Most AI visibility reports show one number where there should be two, and the fused version systematically overstates how well corroborated a brand is.
The two measurements answer different questions. Both are useful. Neither substitutes for the other.
Retrieval sources
A retrieval source is a distinct place an AI engine can pull information about you from.
If your brand is listed on four separate domains, ChatGPT, Claude, Gemini and Perplexity each have four distinct documents they might surface, each with its own authority signals, its own crawl behavior and its own chance of matching a given query.
More retrieval sources is genuinely better. Four listings give you four opportunities to appear in an answer where one listing gives you one. Nothing about shared ownership between those domains reduces that.
This is what a citation tracker measures, and it measures it correctly. It is a count of surfaces.
Independent judgments
An independent judgment is a separate party that actually reached a conclusion about you.
Three properties that run on one shared review catalog present three surfaces and carry one judgment, because a review submitted to any one of them appears on all three. The same rating, rendered three times.
This is not a hypothetical arrangement. Capterra, GetApp and Software Advice are widely reported to share a single review catalog, and G2 acquired all three from Gartner in a deal announced in January 2026 and since closed. The catalogs are still separate between G2 and the other three today, which means a G2 review remains a genuinely different judgment. One corporate decision away, it may not be.
A count of judgments is a count of opinions. It is a completely different measurement from a count of surfaces.
Why the fused number breaks
When a report combines the two, it takes the higher number and applies the more valuable label to it.
Four surfaces get reported as four corroborations. The score rises accordingly. Nobody involved is lying, and the number is nonetheless wrong about the thing that matters.
Two consequences follow.
Overstated confidence. The evidence base looks more robust than it is. A brand reading four independent sources reasonably concludes that four separate parties have formed a view about it.
Hidden correlation risk. Surfaces that share a source move together. A position that looks diversified is actually concentrated, and that only becomes visible on the day the shared source changes.
Reddit’s citation share in ChatGPT fell 86 percent this month for reasons that are still not fully explained. Any brand whose evidence was concentrated in one place learned what concentration costs, with about six days of warning.
What a report showing both looks like
The display is not complicated:
Retrieval sources: 4. Four separate domains, four chances to be pulled into an answer.
Independent judgments: 2. One shared review catalog counts once. The second is a genuinely separate party.
Context sources: 1. Appeared in the answer, said nothing about this brand, recorded but not counted.
Every excluded item says why it was excluded, in one sentence, next to the thing excluded.
A brand reading that can tell at a glance where its position is strong, where it is concentrated, and what a single change would cost it. A brand reading “4 independent sources” cannot tell any of that.
Three questions for any vendor
Does your report separate how many places cite me from how many parties have formed a view about me?
How do you determine whether two sources are independent of each other? The answer should mention shared content, not only corporate ownership, because ownership alone is a weaker test.
What happens to a source you cannot classify? If the answer is that everything gets assigned to something, the report cannot express uncertainty, and a system that cannot say “I do not know” will invent an answer to fill the column.
A worked example
A B2B software brand runs a scan and the report returns six sources.
Three are software marketplaces that share a review catalog. One is a trade publication that covered a funding round. One is a community thread. One is the brand’s own documentation site, surfaced because it answered the question directly.
The naive count is six independent sources.
The honest read is three independent judgments: the marketplace group counts once, the publication counts once, the community thread counts once. The documentation site is owned, so it is not independent corroboration at all, though it is a genuinely valuable retrieval surface.
Six retrieval sources. Three independent judgments. One owned. Every one of those facts is actionable and none of them survives being averaged into a single score.
We applied this to our own scoring in August and our own Evidence score halved as a result. We published the change rather than shipping it quietly. A measurement layer that will not disclose its own methodology corrections is asking the market to take its word for it, which is exactly the position we tell brands not to accept.
FAQ
Which number matters more, retrieval sources or independent judgments?
It depends on the question. Retrieval sources tell you how many chances you have to appear in an answer. Independent judgments tell you how well corroborated you are and how much a single source change would cost you.
Can a brand have many retrieval sources and only one independent judgment?
Yes, and it is more common than it looks. Syndicated listings, shared review catalogs and republished content all create multiple surfaces from a single underlying source of opinion.
Does this affect ChatGPT, Claude, Gemini and Perplexity differently?
The retrieval side varies by engine, since each has different crawl coverage and different citation behavior. The independence question is identical across all of them, because it is a property of your source relationships rather than of any engine.
How would I calculate independent judgments manually?
List every source citing you, look up the parent company of each, then check whether any of them share content such as a review pool or a syndicated listing. Group any that share content, and count the groups rather than the sites.
Is a lower independent judgment count bad news?
It is more accurate news. The count was always what it was, and knowing the real figure tells you where to build next rather than leaving you confident about a position you do not have.
Should retrieval sources be dropped from reporting entirely?
No. Retrieval surface is real and valuable, and a brand actively building listings should be able to see that work reflected. The error is labeling it as corroboration.
ABOUT AXIS SUITE
Axis Suite by TrendAxis is the independent intelligence layer that explains what AI believes about your brand, why it believes it, and what decision that belief ultimately drives.
Most tools count citations. Axis Suite tells you which of those citations are actually evidence about you, how many separate parties they represent, and what would remain if your largest source disappeared.
Run a scan at axissuite.ai or read the methodology in the Proof Center.