All resources

How We Calculate the Noble Mention Score

Delia Rowland

November 12, 2026

•

4

minutes read

Blog

Every mention offer you receive is priced by its Noble Mention Score. You should be able to see what goes into that number and why.

‍

So this post lays out the whole thing: the formula, what each part measures, and the research behind each choice. Anything that was up to our judgment we called out too. 

‍

The score in one line

‍

The Noble Mention Score is the average of two numbers: a website's domain rating, and three times that website's citation rate on your prompts.

‍

Noble Mention Score = (domain rating + 3 × citation rate) ÷ 2

‍

Domain rating is a score from 0 to 100 for how established a website is, based mostly on how many other websites link to it. It's one of the most widely used measures of a site's strength.

‍

Citation rate is how often AI cites a website's most-cited article when it answers the questions you've told us your buyers ask (i.e., your prompts). If that article shows up as a source in 30 of every 100 answers to your prompts, the site's citation rate is 30. Using the site's top article, rather than any one specific article, keeps the score steady (more on that below).

‍

We take citation rates from the AI visibility tool you've connected to Noble. If we don't integrate with your tool, we use data from Scrunch. A site's citation rate stays fixed until we measure it again.

‍

Because citation rate is tripled before the average is taken, one point of citation rate adds three times as much to the score as one point of domain rating.

‍

Here's what that looks like for two made-up websites:

Website Domain rating Citation rate on your prompts Noble Mention Score
Site A 80 10 (80 + 30) ÷ 2 =55
Site B 40 30 (40 + 90) ÷ 2 =65

Example websites. Score = (domain rating + 3 × citation rate) ÷ 2

‍

Site A looks stronger by the usual measure, but Site B is the one AI actually reads when your buyers ask about your category, so it scores higher, and a mention there is worth more to you.

‍

Why we score websites, not individual articles

‍

We score the website an article sits on, not the article itself, because single articles are too unpredictable to price.

‍

The research backs this up. Ahrefs tracked about 37,500 citations across five AI tools, every day for 20 days. On ChatGPT and Google AI Mode, roughly two thirds to three quarters of the individual pages AI cited changed from one day to the next. For "best X" questions, where AI recommends brands, about 75% changed daily.

‍

Websites move around too, but the ones AI relies on most are far steadier. In a related Ahrefs dataset described in the same study, the brands and websites AI cited most often changed by only 11 to 14% from day to day. Everything else changed by 61 to 78%.

‍

So a single article is a shaky thing to put a price on. Whether AI cites it today says little about whether it will tomorrow. On the other hand, the website it lives on gives us a much steadier read.

‍

Why citation rate on your prompts counts most

‍

When a buyer asks AI a question, it often searches the web, builds its answer from the pages it finds, and lists some of them as sources. Those are its citations. Most of them aren't your website: when AI mentions a brand, 85% of the sources it cites are third-party, not the brand's own site. That’s why we say to show up in AI answers, you need to be in the source.

‍

Being mentioned on other sites goes hand in hand with being visible in AI. Across 75,000 brands, Ahrefs found that how often a brand is mentioned across the web tracks its AI visibility far more closely than domain rating or backlinks do.

‍

Citation rate carries the most weight because it measures what you're paying for: whether AI actually reads a website when your buyers ask about your category. A mention on a site AI never cites for those questions is unlikely to reach the answer.

‍

It's also measured on your prompts specifically. The same website can have a high citation rate for one company's prompts and a low one for another's, so we calculate it separately for each customer.

‍

There's a good reason for that. Profound looked at 11.8 billion AI citations across 29 industries and eight AI models. The kinds of websites AI leans on vary a lot, even between companies in the same industry. Within a single industry, the share of citations coming from independent publications differed by as much as 43 percentage points between companies at the high and low ends of the normal range. So even an industry-wide average would misprice mentions for many companies in that industry.

‍

A single score for every customer would overcharge some and undercharge others.

‍

That's why our rule is simple: if a domain is cited for the prompts you care about, it's relevant, and the more often it's cited, the more it's worth.

‍

Why domain rating is still in the score

‍

Domain rating is a secondary check. It tells us little about whether AI will read a site, but it helps screen out sites that aren't credible.

‍

The same Ahrefs study shows both sides of this. Pages AI cited for one day and pages it cited for 15 days or more came from similarly strong websites, with typical domain ratings between 64 and 78. Past a certain point, a higher rating didn't keep a citation around longer.

‍

At the bottom end it was different. Pages with no search traffic, no links from other sites, or very low authority showed up more often among the citations that vanished after a day.

‍

Semrush data analyzed by Kevin Indig points the same way. Across 1,094 categories in ChatGPT, the brand AI named most often had a higher authority score than the runner-up only 52.5% of the time. That study looked at brands named in answers rather than sources cited, but Indig's reading fits ours: strength may get you into the answer, while relevance to the question decides who leads.

‍

That's the job domain rating does in our score. It works as a floor, so a site with high citations and no credibility doesn't get priced like a trusted one.

‍

Why a snapshot can still price a moving target

‍

Both numbers in the score are measured at one point in time, and AI citations don't sit still. Josh Budman, our co-founder and CTO, who built the score, is upfront about that: "You're measuring a moving object at one point in time."

‍

It's a real limit. What the research shows is that the movement follows a pattern. In the Ahrefs study, 50 to 83% of citations were gone the next day, depending on the AI tool. But about 75% of the ones that dropped out came back within the 20 days, and about half returned within three days.

‍

How long a citation had already lasted mattered most. On ChatGPT and Google's AI tools, 96 to 99% of citations that had held for eight days or more came back after dropping out.

‍

That's also why the score uses a rate rather than a single check. When SparkToro had volunteers ask AI tools the same 12 recommendation questions roughly 3,000 times, there was less than a 1 in 100 chance of getting the same list twice. But how often a brand appeared across many runs was, in SparkToro's words, probably a reasonable way to tell how prominent it was.

‍

Over longer stretches, what AI relies on holds up. In the Semrush data, a brand that clearly led its category in ChatGPT kept first place in 90.4% of month-to-month comparisons. Again, that's brands in answers rather than cited sites, but it shows the same thing: day-to-day noise sits on top of a steadier pattern.

‍

That's the basis for pricing on a snapshot.

‍

A credible site that AI cites regularly for your prompts tends to keep returning, even when it disappears for a few days. No snapshot can promise a site will be cited next month, but how often AI cites a website today is a reasonable guide to whether it will keep reading it, and that's what decides what a mention there is worth.

‍

What the research hasn't settled yet

‍

Every input in the score points back to research, but the truth is that nobody has all of the answers about why LLMs choose to cite one source over another. So here’s what we’re still waiting to find out: 

‍

The weighting is our judgment. No study has tested how much citation rate should count against domain rating when pricing a mention. We chose three to one because the evidence says citation rate tells us far more about what AI reads.

‍

The evidence is recent and short. The Ahrefs study we lean on followed citations for 20 days. It shows how citations come and go, not how a website will behave over a full year. Ahrefs also notes that the pattern itself may shift when AI models are updated.

‍

The score measures the source, not your result. It tells you how often AI reads a site for your prompts. It can't tell you exactly how much one mention there will lift your visibility, and nobody has measured that cleanly yet.

‍

What this means for what you pay

‍

The score puts each mention into one of four price tiers. Full details are on our pricing page.

‍

Tier Noble Mention Score Noble fee per live mention Share of mentions in this tier
Standard 1 to 29 $100
11%
Plus 30 to 49 $200
43%
Premium 50 to 69 $400
38%
Elite 70 to 100 $600
8%

Publisher fees, typically $100 to $300, are separate. Full details on our pricing page.

‍

Some publishers also charge a fee to include your brand, typically $100 to $300 per mention. That's separate from Noble's fee, and you see it on each placement. On average, a mention costs about $400 all in.

‍

You also only pay for mentions that go live. Meaning your budget draws down when a mention publishes, not when an offer is made or accepted. Unused budget rolls over month to month, and anything still unused at the end of your annual term is refunded 90 days after the term ends.

‍

The research on AI Search is still young, and new studies keep arriving. When they change what we know about which sources AI reads, we’ll update this post. 

Let’s win AI search

GET DISCOVERED ON