The Beige-ometer is Ghostart's check on a draft before it's published. It reads the draft against ten tests and gives two scores out of 10: Craft, for how well it's written, and Cite, for the evidence it carries.
Whether higher scores go with more AI citations is still open. Ghostart Labs registered a study on 10 September 2026 to test it, and we'll publish the result whichever way it falls.
Key facts
- It's a pre-publication check built into Ghostart.
- It gives you two scores out of 10, Craft and Cite, each with its reasons.
- It runs ten tests, five behind each score.
- It reads one draft at a time.
- Ghostart built it, drawing on The Beige Code by Andrew Seel and Nigel Jay Cooper.
What do Craft and Cite measure?
Craft measures the writing. Its five tests are Voice, Story, Emotion, Resonance and Rhythm.
For pages that speak as a company, such as answer pages, comparison pages and category pages, Story and Emotion are marked not applicable and left out of the average.
Cite measures the evidence inside the text. Its five tests are Statistics, Sources, Expert quotes, Answer first and Verifiable claims. A draft that names one real figure and its source scores better than one that gestures at five vague ones.
Cite reads one draft only. Brand mentions and links elsewhere on the web are measured separately, by Ghostart's visibility checks.
How are the scores shown?
Each score is a whole number out of 10, with a colour band and its name in words.
| Score | Band |
|---|---|
| 1 to 3 | A problem |
| 4 to 5 | Needs work |
| 6 to 7 | Doing well |
| 8 to 10 | Strong |
Select any test and it opens a card. The card shows the score, a one-line diagnosis, the lines from your draft it applies to, and a next step.
When you change the draft, Ghostart scores it again.
What does a real before and after look like?
On 8 October 2026, our co-founder Nigel Jay Cooper drafted a LinkedIn post about how AI tools decide which businesses to recommend. The Beige-ometer scored it Craft 5 and Cite 2. The card for Sources said no tool, study or link was named behind any claim about how AI works.
On 10 October 2026, Ghostart researched sources for the draft. The new version named Kevin Indig's "The Consensus Gap" (Growth Memo, 11 May 2026), which found that only 2.37% of cited URLs show up across all three engines for the same prompt. Cite rose from 2 to 6.
Craft stayed at 5, because the research changed the evidence and left the writing as it was.
Nigel then rewrote the post in his own words and moved the sources into footnotes. Voice rose from 7 to 8 and Craft to 6. Sources fell from 8 to 6, and Cite settled at 4.
The card gave the reason: "Two sources are named, but neither carries a link or full title."
| Score | Before (8 Oct) | After research (10 Oct) | After rewrite |
|---|---|---|---|
| Craft | 5 | 5 | 6 |
| Voice | 7 | 7 | 8 |
| Cite | 2 | 6 | 4 |
| Statistics | 2 | 8 | 6 |
| Sources | 1 | 8 | 6 |
This is one draft, scored three times. It shows how the scores respond to changes in the text. It doesn't show how AI tools responded to the post.
Does a higher Beige-ometer score mean more AI citations?
We think some of what the Beige-ometer measures may go with being cited more often. We don't claim it's proven.
Ghostart Labs registered a study on 10 September 2026 to test whether content with higher Beige-ometer scores is cited in AI answers at a higher rate than lower-scoring content. If the data doesn't support the hypothesis, we'll publish that too.
Until then, the scores are diagnostic checks of a draft. They aren't probabilities, and they don't predict whether any AI tool will cite a page.
Two pieces of published research explain why we test this carefully.
A 2024 KDD paper by Aggarwal and colleagues (Princeton and IIT Delhi) measured how changes to a source affected its visibility in generative-engine responses. Adding statistics scored about +33% and adding quotations about +41%. Citing sources scored about +28%, while keyword stuffing scored −8%.
The paper tests those edits, not the Beige-ometer.
A study by Schulte, Bleeker and Kaufmann (University of St. Gallen) found that only 34–42% of cited sources overlap between consecutive days. A single check is a snapshot, which is why we measure citations by repeated sampling.
Where does the Beige-ometer come from?
It comes from The Beige Code, by Andrew Seel and Nigel Jay Cooper, who founded Ghostart. The book asks why capable people and organisations so often produce writing that's competent and polished, yet forgettable.
Its argument is that expertise, evidence, specificity and judgement matter more as AI makes competent content easy to produce.
The Beige-ometer applies that argument to one draft at a time. Explore The Beige Code.
What does Ghostart do with the scores?
Ghostart is an AI visibility platform. It shows where ChatGPT, Claude, Gemini, Perplexity and Google AI Answers recommend your business, builds you a plan, creates content to plug the gaps and helps you build your AI visibility.
The Beige-ometer sits at the point where a draft becomes a page. Every draft gets scored before it goes out and scored again after each change, so you can see what moved and why.
Your team publishes content with its evidence named and its voice intact.
References
- Kevin Indig, "The Consensus Gap", Growth Memo, 11 May 2026. Found that only 2.37% of cited URLs show up across all three engines for the same prompt.
- Aggarwal et al. (Princeton / IIT Delhi), arXiv:2311.09735, KDD 2024. Measured the visibility effect of adding statistics, quotations and cited sources, and of keyword stuffing, in generative-engine responses.
- Schulte, Bleeker and Kaufmann (University of St. Gallen), arXiv:2604.07585. Found 34–42% overlap of cited sources between consecutive days.