Author Credentials and Citation Mechanics: Schema.org Says knowsAbout Suggests Expertise Without Implying It
The vocabulary that carries an author's credentials is explicit about how little it asserts. Here is what each property is documented to state, and the three things the engines asked for that no byline can carry.
Published by AI Knows Us · Updated 16 October 2026
Read the definition of the property everybody reaches for first. Schema.org defines knowsAbout, on the Person page read on 16 October 2026 at schema.org/Person, as a way "to indicate a topic that is known about", then adds in the same sentence that this is "suggesting possible expertise but not implying it." The vocabulary declines to make the claim the market makes on its behalf. The same page defines hasCredential as "A credential awarded to the Person or Organization", which records an award rather than verifying one. So an author block is a set of structured assertions about a person, published by that person's employer. That is useful and it is not evidence. When we asked ChatGPT and Gemini directly, in 102 build spec conversations on 15 October 2026, what a page would need before they would cite it, the top items were a date with a refresh frequency, a sample size and a published prompt set. A byline carries none of those three. It carries one of the ten, and that one is worth knowing.
What each credential property is documented to state
Every definition below is quoted from the schema.org Person page as read on 16 October 2026. None of it is an inference about what an engine does with the property.
| Property | What schema.org says it is | Who can check it |
|---|---|---|
| knowsAbout | Indicates "a topic that is known about", "suggesting possible expertise but not implying it". The page adds: "We do not distinguish skill levels here" | Nobody. It is a topic label |
| hasCredential | "A credential awarded to the Person or Organization" | The awarding body, if you name it and link its register |
| honorificSuffix | "An honorific suffix following a Person's name such as M.D./PhD/MSCSW" | The institution that conferred it |
| worksFor | "Organizations that the person works for" | Anyone, against the employer's own site and a public filing |
Notice the right hand column. Two of the four point at a body that is not you, and those two are worth the effort. Google's Article structured data documentation, read on 16 October 2026 at developers.google.com/search/docs/appearance/structured-data/article and carrying its own last updated date of 8 September 2026, points the same way on types: "Use the Person type for people, and the Organization type for organizations. Don't use the Thing type, and don't use the wrong type." Precision about which kind of entity you are describing is the job. Confirmation comes from somewhere else.
The three asks a byline cannot satisfy
Set the author block beside what the engines actually asked for and the division is clean. Of the 102 build spec conversations on 15 October 2026, all 102 asked for a date, a last verified line and a refresh frequency, 80 asked for the sample size N, and 78 asked for the prompt set to be published so the reading could be taken again. A named author with a doctorate and a full credential block moves none of those three. They are properties of the work, not of the person who signed it.
One ask does land on the byline, and it is the one most pages leave out. 64 of the 102 asked for the commercial interest to be disclosed on the page itself, and an author block is where that sentence belongs, because the person named is the person with the interest. A line saying who employs the writer and what that employer sells is a credential of a kind: checkable, specific, and almost nobody in this category publishes it.
What the citation data says about where authority sat
Take the mix from the widest run we hold, dated 16 October 2026, in which 630 blind buying questions produced 1,260 readings with none discarded and 20,143 citations over 4,947 distinct domains. Sorted by who owned the page, 82.0% sat on a vendor or agency own site and 12.3% on official or platform documentation. News or a large publisher took 3.2%, social or community 1.9%, and review directories 0.7%.
Two readings of that mix matter for an author page. The 12.3% layer is documentation, where the publisher is an institution rather than a named human, so a byline is not what put it there. And four fifths of the field is self published, which means an author page competes with other author pages rather than with an academic record. The 4,947 domains carrying 20,143 citations work out at 4.07 citations per domain, so the field is wide rather than closed. Where that authority comes from on the firm's side is worked through in what AI assistants actually cite across 597 readings of ChatGPT and Gemini in three markets.
How to build an author page that holds something checkable
- Name the awarding body, not the letters. hasCredential records an award, so put the body's name and its public register page beside it and the claim has a destination.
- Use knowsAbout as a topic index, which is what it is for. Two or three subjects the person genuinely publishes on, treated as navigation.
- Put the disclosure in the bio. One sentence: who pays this writer and what that company sells.
- Date the credential. A board certification, a licence or a chartered status has a year on it and usually a renewal, and the year is the part a reader checks first.
- Keep the markup and the visible page saying the same thing, and fetch the page without a browser to confirm the block is in the response the server sent. Markup injected after load is markup some agents never read.
Building a public profile for a named individual is a discipline of its own, and what a personal branding engagement contains and how to brief it sets out that work when it is bought in.
The study this page is not, and the one that would settle it
A measured citation rate for pages with Person schema against anonymous pages does not exist in our records, and no number of that shape appears above. Producing one needs matched URL counts on one domain, a frozen prompt set that never names the business, publication and confirmed indexing dates logged per URL, at least three URLs in each arm, repeats so the rate carries a range, and a holdout set left untouched. Page age, topic and link position all move with author markup on a real site, so a clean reading needs them matched. The round that produces the figure will arrive with its own numerator, denominator and date.
Questions about author credentials and AI citations
Does adding Person schema get an article cited more often?
Nobody has shown that it does. The vocabulary describes identity and topic, OpenAI and Google each document what their own markup is for without describing it as something that decides a citation, and we have run no test of our own. Add the markup because it is cheap and correct, then measure whether the engines name you.
What does knowsAbout tell an engine about expertise?
By its own definition, that a topic is known about, "suggesting possible expertise but not implying it". The page adds that it does not distinguish skill levels. It is a topic label.
Do outbound citations to academic sources make a page more citable?
We have not measured that and no figure for it appears here. What the measurement shows is where the engines went instead: 12.3% of the 20,143 citations recorded on 16 October 2026 landed on official or platform documentation and 0.7% on review directories. Linking a primary source is good practice for a reader whatever it does downstream.
Should a company byline be used instead of a person?
Google's documentation says to use the Person type for people and the Organization type for organizations, and not to use the wrong type for either. If a named person wrote the piece, name them and give them a page.
Which credential field is worth the most effort?
hasCredential, with the awarding body named and its register linked, because it is the one entry on an author page that somebody other than you maintains. A credential nobody can confirm is a sentence about yourself.
Which engines produced the citation figures above?
ChatGPT and Gemini, read on 15 and 16 October 2026. The other three surfaces our product covers, Perplexity, Claude and Google AI Overviews, were never put these questions, so no figure here reads on them.
What to do with the author page you already have
Open it and count how many claims on it somebody outside the company could confirm this afternoon. On most author pages the answer is zero: a photograph, a job title, a list of interests and a sentence about years of experience. Replace one of those with a credential whose awarding body is named and linked, and add the line saying what your employer sells. Two edits, and between them they move the page from asserting standing to publishing something a sceptical reader can check. AI Knows Us wrote this page and sells the measurement and the publishing that follows it, which is the disclosure 64 of 102 build spec conversations asked for, sitting where it belongs rather than at the foot of a terms page.