Executive Summary
Most brands are optimizing only half of AEO. On-site work — cleaner pages, better schema, more blog content — is necessary but insufficient for earning AI citations. The real leverage sits off-site, in earned media.
The Most Common Mistake in AEO
Most brands and agencies treat Answer Engine Optimization as a content project. They write clearer pages, add more schema, and refresh your blog. This work matters, but it is not enough.
In 2025, a study from the University of Toronto analyzed thousands of AI citations across multiple large language models. The finding was stark: AI systems cite earned media approximately five times more frequently than they cite brand‑owned websites.
Your own site is part of the story. But the AI is reading your press mentions, your podcast appearances, your Wikipedia references, your analyst coverage, your Substack archives, your third‑party reviews, and your competitor’s comparison pages that happen to name you. Those sources shape what the AI says about you, often more than your own copy does.
A second finding from the same body of research: entity authority correlates 4.8 times more strongly with AI citations than technical SEO optimization does. Page speed, meta tags, and internal linking structures are not the primary drivers. Rather, how consistently and credibly your brand entity is referenced across the open web is the driver.
On‑site AEO is necessary. Off‑site AEO is where the leverage lives.
What the Research Actually Says
University of Toronto (2025): Earned media is cited approximately 5x more frequently than brand‑owned content in AI‑generated answers.
5W PR study (2025): 85.5% of AI citations come from earned media sources, not from corporate domains.
PromptWatch (2025): Entity salience correlation with AI citations = 0.386. Technical SEO correlation = 0.080. The gap is nearly fivefold.
These numbers are measured from actual AI outputs across ChatGPT, Perplexity, Claude, and Gemini.
What Counts as Earned Media for AEO Purposes
Not all third‑party mentions are equal. The AI’s retrieval layer prioritizes sources that have their own authority, structure, and permanence.
Wikipedia and Wikidata entries are among the most trusted signals. A well‑maintained Wikipedia page or a complete Wikidata entry gives the AI a stable, citable entity reference.
Professional association listings such as IEEE, AMA, ABA, or AICPA carry domain‑specific authority. If you are listed there, the AI treats those mentions as verified credentials.
Also, analyst reports from Gartner, Forrester, and IDC are heavily weighted. Being cited in a report, even briefly, creates a durable citation the AI will return to.
Furthermore, verified review platforms like G2, Capterra, and Trustpilot matter, especially when they use structured data markup. The AI reads the text of the review, not just the star rating.
In addition, podcast episode transcripts that are indexed and searchable function as mini‑articles. If a podcast publishes show notes as HTML, the AI can cite your exact quote.
Substack posts and independent newsletters with established domain authority are treated similarly to traditional media. A well‑archived newsletter is a permanent citation.
News articles from outlets that publish structured metadata (author bylines, publication dates, schema.org markup) are easier for AI to parse and cite correctly.
University and .edu mentions carry an authority premium. A faculty profile, a research acknowledgment, or a guest lecture announcement all count.
Government data sources (.gov) are treated as highly authoritative for factual claims. If your brand appears there, the AI will favor that signal.
Content on your own blog, even if well‑structured, is treated as self‑attested. The same claim made on a third‑party site with independent authority carries more weight with the AI.
The Off‑Site AEO Audit
Most brands have never looked at their earned media footprint through the lens of AI retrieval. Here is what you will find when you do.
Old press releases that describe products or pricing you no longer offer, news articles that quote a former executive who left three years ago, podcast transcripts where you said something that is no longer true, review site listings with outdated feature sets, and wikipedia entries that are thin or missing entirely.
These sources are not neutral. They are actively shaping what AI says about you, often in ways that are obsolete or flatly wrong. And because they live on domains you do not control, you cannot edit them with a CMS login. You have to engage, correct, update, or replace them.
That is the off‑site AEO workflow. It is not a content production treadmill. It is a reputation and entity governance process.
How to Build a Footprint That AI Trusts
Publish to earned media channels deliberately, not reactively. When you launch a feature, do not just put it on your blog. Pitch it to an industry publication that will publish it with stable URLs and clean metadata.
Get your Wikipedia page right. If you do not qualify for a Wikipedia page, get mentioned on relevant Wikipedia pages as a cited source.
Appear on podcasts that publish show notes and transcripts as searchable HTML. The AI reads the transcript.
Claim and verify your profiles on Wikidata, Crunchbase, and professional directories. Fill out every field. Keep them current.
Encourage detailed, structured reviews on platforms that AI trusts (G2, Capterra). The AI reads the text of the review, not just the star rating.
Audit your old earned media at least quarterly. Flag what is outdated and contact the publisher to correct or remove. For sources that cannot be changed, publish new, authoritative content that supersedes them. The AI’s recency bias will favor the newer signal.
Read more about how to optimize your business for AI here.
Why the AIBAI Diagnostic Looks Outside Your Site
The AIBAI Diagnostic does not stop at your domain. It maps the entire retrieval‑layer footprint across all four major AI platforms: your site, your earned media mentions, your reviews, your directory listings, your competitor’s pages that mention you. It shows you where the AI is pulling from and where the gaps and misrepresentations live.
On‑site AEO is half the job. Off‑site AEO is the other half. Most brands have not even started the second half.