each response, and time out after 12 seconds.
Do you obey my robots.txt?
No, and we'd rather say so plainly. A scan is a single fetch a person explicitly asked for, not a crawl — and honouring Disallow would mean refusing to audit the staging sites and locked-down paths that people most want checked. We read your robots.txt to report on it, not to take instruction from it.
We send a stable user agent, LaunchCheckBot, with a contact URL. If you want us gone, block that user agent at your server or WAF and every request will fail closed.