Every score a Juma audit produces comes from the rubric on this page — six weighted dimensions, every sub-check documented, every weight versioned. Other GEO tools hide their methodology. We publish ours.
Each of the six dimensions produces a 0–100 sub-score. The overall score is a weighted average, rounded to the nearest integer.
Dimension
Weight
AI Crawler Access
20%
Citability
25%
Schema Markup
15%
Technical SEO
15%
Content Authority
15%
Brand Presence
10%
Overall score
100%
Weights reflect what most directly predicts AI citation today. Crawler Access is binary-ish — a block there zeroes out everything else — but Citability carries the highest weight because it's the most actionable lever for content teams.
Score bands
Excellent
80–100
Good
60–79
Needs Work
40–59
Critical
0–39
Dimensions, in detail
AI Crawler Access
20% weight
Whether AI search engines are actually allowed to read your site.
If GPTBot, PerplexityBot, ClaudeBot, or Google-Extended are disallowed in robots.txt, the other five dimensions don't matter — your content will never reach the index that answers user prompts.
Data source — Fetches /robots.txt directly; parses user-agent blocks and Disallow/Allow directives with wildcard fallback.
Sub-check
Points
Criteria
GPTBot (OpenAI / ChatGPT)
20
Allowed in robots.txt
ChatGPT-User (live browsing)
20
Allowed in robots.txt
PerplexityBot
20
Allowed in robots.txt
ClaudeBot (Anthropic)
20
Allowed in robots.txt
Google-Extended (Gemini / AI Overviews)
20
Allowed in robots.txt
Googlebot (baseline crawler)
5
Allowed in robots.txt
Bingbot (baseline crawler)
5
Allowed in robots.txt
Sub-scores are capped at 100 before the overall weight is applied.
Citability
25% weight
How easily an AI model can quote a coherent answer from your page.
LLMs preferentially cite content that is scannable, hierarchically structured, and leads with a substantive answer. Thin pages or walls of unstructured text get paraphrased away.
Data source — Cheerio-parsed HTML from Firecrawl; counts tags, inspects first paragraph, detects FAQ markers.
Sub-check
Points
Criteria
H1 present
8
Exactly one H1 in the document
H2 subheadings
7 / 4
≥2 H2 → 7pts; exactly 1 → 4pts
H3 sub-sections
5
At least one H3 tag
Answer-first opening
25 / 15
First <p> ≥50 words → 25pts; 30–50 words with question pattern → 15pts
Lists (ul/ol)
12 / 6
≥3 lists → 12pts; 1–2 → 6pts
Tables
8
At least one <table>
FAQ section
15
FAQPage JSON-LD, FAQ class/id, or FAQ heading detected
Content depth
up to 20
1000+ words earns full 20; prorated below
Sub-scores are capped at 100 before the overall weight is applied.
Schema Markup
15% weight
Whether structured data is present and valid for AI parsers.
Schema.org JSON-LD lets models resolve entities, authorship, and answer structure without guessing. Incomplete schemas earn partial credit because they still help, just less reliably.
Data source — All <script type="application/ld+json"> blocks (including @graph); validates required fields per type.
Sub-check
Points
Criteria
Organization (name, url, logo)
20 / 10
Full → 20pts; missing any required field → 10pts
Article (headline, author, datePublished)
20 / 10
Full → 20pts; partial → 10pts
FAQPage (mainEntity with questions)
20 / 10
Full → 20pts; partial → 10pts
HowTo (name, step with items)
15 / 8
Full → 15pts; partial → 8pts
Product (name, offers)
15 / 8
Full → 15pts; partial → 8pts
BreadcrumbList (itemListElement)
10 / 5
Full → 10pts; partial → 5pts
Sub-scores are capped at 100 before the overall weight is applied.
Technical SEO
15% weight
Baseline hygiene signals that AI crawlers use to trust a page.
Missing titles, broken canonical tags, or 8-second response times don't stop an AI model, but they make your page less trustworthy to rank and cite.
Data source — Cheerio HTML inspection + Firecrawl response metadata (timing, content-type).
Sub-check
Points
Criteria
Meta title (30–60 chars)
15
Passes length check
Meta description (120–160 chars)
15
Passes length check
Canonical tag
10
<link rel="canonical"> present
Open Graph tags (title + description + image)
15
All three present
Viewport meta
10
<meta name="viewport"> present
HTTPS
10
URL uses https://
Response time
10
First-byte response < 3000ms
Content-Type header
5
Contains text/html
Language attribute
5
<html lang="…"> set
Single H1
5
Exactly one H1 tag
Sub-scores are capped at 100 before the overall weight is applied.
Content Authority
15% weight
Signals that tell a model this page is worth trusting.
AI engines systematically over-cite pages with named authors, original data, and outbound links to recognized authorities. These are the strongest levers after citability.
Data source — HTML heuristics: author/byline selectors, table density, statistic regexes, outbound-link host matching, credential regexes.
Sub-check
Points
Criteria
Author attribution
25
rel=author, byline class/id, itemprop=author, or 'Written by …' pattern
Original data
20
Table with >3 rows OR ≥5 statistics (%, $, comma numbers) in body
Authority citations
up to 20
Links to gov/edu/Nature/Reuters/etc. — 5+ earns full 20
Credential signals
up to 15
Mentions Ph.D., MD, founder, certified, years of experience — 3+ unique earns full 15
Content depth
20 / 15 / 10
2000+ words AND 4+ sections → 20; 2000+ words → 15; 1000+ → 10; prorated below
Sub-scores are capped at 100 before the overall weight is applied.
Brand Presence
10% weight
Whether your brand is being talked about where AI models forage for context.
LLMs condition on where a brand is mentioned, not just on your own site. Reddit, LinkedIn, and general web mentions disproportionately influence how models describe you.
Data source — DataForSEO SERP API — quoted-brand searches scoped to reddit.com, linkedin.com, and the open web.
Sub-check
Points
Criteria
Reddit mentions
up to 30
10+ results earns full 30; prorated below
LinkedIn mentions
up to 30
10+ results earns full 30; prorated below
General web mentions
up to 20
10+ results earns full 20; prorated below
Platform diversity
20 / 10 / 5
3 platforms → 20; 2 → 10; 1 → 5
Sub-scores are capped at 100 before the overall weight is applied.
Versioning
v1.0 · 2026-04-15 — Initial published methodology. Six weighted dimensions; rubric pinned to the analyzer source of record.
Rubric changes will be versioned and dated here. Older reports stay tied to the methodology version that produced them.
Audit a site against this rubric
Every score uses the same published rubric. No paywall, no login.
