A real scan of one live homepage, taken with the same analyser the free scanner runs on 2026-11-05. It is here so you can see the shape of the result - every failing check, its evidence and its recommended fix - before typing your own domain.
Published as a measurement of one homepage on one date, the same way the 30-site study publishes its rows. It is not a review of the company and not a ranking of its product.
77/100
overall GEO score, grade C
26/40
checks passing
13
failing checks listed
83%
of the 620 benchmark homepages scored lower
Illustrative photography (Unsplash License) — not the analyst who ran this scan, and no client or employee is depicted.
Dimension profile
The twelve weighted dimensions as one shape. A round profile means evenly ready; a spike means one dimension is carrying the score.
Ring = 25 / 50 / 75 / 100 · hover a point for its score
Dimension scores
Dimension
Score
AI Crawler Access
94
Machine Readability
100
Content Depth
73
Citability & Evidence
45
Answer Readiness
50
Trust & Authority
100
Semantic Structure
100
Metadata & Discoverability
86
AI Context Files
83
Freshness
0
International Readiness
100
Delivery & Mobile
50
Check results
13 need attention · 40 checks
6High7Medium0Low26Passed1N/A
Every failing check is listed below with the evidence it produced and the fix guide for its rule. The counts come from the stored scan of vercel.com on 2026-11-05; the weighted total is 77/100, grade C.
Score breakdown
12 dimensions, with the gaps visible
Score · weight
AI Crawler Access
94
16%
Machine Readability
100
12%
Content Depth
73
11%
Citability & Evidence
45
11%
Answer Readiness
50
10%
Trust & Authority
100
10%
Semantic Structure
100
8%
Metadata & Discoverability
86
7%
AI Context Files
83
5%
Freshness
0
5%
International Readiness
100
3%
Delivery & Mobile
50
2%
The score measures whether the page is technically ready to be cited. It does not measure whether an AI engine mentions the brand.
What does a 77/100 score mean?
It means the homepage passed 26 of 40 published checks and earned 77 of the 100 weighted points. The grade is C: average readiness — some of what a citation needs, and several dimensions need work. A scan measures whether the page is in a state that makes being cited possible - not whether an engine already cites it.
The weighted dimensions are published on the methodology page; the checks themselves are documented one by one on the check index. This sample is a C, which is the ordinary result: a homepage can be fully readable and still miss the evidence a quoted answer needs.
Which checks failed on vercel.com?
The 13 failing checks, highest severity first. Each row carries the evidence the analyser recorded and the fix it recommends; the full rule text is on the linked check page.
highAI Crawler Access
No content policy in robots.txt
What we found: robots.txt has no Content-Signal directive.
Fix: Add a line such as: Content-Signal: search=yes, ai-input=yes, ai-train=no. Early adoption and weighted low, but it states your position in the one file every crawler already reads.
highCitability & Evidence
No quotations
What we found: No <blockquote> element found.
Fix: Quoting a named expert is one of the highest-yield additions in the published research.
highAnswer Readiness
No question-shaped headings
What we found: None of the 17 H2 headings read as a question.
Fix: Engines retrieve spans that match a question. Label-style headings give them nothing to match.
highAI Context Files
No markdown alternate
What we found: No <link rel="alternate" type="text/markdown"> in the document head.
Fix: Publish a markdown version of the page and point at it with <link rel="alternate" type="text/markdown" href="...">, so an agent can fetch the content without the chrome around it.
highFreshness
No date is published
What we found: No dateModified, datePublished or Last-Modified header found.
Fix: Publish a machine-readable modification date.
highFreshness
No current year on the page
What we found: The year 2026 does not appear in the visible text.
Fix: Update the site so a current year is visible; stale date signals suppress generated-answer selection.
mediumContent Depth
Moderate body text
What we found: 550 words of visible text.
Fix: Generative engines favour sources that cover a topic fully. Aim for 800+ words on pages you want cited.
mediumContent Depth
Only one extractable format
What we found: Lists found, no tables.
Fix: Lists and tables are the formats engines extract most reliably. Add both where they fit.
mediumCitability & Evidence
Few quantified claims
What we found: 3 numeric claim(s) detected.
Fix: Add specific figures. Published GEO research reports the largest visibility gains from statistics and citations.
mediumCitability & Evidence
Only one authoritative citation
What we found: Cited: github.com.
Fix: Citing several independent authoritative sources is strongly associated with being cited in turn.
mediumAnswer Readiness
Q&A content without FAQPage schema
What we found: <details>/<summary> markup found but no FAQPage node.
Fix: Add FAQPage schema that matches the questions already visible on the page.
mediumMetadata & Discoverability
Meta description length is outside the useful range
What we found: 45 characters.
Fix: Aim for 50-160 characters.
mediumDelivery & Mobile
HTML payload is large
What we found: 578 KB of HTML before compression.
Fix: Trim the markup; parse budget is finite on both crawlers and devices.
What can this sample not tell you?
It reads one homepage, once, from one network location. It does not ask any AI engine about the brand, so it cannot say whether the site is mentioned, recommended or cited - that is the paid measurement. It also does not see the firewall a different address would meet, or the pages beyond the homepage.
The limits are the ones on the methodology page, stated in the same words there and in the scan result.
