A preliminary analysis of 599 mentions across 79 brands and an 18-week window. 

I noticed an interesting pattern while writing case studies for Noble. It seemed that after getting a certain amount of brand mentions onto the third-party sites LLMs are citing, what we call “offsite mentions”, AI visibility seemed to jump up more significantly and hold there. 

I couldn't quite pinpoint when that shift happened, just that something was changing once enough offsite mentions were live after a certain period of time.

So, to see if there was anything behind my hypothesis, we took a look at our customer data to work out how many offsite mentions it takes before you start to see a sustained lift in visibility, how quickly the lift can happen, and whether your starting AI visibility is a limiting factor to how much you can lift.  

How we ran this analysis

We observed and analyzed weekly visibility data from 79 brands in Noble's customer base: 599 brand mentions in third-party sites across 18 weeks, between January 26 and May 25, 2026. Visibility, to be clear, is measured by looking at how often your brand actually shows up in AI search answers. Meaning the percentage of questions where you show up is your visibility score. For example, a brand with 20% visibility is appearing in roughly one out of every five AI answers and a brand at 5% is in one out of 20.

The fuller methodology, including sample structure, definitions, and limitations, is in the section at the end of this piece.

Here's what we found.

Sustained visibility does depend on offsite mention volume

Across the 79 brands we focused on, 68% reached a visibility peak above where they started at some point after their first mention went live.

That isn't a particularly noteworthy finding on its own. Adding any mention is likely to nudge visibility somewhere above where it started, at least momentarily. The more interesting question is what happens to those lifts over time.

Does your brand still show up more often at the end of the window than it did at the start, or does the bump fade back out as the weeks go on?

That's what we mean by sustained lift. And whether a brand reached that sustained lift came down, more than anything else, to how many mentions it earned.

Here's what sustained lift looked like across mention volumes:

Look at where the sustained-lift numbers jump. 

At one to three mentions, only about a quarter of brands held their gains. At four or more, that more than doubles. And the gap is unlikely to be chance: brands with four or more mentions sustained a lift at more than double the rate of our control group (brands with no mentions at all), a difference that would turn up by random luck only about 3 times in 1,000.

Four mentions seems to be the starting line: below it, holding onto a lift is the exception; at four and above, it becomes the more likely outcome. The rate of sustaining a lift also doesn't tail off as mentions climb, but rather holds through the middle of the range and tops out at the very top, where all five brands with 21 or more mentions kept their lift. That's a small group, so we wouldn't read too much into the exact 100%. 

But it points the same way as the rest of the table: the brands most likely to hold their visibility were the ones that kept earning mentions, not the ones that stopped at a few.

How quickly visibility can lift

To measure the speed at which visibility can climb, we found the cleanest picture within a specific subset of the data: 27 brands that started essentially invisible in AI search responses (under 1% baseline visibility) and went on to secure four or more mentions.

Within four weeks of their first mention, this group went from effectively 0% visibility to 19%. That's a brand starting with no AI search presence at all and ending up cited in roughly one in five AI search responses within a month. 

This shows just how responsive AI engines are to quality information in credible, third-party sources. Place brand mentions in the right sources these models already read, and the models respond quickly.

One caveat is that results don't always move this fast. A month-to-meaningful-visibility climb is what's possible when mentions land in the right places, not a guarantee for every brand. But it happened cleanly across 27 brands here, and that's enough to show that strategic offsite mentions do work to quickly boost visibility.

And what about brands that already had some visibility going in?

They moved on the same timeline, just not by the same amount. Brands that started above 5% visibility and earned four or more mentions saw their lift land inside the first month too. The climb was simply smaller:

Both groups responded on the same clock: when the mentions landed, visibility moved after about a month. The only difference is the amount of growth. A brand starting from near-zero has room to leap to one in five answers; an established brand at 20% is more likely to add a few points, moving to roughly 24%. 

That might not sound dramatic, but think about what 4 points represents. A brand showing up in 20% of AI answers shifting to 24% is appearing in one additional question out of every 25. 

Across a category where users are running thousands of AI searches a day, that adds up.

Why this is happening

We can't open up ChatGPT or Perplexity and watch them decide which third-party sources are best. But the pattern in our data lines up with what's now well documented about how these engines choose what to cite. It comes down to two things.

More sources means more corroboration.

When an AI engine answers a question, it's synthesizing many sources: articles, reviews, comparison guides, forums. Outside research backs this up: when Ahrefs studied 75,000 brands, offsite mentions turned out to be the best predictor of whether a brand showed up in AI search. The more data points you provide for the AI, the more likely you are to see an increase in visibility. 

Different LLMs rely on different sources.

One analysis of 680 million citations found that only 11% of domains were cited by both ChatGPT and Perplexity. ChatGPT leans toward reference sites like Wikipedia, Perplexity toward community forums like Reddit, Google toward YouTube, Claude toward blogs… and those preferences shift over time. 

No single source even dominates within a platform: across 200 million prompts, Evertune found that even the most-cited domain on any given engine rarely accounts for more than about 5% of its citations. So, aiming for placements in a short list of sources doesn’t work. 

That same movement is why placing offsite mentions isn't a one-time exercise. The sources carrying your visibility today can drop out of an engine's rotation tomorrow, so presence has to be replenished to hold. 

More on what that means for your strategy below.

The takeaways

1. Brands need ongoing volume to produce sustained results. Below four mentions, lifts tend to be peak moments rather than sustained results. At four, they start to hold. Past it, visibility climbs higher and holds more reliably. The most-mentioned brands in our data were the most visible and the most stable.

2. Plan for at least a month before starting to measure results. AI engines need time to read and incorporate new sources. As the data showed, in the best cases tangible results occurred a month after a bulk of mentions went live. The key is to focus on continually placing mentions to create the most opportunity for a sustained lift. 

3. Your brand’s visibility starting position isn’t a limiting factor. The brands we analyzed that were invisible in AI search moved to one in five AI answers, and brands that already had a foothold lifted too once they added enough mentions. Established brands saw smaller gains in raw points, but that's mostly because they had less room to climb. Wherever you're starting from, the growth here is ripe for the taking.

How we conducted this research, and what it can't tell us

The dataset. Eighteen weeks of weekly visibility readings from Noble's customer base, between January 26 and May 25, 2026. Each row represents one brand-week.

The sample. 121 unique brands observed in total. Of those, 120 had visibility data: 79 also earned at least one mention during the period (these are the brands we focused on for the findings above), and 41 had no mentions, which we used as a control group to sense-check the threshold result. The one remaining brand earned mentions but had no visibility reading, so it falls out of the analysis. The total number of mentions recorded was 599. We count each brand individually, even when an agency manages multiple brands in our platform. Agency-level scores blend client outcomes together, so we look at each client brand on its own.

The customer profile. The brands in our dataset skew heavily toward B2B SaaS, with smaller representation from fintech, agencies, professional services, and a handful of consumer brands. Findings should be read as observations from a primarily B2B SaaS context.

Definitions. A mention live is the moment Noble has secured a brand placement in a third-party source that AI search engines treat as a citation. A baseline is a brand's visibility immediately before its first mention went live. A sustained lift means a brand's visibility at the end of the window was at least half a percentage point higher than its baseline, and we use that threshold to filter out tiny movements that could just be noise.

What this dataset can't tell us.

• 18 weeks isn't very long. Enough to see the threshold pattern, but not enough to say what happens over six or twelve months.
• The 21+ mention tier is based on five brands. The 100% finding there is striking, but with only five brands in that group, we'd treat it as a strong signal that needs more data rather than a fixed rule.
• Source effects aren't measured. A mention in a major publication and a mention in a niche review might land differently, but this analysis treats them the same. That's a question for a future piece.
• Whether this generalizes beyond B2B SaaS is an open question. Treat the threshold finding as well-evidenced for B2B SaaS specifically, and as a working hypothesis for everyone else.
• We watched what happened, we didn't run an experiment. We didn't randomly assign mentions to some brands and withhold them from others, so the pattern we're describing is what we observed, not what we can prove caused the visibility change.
• Brands at different mention volumes had different starting positions. Brands earning more mentions during our window tended to start with lower visibility. The 21+ mention group, for example, was made up entirely of brands starting near zero. This means direct comparisons of how much visibility grew across mention tiers aren't strictly apples-to-apples. The sustained-lift finding (binary: did the lift stick?) compares each brand against its own baseline, so it isn't affected. The trajectory section restricts to a comparable subset (cold-start brands) for the same reason.

A note on data dynamics: This analysis is a snapshot in time. AI search visibility, as measured by Scrunch and similar tools, is recalculated as the underlying AI engines update their responses to prompts. Historical readings can shift as those measurements refine. We're treating this as a preliminary look at a pattern we plan to continue tracking, and we expect to report on how findings evolve as our dataset grows and refreshes. This article was last updated on May 28, 2026.