A source about your brand can come to exist in at least six different ways, and no AI visibility report currently distinguishes between them.
That gap did not matter much in June. It matters now, and this post explains why.
The six states
Earned. A journalist found you and decided independently that you were worth writing about.
Editorially selected after a pitch. You raised your hand. An editor read it and agreed on the merits.
Contributed. You wrote it. The publication ran it under your byline.
Placed. An intermediary arranged it, possibly with genuine editorial review, possibly not.
Sponsored. Paid for, disclosed, frequently accurate and useful.
Paid. Paid for, undisclosed or disclosed so lightly that a reader would not notice.
None of these is automatically illegitimate. A placed article can go through real editorial scrutiny. A sponsored piece can be more accurate than a rushed news item written under deadline. Contributed columns are a normal part of trade publishing and have been for decades.
But they are not the same evidence about your brand, and that is the part that breaks.
The scan sees the publication, not how the coverage got there
When ChatGPT, Claude, Gemini or Perplexity cites a domain in an answer about your category, your visibility tool records a citation. It logs the domain, sometimes the URL, sometimes the sentence.
What it does not log is how that page came into existence.
An earned feature and a placed article arrive identically. Same domain. Same authority signal. Same increment to your number.
You can test this on your own report in about ten minutes. Take the source list, and for each one write a single word: earned, pitched, contributed, placed, sponsored, or paid.
If you cannot answer for a source, that is the answer.
The state most brands have the most of
Contributed content is the one worth thinking about hardest, because most B2B brands have far more of it than they realise.
A bylined column in a trade publication is your words, published under your name, on somebody else’s domain. It is not deceptive. It is a normal and long-standing part of trade publishing. Readers understand the convention.
But an AI system reading that page sees a page on an authoritative domain discussing your category and naming your company. Structurally it looks a great deal like a journalist writing about you. The convention that a reader understands is not encoded anywhere the retrieval system can see.
That is not an argument against contributed content. It is an argument for knowing how much of your evidence base is made of it, because a position built largely on your own bylines is a position built on one source of judgment wearing several mastheads.
Why this stopped being theoretical in August 2026
Until recently, improving your position in AI answers has been a diffuse activity. Publish more. Get mentioned more. Hope some of it lands.
On 25 August 2026, Featured launched a product called GEO Audit that identifies which publications AI engines are actually citing for a given category and ranks those publications as PR targets. It will draft the pitches. Their launch data covers 22,881 citations across 405 audits.
That is a legitimate product doing a legitimate thing. Working out which outlets influence AI answers and then pitching them is public relations with better targeting than guesswork.
But once the five sources that move your score have names, the work stops being diffuse and becomes a target list. Marketers will pursue those domains. That is not corruption. It is marketing responding rationally to a channel that has become measurable.
The consequence lands on the measurement rather than on the marketers. The population of sources stops being a natural sample and starts being an optimized one. A metric built on the assumption of a natural sample does not tell you what it used to tell you.
The same distinction just appeared in European law
On 2 August 2026, the transparency obligations in Article 50 of the EU AI Act became applicable. Providers of generative systems must mark synthetic output in a machine-readable format so it can be detected as artificially generated. Systems already on the market before that date have until 2 December 2026 to comply.
There is a carve-out worth reading closely. AI-generated text published to inform the public on matters of public interest does not require disclosure if it has been subject to human review and editorial control, with somebody identifiable holding editorial responsibility for it.
So in the disclosure regime, the line between labelled and unlabelled is editorial responsibility.
In the provenance question, the line between earned and placed is also editorial responsibility.
Same hinge, two regimes, and neither is visible to a system that counts domains.
What to do about it now
Three things, none of which require a new tool.
Answer the one-word question for every source in your report. Earned, pitched, contributed, placed, sponsored, paid. Keep the list.
Check the section, not just the domain. Many publications run branded content under a section name that reads like editorial. Brand voice, partner content, corporate news. A scan sees the domain.
Ask the question that actually decides things. If a competitor bought the same five placements next month, would your report be able to tell you why both of your numbers moved?
If the answer is no, then the metric is partly measuring spend, and it is measuring it in a currency nobody has priced.
FAQs
What is source provenance in AI visibility?
Provenance describes how a citation came to exist rather than where it appears. It covers whether coverage was earned independently, pitched and accepted, contributed, placed by an intermediary, sponsored, or paid for outright.
Is placed or sponsored coverage bad for AI visibility?
Not inherently. Placed and sponsored coverage can be accurate and can pass editorial review. The problem is that measurement tools record it identically to independently earned coverage, so the number cannot tell you which kind of position you hold.
Can an AI visibility tool detect whether coverage was paid?
Not reliably. Most provenance states cannot be determined from a URL, and for several of them nobody outside the publication and the client has the information at all.
What did the EU AI Act change on 2 August 2026?
Article 50 transparency obligations became applicable, requiring machine-readable marking of AI-generated output. AI-written text informing the public is exempt from disclosure where a human has reviewed it under editorial control and holds responsibility for it.
How do I audit my own citation provenance?
Take the source list from your visibility report and assign one word to each source: earned, pitched, contributed, placed, sponsored, or paid. Sources you cannot classify are themselves a finding.
Why does provenance matter more now than it did last year?
Tools launched in 2026 can identify the specific publications AI engines cite for a category, which makes source acquisition targetable. Once sources can be acquired deliberately, a citation count stops being a neutral sample of what the world says about you.
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.