Getting started
How to read your SEO, AEO, and GEO scores
Three numbers, one verdict. What each score means, how they're calculated, and what counts as "good enough."
Last updated April 9, 2026
Every VistaCite audit returns three scores between 0 and 100: SEO, AEO, and GEO. They measure different things, and a site can be strong on one and weak on another. This doc explains how to read them.
The three pillars
Classic SEO is what Googlebot has always measured. Title tag length, meta description, H1, heading hierarchy, canonical tag, robots directives, Open Graph tags, image alt text, image dimensions, and internal linking. If you've run Ahrefs, Screaming Frog, or Semrush before, this pillar will feel familiar — the check list isn't radically different. What's different is we only audit a single URL at a time, so "site-wide" checks like broken-link crawls or robots.txt coverage against your whole domain aren't in scope.
AEO — Answer Engine Optimization is whether AI answer engines (ChatGPT, Perplexity, Google AI Overviews) can actually find and parse your page. This is where most sites are weak. The checks include: are AI crawlers like GPTBot, ClaudeBot, and PerplexityBot allowed in your robots.txt? Is there Organization and Article JSON-LD schema so engines know who you are and when the content was published? Are your H2 headings framed as questions (how answer engines index content)? Do you have direct-answer paragraphs immediately after each question heading? Is there a visible author byline and a "last updated" date? Does your title actually carry a claim or is it a generic topic?
GEO — Generative Engine Optimization is whether generative models will quote your page when they answer a question. This rubric comes mostly from Princeton's 2024 GEO paper and Ahrefs' 17M-citation study. The checks include: quotable sentence density (short, declarative, self-contained), statistic density (numeric claims per 1000 words), fact density, citation attribution phrases, authoritative outbound links to .gov / .edu / Wikipedia, direct quotations in blockquotes, structured lists and tables, paragraph chunkability for RAG retrieval, entity sameAs cross-links, and total content length.
How the scores are calculated
Each pillar has 10 checks. Each check returns pass / warn / fail. The pillar score is a weighted average where critical checks count more than minor ones. Perfect score (100) means every check passed. Zero means every check failed.
Scoring is deterministic. The same HTML always produces the same score. We do NOT use an LLM to grade pages — LLM grading is non-reproducible and hard to debug. Only the copy-paste fixes are LLM-generated, and those are advisory, not part of the score.
What counts as "good enough"
| Score | Interpretation |
|---|---|
| 90–100 | The page is well-structured for its pillar. Focus on off-page signals (backlinks, brand mentions) and content quality. |
| 70–89 | You're in decent shape but missing a few high-leverage fixes. The fix_it_plan.md file will show you exactly what. |
| 50–69 | Major gaps. The audit will flag multiple "This Month" improvements you should tackle before pushing hard on traffic. |
| Below 50 | Foundational issues. Critical checks are failing — these block AI engines from even understanding your page correctly. Start with the "This Week" tier. |
The average site we audit scores roughly: SEO 65, AEO 35, GEO 50. AEO is where most sites are worst — because it's the newest pillar, and because traditional SEO tools don't measure it.
Why the Crystal Ball is a separate signal
The Crystal Ball you see on the results page isn't part of the scores. It's a literal query: we ask an LLM what it knows about your domain, verbatim. If the answer is "I don't have reliable information about this domain," that's usually the diagnosis — no amount of on-page optimization will fix the fundamental problem that your brand isn't in the LLM's training data yet.
Fixing the Crystal Ball result is an off-page problem: Wikipedia presence, press mentions, podcast appearances, citations in published research. VistaCite can tell you the Crystal Ball is red, but we can't fix it for you.