Domain authority now explains just 3% of AI citation selection, E-E-A-T explains 66%. Here is the author schema markup that closes the gap on every post.
A high domain authority score no longer carries an AI citation. A 2026 analysis of AI citation signals across major platforms found domain authority correlates with AI citation selection at r=0.18, explaining roughly 3% of the variation, while E-E-A-T signals correlate at r=0.81, explaining about 65% (Clairon, 2026). The engine reading your page is checking who stands behind it, and an unsigned, unaccountable AI post fails that check no matter how strong the domain is. Author schema, a real machine-verifiable Person entity attached to every post, is the concrete fix.
This is the part most teams shipping AI content have not adjusted to. They moved the volume up and left the byline blank, betting the domain would carry it. That bet stopped paying. If you want to be the source an answer engine quotes, the work has shifted from the domain to the person, and that is a structural change you can actually make. It builds on the strategies in answer engine optimization, and it leans on the same AI Overview behavior we covered in how to show up in Google AI Overviews.
Because the engine is grading the trustworthiness of the source, not the strength of the domain. Squared, an r=0.18 correlation explains roughly 3% of the variation in which pages get cited, so domain authority is now close to noise. E-E-A-T, at r=0.81, explains about 65.6% of it. An AI Mode Boost analysis of AI Overview citations across 63 industries, reported by Wellows in December 2025, found that 96% of the content pulled into Overviews comes from sources with strong, verifiable E-E-A-T signals.
A domain you spent years building is not worthless, but it no longer determines whether an individual post gets cited. The model is asking a different question: can I tell who wrote this, are they credible on this topic, and is anyone accountable for it. A blank byline on a high-authority domain answers none of that, so the page gets passed over for one that does.
Rank is no longer a proxy for citation either. An Ahrefs analysis of 863,000 keywords and 4 million AI Overview URLs found only 38% of cited pages ranked in the top 10, down from 76% just months earlier (Ahrefs, 2026). BrightEdge, tracking a separate 16-month rolling panel, puts the overlap even lower, at 17% (BrightEdge, 2026). The Overview is not reading down the ranked list and quoting the top result. It is assembling an answer from sources it trusts, wherever they sit, and trust is where the author becomes load-bearing. A page ranking eighth with a credible, verifiable author can be cited over the page ranking first without one.
A machine-verifiable author is a real Person entity an engine can resolve, attribute, and trust, not a name string at the top of a post. "By Jane Doe" in bold text is invisible to a model as an identity. It cannot tell whether Jane Doe is a real person, which Jane Doe she is, what she knows, or whether she exists at all. Turning that string into an entity is what makes the byline count.
Three things turn a name into an entity: structured markup that names the author as a Person, a real bio page that uniquely identifies them, and external profiles that confirm they are who the page says they are. Each one is cheap on its own. Together they are the difference between a byline a model ignores and one it can stand behind.
Start with the markup. In schema.org's Person type, the fields that carry weight for authorship are name, jobTitle, worksFor wired to the Organization that employs them, url pointing at a real bio page, knowsAbout for their areas of expertise, and hasCredential for qualifications. The knowsAbout field is also where author entities meet entity-based semantic SEO: it tells an engine what this person is an authority on.
Google's article structured data guidance is strict about how you fill these in. Put only the author's name in author.name: no job titles, no honorifics, no "Dr." or "Senior Editor" jammed into the string. Those belong in jobTitle and the dedicated honorific properties. Use the Person type for a human author, never the generic Thing type. And author.url has to be a page that uniquely identifies that one author, like their bio page or a profile, not a shared "/team" page that lists everyone.
sameAs is how an engine confirms this is the same person across the web. It takes a list of URLs that point to the author somewhere else, and each one is a reconciliation point: LinkedIn, GitHub, Crunchbase, and Wikidata. Wikidata feeds the knowledge graphs these engines lean on, making it the strongest entity signal you can provide for a publicly notable individual - it has notability requirements, so entries without meeting those criteria are routinely removed. Google accepts sameAs as an alternative to author.url when disambiguating who an author is.
The rule that trips people up: every link has to resolve and clearly identify the same person. A sameAs pointing at a dead LinkedIn URL, a renamed GitHub handle, or the wrong person with a shared name is worse than no link, because it breaks the reconciliation instead of completing it. This matters per engine too. On ChatGPT, LinkedIn now ranks as the second most-cited domain behind Wikipedia (SatelliteAI, 2026), which tells you how much weight profile and author entity signals carry in AI citation.
The author.url should point at a real page that exists only to identify that author, ideally marked up with ProfilePage structured data, which Google designed for exactly this: pages where a named creator shares first-hand perspective. A proper bio page gives the entity somewhere to live: the photo, the role, the expertise, the same sameAs links, all in one place a model can read.
Behind the markup sits the part that is not technical at all. A real author means a named human who is accountable for the post, and an editor or reviewer who signed off on it. That accountability chain, writer plus reviewer plus a byline that resolves to a real person, is what "trust" actually decomposes into when an engine grades a source. Markup without a real human behind it is a costume. The point is to make the costume unnecessary by having the real thing.
Here is the shape, with the rules baked in. The author is a Person, not a Thing; the name field holds only the name; worksFor links to the Organization; url points at a bio page that resolves; and sameAs lists profiles that confirm the identity.
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Author schema for AI citations: the 2026 E-E-A-T fix",
"author": {
"@type": "Person",
"name": "Mitrasish",
"jobTitle": "Co-founder",
"worksFor": {
"@type": "Organization",
"name": "Lyra",
"url": "https://www.trylyra.ai/"
},
"url": "https://example.com/team/your-name/",
"knowsAbout": ["SEO automation", "answer engine optimization", "content pipelines"],
"hasCredential": {
"@type": "EducationalOccupationalCredential",
"name": "Co-founder",
"credentialCategory": "Professional Experience"
},
"sameAs": [
"https://www.linkedin.com/in/mmitrasish/"
]
}
}Two rules decide whether any of this works. First, when a post has more than one author, give each one their own author field rather than merging them into a single string. Second, and this is the one that quietly kills most implementations: the markup has to match what is visible on the page. If the schema names an author the page does not, or claims a role the bio contradicts, Google may ignore the structured data entirely. The markup is a machine-readable copy of what a reader can already see, not a separate story you tell only the crawler. Correct, consistent implementation is what makes any of it count.
These are two different gates, and AI search wants both. Fact-checking proves the claims in a post are true: every stat sourced, every link live, every number dated. Author schema proves a credible, accountable person stands behind those claims. One verifies the content, the other verifies the source. A post can pass one and fail the other, and a model reading for citations notices the gap.
Think of it as two questions an engine asks before it quotes you. Is this accurate, and who says so. Verified facts answer the first. A resolvable author entity answers the second. Content that nails both is what gets pulled into an answer; content missing either gets skipped for a competitor who covered both. They reinforce each other, which is why a strong author signal pairs naturally with the extractable, answer-first structure covered in how to rank in ChatGPT: SatelliteAI's analysis of ChatGPT citation patterns found 72.4% of cited pages put a short, direct answer right after a question-based heading.
Most AI content ships unsigned and unowned, and that is exactly the test it fails. The output of a bulk generator has no author because no person wrote it and no person checked it. It is published into a CMS under a generic account, or under no author at all, with no bio page, no sameAs, and nobody accountable for a word of it. To an engine grading sources, that is a blank where the trust signal should be.
The numbers track this. A 16-month SE Ranking study of 2,000 AI-written articles across 20 brand-new domains, reported by Search Engine Land, found that only 3% of AI-generated pages held top-100 rankings after month three, down from a peak of 28% earlier in the test period. The fix is structural, not cosmetic: you cannot bolt a credible author onto a flood of anonymous posts after the fact and expect it to read as genuine. It has to be built into how the content is produced. Visible author credentials, the markup and metadata that make a person identifiable to an engine, lift AI citations by around 40% (Clairon, 2026), a large return for a structural change most pipelines skip. The accountability gap in mass-produced AI content is precisely what separates content that compounds from content that disappears.
A pull-request pipeline closes the gap by making authorship a property of how every post is shipped, not an afterthought. The mechanics are simple once they are structural. Every post is drafted and committed under a named human. The pull request is tagged for a human reviewer who reads the diff and merges it, so a real person signs off before anything goes live. And the Person JSON-LD, the sameAs links, and the bio link ship with the draft automatically, matched to what the page shows.
That is the model Lyra runs. She writes in your blog's existing voice, fact-checks every claim and link as a hard blocker, and opens the post as a GitHub pull request you review and merge. Nothing auto-publishes, so a named human is accountable for every post by construction, which is the exact accountability chain an AI engine is grading. The author markup is not a plugin you remember to configure; it is part of the draft, on every post, the same way the byline is. Want author schema and a real reviewer on every post you ship? You can request early access and we will walk through your blog.
Author identity is the one field most AI content pipelines systematically skip. The engines have shifted from grading the domain to grading the person behind the post. A real byline, a resolvable Person entity, and a named reviewer who signs off before publication form the accountability chain that separates content that gets cited from content that gets passed over. Sign the post, mark up the Person, and route every draft through a named reviewer who merges it.
Author schema is only worth anything if it ships on every post and matches a real, accountable human; Lyra writes under a named author, marks up the Person entity, and opens a pull request you merge.
Step by step
Publish a real author bio page
Create a uniquely identifying author page with a photo, role, expertise, and links out. Mark it up with ProfilePage structured data so engines read it as a profile, not a generic page.
Mark the author up as a Person entity
Add Article.author as a Person with name (name only), jobTitle, worksFor wired to the Organization, url to the bio page, knowsAbout, and hasCredential. Never use the Thing type.
Wire sameAs to profiles that resolve
List the author's LinkedIn, GitHub, Crunchbase, and Wikidata URLs in sameAs. Confirm each one loads and points to the same person before shipping.
Match the markup to the visible page
Make sure the author name, role, and bio in the schema match what a reader sees on the page. If the structured data and the page disagree, Google may ignore the markup entirely.
Make a named human accountable for every post
Commit each post under a real person and route it through a reviewer who signs off. The byline, the markup, and the sign-off form the accountability chain AI search is grading.
FAQ
Yes. AI answer engines weight E-E-A-T signals far more heavily than domain authority, and a clear author is the cheapest E-E-A-T signal you can add. Author schema marks up a real Person entity an engine can resolve and attribute, which is what lets it trust the page enough to cite it. A visible author byline backed by matching markup is one of the highest-return edits you can make.
Use the Person type for a named human author, and Organization only when the publisher itself is the author. Google's guidance is explicit: use Person for people, do not use the generic Thing type, and put only the name in author.name with no job titles or honorifics. Job title goes in its own jobTitle property, not in the name string.
sameAs is a list of URLs that prove the author is a specific, real person rather than a name string. Pointing it at a LinkedIn profile, a GitHub account, a Crunchbase entry, or a Wikidata item lets an engine reconcile the byline to a known entity. Every sameAs link must resolve and clearly identify the same person; a dead or mismatched profile hurts more than no link at all.
For AI citations, yes. In 2026 analyses, E-E-A-T signals correlate with AI citation selection at r=0.81, explaining about 65.6% of the variation, while domain authority correlates at only r=0.18, explaining about 3.2%. A high-authority domain no longer carries an unsigned, unaccountable post into an AI answer on its own; the trust signals around the author and the claims do.
Built by the tool you're reading about
Lyra finds the topics worth ranking for, writes them in your repo's voice, fact-checks every claim, and opens a pull request scored and ready to merge. You review and hit merge. Want to see what she'd write for you? Tell us about your blog and the founder will walk through it with you.
Keep reading
The robots.txt for AI bots setup that blocks training crawlers like GPTBot and ClaudeBot while allowing the search crawlers that get you cited in AI answers.
AI citation tracking, honestly: the free GA4 regex setup, the paid trackers compared, and why most of your AI search visibility stays invisible by design.
Programmatic SEO in 2026 survived the scaled-content-abuse crackdown. The new bar: one unique, verifiable data point per page plus a real editorial gate.