Signal Harbor Weekly: June 16, 2026
· Signal Harbor · Originally published in Signal Harbor Weekly
Welcome back to Signal Harbor Weekly.
One of the easiest mistakes to make in AI search right now is assuming that being cited means you're influencing the answer.
It doesn't.
A lot of companies are celebrating because their website appeared as a footnote in ChatGPT, Perplexity, or Google's AI results.
The assumption is simple:
"If the AI cited us, we must be winning."
But a recent study analyzing more than 21,000 AI search citations suggests the reality is more complicated.
THE INSIGHT
The researchers looked at two different concepts:
Citation Selection
Did the AI choose to cite your page?
Citation Absorption
Did the AI actually use information from your page when constructing its answer?
That distinction matters.
Because getting selected and getting absorbed are not the same thing.
A page can be cited without significantly influencing the response.
And a page that heavily influences the answer may end up contributing far more value than a page that simply appears in the source list.
THE NUMBERS
The study examined AI-generated answers across ChatGPT, Google AI, and Perplexity.
Perplexity cited the most sources on average:
Perplexity: 16.35 citations per prompt
Google: 12.06 citations per prompt
ChatGPT: 6.88 citations per prompt
At first glance, you might conclude that more citations are better.
But when researchers measured citation influence- the degree to which a cited page appeared to shape the final answer- the results looked very different.
Average influence scores were:
ChatGPT: 0.271
Perplexity: 0.065
Google: 0.058
The takeaway isn't that one platform is better than another.
It's that citation volume and answer influence are different measurements.
A company can be cited frequently without meaningfully shaping the response.
THE B2B PROBLEM
This is especially important for software companies and other complex purchases.
Imagine a buyer asks:
"What is the best construction management software for a residential remodeling company?"
You don't just want your homepage sitting in a list of sources.
You want the AI to understand:
Who your product is built for
What problems it solves
How it differs from alternatives
Why a buyer should choose it
In other words:
You don't just want to be in the bibliography.
You want to shape the recommendation.
THE OTHER FINDING
One result surprised me.
For months, marketers have repeated the idea that AI systems love Q&A content.
The study found little evidence for that.
Pages formatted as Q&A showed slightly lower average influence than non-Q&A pages.
That doesn't mean Q&A content is bad.
It simply means that formatting alone isn't enough.
Structure helps.
But useful information appears to matter more.
WHAT I'M WATCHING
As AI search evolves, I think we're going to see a shift away from simple visibility metrics.
Citation count is useful.
But citation count alone can be misleading.
The more interesting question is:
How much of your message is actually making its way into the answer?
That's where I believe this industry is headed.
Not just measuring whether you're cited.
Measuring whether you're understood.
That distinction is one of the reasons we built Signal Harbor.
Most AI visibility tools focus on whether a company appeared in an answer.
We're increasingly interested in a different question:
How much of the company's message actually made it into the answer?
Because being cited and being influential are not necessarily the same thing.
As research in this area develops, I think businesses will need to measure both.
And that's a very different problem.
The companies that win in AI search won't just be the ones that get cited.
They'll be the ones whose information gets absorbed.
Sebastian Miller
Co-Founder, Signal Harbor
signalharborconsulting.com
sebastian.miller@signalharborconsulting.com
402-306-2213
Source: Zhang, K., He, X., & Yao, J. (2026). From Citation Selection to Citation Absorption: A Measurement Framework for Generative Engine Optimization Across AI Search Platforms.
Research Note
Signal Harbor's work is informed by emerging research in generative search, information retrieval, recommendation systems, and AI trust, combined with ongoing analysis across major AI platforms. As this field develops, Signal Harbor will continue evaluating new research and refining how AI recommendation performance should be measured.
This article first appeared in Signal Harbor Weekly, the Signal Harbor newsletter on AI visibility.