Evidence Surfaces
The same record does not produce the same outcome when it is published in two different places. A verified date, a correctly spelled name, a decision with a known author and a known date: the material itself does not change, yet the surface it sits on decides whether it is picked up and cited. This page sets out the distinction between surfaces, the measured citation share of each one, and the text format that a single record takes on for each surface.
We use the word surface in one sense only: the publication location where a record stands in public. How the material is extracted, and what counts as a record in the first place, are covered on two separate pages. For the definition, see a company story is a record; for the collection method, see story extraction. This page deals with the distribution layer alone.
Three surfaces and their measured citation shares
The Omniscient Digital analysis, a vendor sourced study (8 January 2026; 240 branded queries, five answer engines, 23,387 citations), measured where citations come from on branded queries. The result splits across three surfaces.
| Surface | What it means | Citation share |
|---|---|---|
| Owned | The brand's own website and its own properties | 23 percent |
| Earned | Press, trade publications, association newsletters and comparable outlets where the publishing decision belongs to somebody else | 48 percent |
| Third party | Content produced about the brand by others | 30 percent |
The content type cut of the same data set shows how much room the customary address of the corporate narrative actually occupies: About Us pages 1.92 percent, homepages 1.82 percent, frequently asked questions 0.41 percent. These two distributions are two readings of one study, not two independent pieces of evidence. The scope is branded queries and the source is a vendor study; we state both alongside the figures.
The conclusion we draw is this. The larger part of a record is not prepared to look good on your own website; it is prepared in a form that somebody else can take over. Material that can be taken over carries a date, carries a name, can be quoted in a single sentence, and can be traced to a source.
The text format of one record per surface
The material comes out of a single interview cycle and a single archive pass. What changes is not the material but its form. The same record turns into the following formats.
| Text format | For which surface | What it contains |
|---|---|---|
| Interview transcript | Owned | Edited question and answer text, the speaker's name, the speaker's role and the interview date |
| Chronicle | Owned | A year by year list of events; every line carries a date, a decision and the person who took it |
| Signed personal article | Third party and professional network | First person text written in an individual voice rather than in a corporate voice |
| Fact file | Earned | A list of dates, names, figures and sources prepared for press and associations; verifiable facts, not claims |
| Quotable block | All of them | A short passage whose republication permission is stated explicitly inside the text, with the source credit embedded in the block |
| Structured data | Owned | The schema layer that renders the record readable by machines |
Structured data sits at the end of the list because it is a carrier, not a cause. The measurement Ahrefs ran between August 2025 and March 2026 compared 1,885 pages that received JSON-LD against 4,000 matched control pages using a difference in differences design. On the AI Overviews surface the effect was minus 4.6 percent, and on the other two surfaces the effect could not be separated from noise. We do not use this finding to dismiss schema work; schema keeps its place on our entity layer and methodology pages. What the finding does is relocate the gap: what is missing is not the technical layer but the record that layer is supposed to carry.
Where a numerical record is involved, publishing a data set is a further formal route. The Google Search Central documentation for Dataset lists name and description as required properties, and creator, license, identifier, temporalCoverage, variableMeasured, measurementTechnique and distribution as recommended ones. What the documentation promises is discoverability in Dataset Search, not citation by artificial intelligence. We pass that on exactly as it stands.
Publishing under a person's name, and the reason for it
We recommend publishing part of the text in an individual voice rather than a corporate one. The reason rests on a measured distribution. The analysis Meltwater published in its own press release (9.5 million citations, 16 B2B categories) reports that roughly 75 percent of citations arriving through the professional network come from individual profiles, and that 51 percent belong to people with fewer than 10,000 followers; in the same study the share of company websites is 18.7 percent. We note that the source is the company's own press release.
The practical consequence is this. A decision in the chronicle becomes capable of standing on a separate surface once it is written in the voice of the person who took it, with its date and its reasoning. Follower count does not appear to be decisive in this distribution, so publishing under an individual name does not require a large account. The authority behind the person conducting the interviews is set out on the founder page.
Text layer only on visual surfaces
Some surfaces are visual by nature. For those surfaces we prepare the text layer alone: transcript, descriptive copy, fact list, source credit and section headings. We do not produce the media itself.
On the video side the only data we hold is vendor sourced and covers a narrow area. According to BrightEdge data, video appears as a strong citation source on Google's answer surfaces (AI Overviews 29.5 percent, AI Mode 16.6 percent), while its share in chat engines is small (ChatGPT 0.2 percent). Without that narrowing we do not make any general statement about video. Nor do we imply in any form that visual originality increases the likelihood of being cited, because we hold no such measurement.
No dependency on a single third party surface
The share held by third party surfaces moves over time. In the tracking Semrush ran across 13 weeks between 14 July and 12 October 2025, covering more than 230,000 prompts and more than 100 million citations, the share of one community platform fell from roughly 60 percent to roughly 10 percent. We do not know the cause of that decline and we assign no causality to it. The single conclusion we draw sits on the distribution side: a record is not tied to one external surface but converted into more than one format.
How we define success
Our success metric is not how many times a page on this website, or on the client's website, has been cited. What we measure is the number of facts repeated on third party surfaces. Fact has a defined meaning here: a date, a name, a figure, or a decision with a known author and a known date. A fact counts when it appears on a surface we did not publish. The report lists which fact appeared on which surface on which date.
One reason for choosing this metric is that ranking and being cited are not the same thing. In the measurement Ahrefs published on 11 August 2025 across 15,000 long tail queries, the average overlap is 11.9 percent; the outlier pulling that average upward belongs to an engine we cannot name in customer copy. Our measurement and reporting setup is described on the methodology page, and our publishing discipline on the SEO safe publishing page.
What we do not promise
- We do not buy placement in earned media and we do not promise publication. The publishing decision belongs to your publisher. What we deliver is the fact file and the text.
- For surfaces that are visual by nature we prepare the text layer only. We do not shoot photography or video; we do not produce corporate films, promotional films, anniversary films, documentaries or animation. We do not offer editing, colour grading, voice over, drone, studio or production crew services.
- We do not imply that visual originality increases the likelihood of being cited. We hold no such measurement. The decorative images on this website come from stock sources and their credits are listed on the transparency page.
- We do not film interviews. The recording is an audio recording made solely for transcription; it is not a deliverable and it is not published. We do not scan, restore or digitise archive photographs.
- We do not guarantee citation, ranking or visibility. What we provide is measurement, a statement of method, and a report.
- We do not claim that structured data produces citations on its own. The Ahrefs measurement cited above indicates the opposite.
- We do not take on work while our capacity is full, and we do not accept anniversary projects when fewer than eight weeks remain before the event date.
Where to go next
The scope counterpart of this page is the scope and exclusions page, which sets out which step ends with which document, how many revision rounds are included, and which items fall outside scope. The answer engine visibility side is covered on the GEO page, the working setup on the working model page, and published work in the case studies section. If you would like to discuss your own record, you can reach us through the contact page.