DraftCamp Page Audit

Audit any page for SEO and GEO/AEO — including what AI crawlers can actually see. Runs entirely in your browser.

Click the icon. Get a complete audit of the page you're on — for search engines and for AI answer engines — in about two seconds. No account. No API key. No sign-up. Nothing is sent to any server. WHY THIS EXISTS Most SEO extensions were built for a web where Google was the only reader that mattered. That stopped being true. People now ask AI assistants such as ChatGPT or Claude what they used to type into a search box. Those systems answer by fetching pages, extracting passages, and citing sources. If your page can't be fetched, parsed, or cleanly quoted, it doesn't appear — no matter how well it ranks. That's a different problem from ranking, and it needs different checks. This extension does both, side by side, without blending them into one misleading number. ROBOTS.TXT, READ PROPERLY Blocking an AI crawler isn't one decision. It's three, and the right advice is different for each. Most tools flag all of them identically and tell you to fix something that isn't broken. RETRIEVAL CRAWLERS — these produce citations These fetch a page to answer a question someone is asking right now, and they are the ones that generate citations. Blocking them removes you from AI answers, so the report treats it as an error. TRAINING CRAWLERS — a legitimate choice These collect text for future model training. Plenty of publishers block them deliberately, and doing so has no effect on whether you are cited today. Reported for awareness, never as a defect. USER-TRIGGERED FETCHERS — narrower cost These act only when a person pastes your link into a chat. Blocking them breaks that specific flow, so the report raises a warning rather than an error. The extension recognises the current user-agent tokens for all three groups across the major providers, and keeps opt-out tokens separate from genuine fetchers, because they are not the same thing. The matcher follows RFC 9309 properly: the most-specific user-agent group wins, the longest path rule wins, and Allow beats Disallow on an equal-length match. A file that blocks everything with a wildcard and then allows one named crawler is read correctly. WHAT IT CHECKS CRAWLABILITY AND INDEXABILITY • HTTP status and the full redirect chain • HTTPS • Robots meta directives — noindex, nofollow, noarchive, nosnippet, max-snippet • X-Robots-Tag response header, which is invisible in the DOM and easy to ship by accident • Canonical: present, absolute, self-referencing, or pointing elsewhere • hreflang set validity — malformed codes, missing self-reference, missing x-default • robots.txt reachability and whether this specific path is crawlable • sitemap.xml existence, and whether this URL is actually listed in it • lang attribute and viewport • rel=prev/next pagination ON-PAGE SEO • Title length measured in pixels, not characters. Google truncates on width, so "Illinois" and "Wowwww" are the same character count and nowhere near the same width. • Meta description length, also in pixels, against the ~920px actually rendered in results • Exactly one H1, and a heading outline with no skipped levels or empty headings • Word count against content-type targets, with the page type inferred from title and heading shape • Image alt coverage — distinguishing a missing alt attribute from a deliberately empty one, since an empty alt is valid markup marking an image as decorative, not a defect • Generic alt text that names the format ("chart") rather than the finding • Images missing width and height, the main driver of layout shift • A lazy-loaded first image, which is a common own-goal on LCP • Link profile: internal, external, nofollow, sponsored, ugc • Broken links — actually fetched, not guessed • Empty anchors, generic anchor text, and javascript: hrefs that no crawler can follow • Open Graph tags, with the og:image downloaded and measured against 1200×630 • Twitter card tags, favicon, text-to-HTML ratio, DOM weight • URL quality: underscores, uppercase, depth, length, dates in the path, and tracking parameters without a canonical STRUCTURED DATA Every JSON-LD, microdata and RDFa block is parsed and validated per type. • JSON-LD syntax errors — a malformed block is discarded silently by every consumer, and nothing warns you • An article block is checked for its headline, image, published and modified dates, and an author with both a name and a URL • Question-and-answer markup is checked for a valid question and answer structure • Step-by-step markup is checked for the presence of steps • Product markup is checked for offers, price, currency, availability and ratings • Publisher markup is checked for name, URL, logo and linked entity profiles • Breadcrumb markup is checked for contiguous positions starting at 1 • Content mismatches — schema headline disagreeing with the visible H1, schema dates with no visible counterpart, and FAQ markup describing answers that aren't on the page (a documented guideline violation that risks a manual action) GEO / AEO EXTRACTABILITY Ten dimensions, each scored 0 to 5: 1. Answer-First Structure — does the intro, and each section, lead with the answer or warm up to it? Openings are checked for length, for hedging, for preamble like "In this article", and for whether they actually engage their own heading. 2. Section Depth — how many H2 sections land in the 120-180 word band? Sections in that range are cited substantially more often, because they hold one complete idea: long enough to stand alone when lifted out, short enough that the answer isn't diluted. 3. Statistics and Data — how often does a sentence carry a real figure, and how often is that figure attributed? 4. Authoritative Quotations — direct quotes with a named source. This is the highest-impact single signal in the rubric. An unattributed pull-quote is decoration and scores accordingly. 5. Sources Cited — outbound links to genuinely authoritative domains, weighed against how many claims the page makes. 6. FAQ Presence — question-form headings answered immediately in 40-80 words, which is the shape an answer engine can lift whole. 7. Heading Hierarchy — the nesting that tells a model which answer belongs to which question. 8. Internal Linking Depth — five or more internal links with descriptive anchor text. 9. Image Alt Coverage — descriptive alt text, not format labels. 10. Freshness Signals — visible dates, schema dates, and current-year references that are actually accurate. CORE WEB VITALS • Largest Contentful Paint, Cumulative Layout Shift and Time To First Byte, read from the live page • INP is deliberately not shown. It measures real user interactions, and there are none to read on a page nobody has interacted with yet. A number there would be invented. MEASURED VS ESTIMATED Every GEO dimension is labelled with how much you should trust it. MEASURED (7 of 10) — the measurement is the definition. Counting the words between two H2 headings is section depth. There is no judgement involved and nothing a language model would add. ESTIMATED (3 of 10) — Statistics and Data, Authoritative Quotations, and Sources Cited. The extension can count quotation marks, attribution patterns and authoritative domains. It cannot judge whether a source genuinely carries weight. So it says so, on the row, every time. That distinction is the honest part. A tool that presents a guess with the same confidence as a measurement is telling you something it doesn't know. TWO SCORES, NEVER ONE The report shows an SEO score and a GEO score side by side, and never averages them. A page can be technically flawless for search and still invisible to every answer engine. It can also be beautifully structured for extraction while being blocked in robots.txt. One combined number hides exactly the situation you most need to see. Every finding carries three things: what was measured, why it matters, and how to fix it. The "why" is what makes the report useful — someone who understands the mechanism can judge whether a fix is worth it on their page, rather than working a checklist on faith. WHAT IT DELIBERATELY DOESN'T DO No keyword-density check. Keyword-density enforcement is a documented failure mode that produces worse writing. It isn't here and won't be. No penalty for length. A page is never marked down for exceeding a word target. Doing so punishes thorough comparison pieces and multi-vendor roundups for being complete. No impersonation. The extension never disguises itself as a crawler, and setting such a user-agent would require a permission it doesn't request. The report says "HTML as served, JavaScript disabled" — which is the condition that actually matters, since no major AI crawler runs JavaScript. No exhaustive link crawl. Link checking samples up to 24 links, four at a time, article links first — and the report tells you it was a sample. Checking 400 links would take minutes and would make your browser look like a scraper to the server on the other end. PRIVACY Everything runs locally in your browser. • No account, no login, no API key • No analytics, no telemetry, no error reporting, no tracking of any kind • Nothing is sent to DraftCamp or to any third party • No remotely hosted code While an audit runs, the extension makes requests only to the origin of the page you're auditing: robots.txt, llms.txt, sitemap.xml, a no-JavaScript copy of the page, plus the outbound links and Open Graph image it verifies. Those requests are made without cookies, so they carry none of your session data. That's not only a privacy choice: a crawler has no session either, so it is necessary for accuracy. Reports are cached in session storage so reopening the popup is instant. Session storage lives in memory only, never touches disk, and clears when you close your browser. The extension reads a page only when you open the popup. It does nothing on pages you haven't asked it to audit. WHO IT'S FOR • Technical SEOs auditing pages one at a time and needing an answer immediately • Content teams checking a draft is extractable before it ships • Developers who want to know what their framework actually serves a crawler • Anyone whose traffic has shifted toward AI answers and who needs to see why • Agencies doing fast client-site triage without opening a full crawler It works on any site — yours, a client's, or a competitor's. BROWSER SUPPORT Built on Manifest V3, so it runs in any current Chromium-based browser. ABOUT DRAFTCAMP DraftCamp helps content teams write and maintain articles that earn search traffic and AI citations. This extension audits one page at a time and is completely free, with no account required. The DraftCamp app runs the same class of analysis across every page on a site, tracks scores over time, and connects to Google Search Console — that's the paid product, and this extension works fully without it. Questions or bug reports: support@draftcamp.ai