GEO Snapshot by sizrok

See what AI search engines see when they look at a web page. Click the icon on any site and GEO Snapshot runs a full audit over the live page. The panel sits on top of the site, so one screenshot captures both the page and what is wrong with it. Twenty eight checks, grouped into tabs, free, with no account and no limits. WHAT IT CHECKS AI access — whether the page is readable without JavaScript, whether robots.txt turns AI crawlers away, whether llms.txt and the XML sitemap are there and real. Basics — indexability, title, meta description, H1, canonical, language and hreflang, and whether the page works on a phone. Content — whether an assistant could lift an answer from the page, the heading outline, word count, reading ease, how fresh the page is, and how much of the markup is actually text. Schema — the structured data on the page validated against the required and recommended properties for each type, with the rich results it makes the page eligible for listed by name. Links and images — internal against external links, empty and generic anchors, broken links, alt text coverage, oversized files, and images with no dimensions set. Speed — real Core Web Vitals measured on this load, on your connection, with the largest element highlightable on the page. THE CHECK MOST SITES FAIL WITHOUT KNOWING Pages that build their content in the browser look perfect to you and are blank to the crawlers that feed AI answers, because those crawlers do not run JavaScript. GEO Snapshot fetches the page's own served HTML and compares it against what the browser ended up with, then tells you how many words and headings were missing from the version a crawler receives. It also parses robots.txt the way a crawler does, rather than searching it for a string: user agent groups, comments, Allow overrides, and the rule that a named group beats the wildcard. Results are grouped by product rather than by crawler, because most AI products run more than one and they do different jobs. One gathers pages in bulk; another fetches a single page at the moment it is about to be cited in an answer. A site that allows the first and blocks the second is invisible in answers while its owner believes it opted in, and that is the most common way this goes wrong. SCAN THE WHOLE SITE, NOT ONE PAGE Most of the checks work on a fetched page, so the extension can read your sitemap and audit your own pages in the browser. Duplicate titles and descriptions, pages with no description, thin pages, accidental noindex, and urls in the sitemap that no longer load. Same origin only, so it reads the site you are standing on and nothing else. No crawler, no backend, no account. Export the results as CSV. SEE THE PAGE THE WAY OTHER PEOPLE MEET IT The search preview draws your result the way a search engine renders it, with truncation measured in pixels rather than characters, because that is how the cut is actually decided. The share preview draws the card at the ratio every messaging platform crops to. Click any finding and the exact elements are outlined on the live page, so you can see the thing rather than read about it. THERE IS NO PRIZE FOR 100 The score is a checklist of things that are present or not. No search engine ever sees it. A page in the eighties that answers the question beats a page at 100 with nothing to say, so the panel says so, next to the number. If a check could not run, it says so, and it is left out of the score rather than counted against the site. Nothing is ever reported as a failure unless it was actually observed. WHAT YOU CAN TAKE AWAY A plain text summary, a Markdown table, the structured data the page should carry, an llms.txt draft built from your own sitemap, CSV of every link and every image, and a ready to run prompt for an AI assistant. The prompt carries the measured values, lists only the checks that failed, and closes with rules that forbid inventing services, locations, prices or claims the audit did not evidence. You get finished replacement text back, not a lecture about why SEO matters. PRIVACY No backend, no account, no API key, no analytics, no telemetry. Nothing is collected and nothing is sent anywhere. It asks for no website access when you install it. It runs on the tab you clicked on. Audit history is stored on your own computer and there is one button that deletes it. Anything beyond the page you are on is off until you turn it on. The settings page lists every extra surface with the exact access it needs, one switch each, and switching it off hands the permission back to Chrome. Free, and free of the usual reasons a free tool exists. Built by sizrok, an AI automation studio in Greater Manchester.