Intro
Premium brands have spent two decades polishing their owned channels — website copy, brand stories, hero images. AI search engines do not reward that effort the way Google did.
A 2025 large-scale comparative study by Chen, Wang, Chen and Koudas (arXiv:2509.08919v1) measured how AI Search engines select sources versus how Google does. The finding is direct: AI Search exhibits a systematic and overwhelming preference for earned media — third-party, authoritative sources — over brand-owned and social content. Google’s mix is more balanced. The AI mix is not.
The pattern holds across multiple verticals, multiple languages, and paraphrased queries. It is not a quirk of one engine or one prompt. It is a structural property of how AI answer engines weigh source authority.
For premium brands, this changes the work. Owning the homepage is still necessary. It is no longer sufficient.
Audit your earned media footprint
What the research actually shows
The Chen et al. paper compares the cited sources behind AI answers (ChatGPT, Perplexity, Google AI Overviews, and others) against Google’s organic results for the same queries.
Three patterns matter for premium brands:
- Source mix is skewed. AI engines lean heavily on earned media: trade press, editorial reviews, established publishers, Wikipedia, expert blogs, structured directories. Brand-owned URLs appear in a minority of citations.
- Brand-owned and social content are deprioritized. Even when a brand’s site is well-optimized for traditional SEO, it tends to be a confirming source rather than the primary source AI engines quote.
- The bias is stable across paraphrases. Reword the prompt, change the language, switch verticals — the earned-media skew persists. It is not noise.
In plain terms: AI engines treat third-party authority as more trustworthy than self-description. That is a defensible heuristic. It is also a hard constraint on any brand that has invested mostly in owned channels.
Why AI engines prefer earned media
The reason is structural, not aesthetic.
AI answer engines are trained and tuned to minimize hallucination and brand-bias risk. When they synthesize an answer, they need sources that are:
- Verifiable — independent of the entity being described.
- Recent enough — refreshed on a cycle the engine can detect.
- Cross-referenced — cited by other sources in the same neighborhood.
- Stable — same URL, same claim, over time.
Brand-owned pages fail two of those tests by definition: they are not independent, and self-citations do not count as cross-references. Even excellent brand content is, from the engine’s perspective, an interested party speaking about itself.
This is why the Capston Core methodology treats earned media inventory as a first-class measurement, not a vanity layer. If an AI engine never sees the brand in third-party sources it already trusts, the brand will be missing from the answer regardless of how good the website is.
What this means for premium brands
The implication is sharper for premium brands than for mass-market ones.
Premium brands traditionally control their narrative through tightly managed owned channels, hand-picked partnerships, and selective press. That model produces beautiful brand-owned content and a thin, curated earned-media footprint. In AI search, thin earned media reads as low authority.
The competitors who win in AI answers are not always the strongest brands. They are the brands with the densest, most consistent, most recent third-party coverage on the specific prompts buyers actually use.
This is where the AI visibility scoring system becomes operational. Scoring without a source view is incomplete. The score has to tell the brand which earned sources AI engines reuse, which competitors own those sources, and which gaps are the cheapest to close.
How to respond without becoming a content farm
The wrong response is to flood the web with low-quality placements. AI engines detect that pattern and discount it.
The right response is a short list of disciplined moves:
- Map the source set. Identify the third-party sources AI engines actually cite for the brand’s prompt set. Not the publications the brand wishes were cited — the ones engines already reuse.
- Prioritize repeat sources. Authority compounds when the same trusted source mentions the brand more than once on different angles.
- Get the facts straight in their content, not yours. Wrong room counts, outdated chef names, stale ratings — AI engines will pick them up from earned sources before the brand-owned site.
- Build cross-references. When two earned sources mention the same fact about the brand, the engine treats the fact as verified.
- Hold the line on tone. Premium brands do not need volume. They need precision in the sources that matter.
This is where the data and evidence layer carries the weight. Every recommended placement is traced back to a specific prompt, a specific engine, a specific source the engine already trusts.
How this fits into Capston Core
Earned media bias is not a side topic. It is one of the load-bearing reasons the Capston Core silo exists.
The mapping connects to:
- The Capston Core methodology — the five-stage process that includes earned-media inventory as a measured stage.
- The AI visibility scoring — where source quality is one of the eight dimensions.
- The data and evidence layer — where every cited source is logged, dated, and re-checkable.
- The Capston Hospitality Scorecard — where the earned-media gap is benchmarked against direct competitors.
→ Back to Capston Core
FAQ
Does this mean brand-owned content no longer matters?
No. Brand-owned content remains the canonical reference for facts AI engines verify against earned sources. It is necessary but not sufficient. The bias is about which sources engines quote, not which sources they read.
Is earned media bias the same across ChatGPT, Perplexity, and Google AI Overviews?
The direction is the same — all three lean to earned media — but the intensity varies. The Chen et al. study documents the pattern across multiple engines; Capston Core measures the per-engine intensity per brand.
How long does it take to shift the source mix for a brand?
Earned media is a slow-moving asset. Realistic horizon for a measurable shift in citation share is two to three quarters, assuming a disciplined plan and accurate facts.
Does paid placement count as earned media?
Not in the sense AI engines reward. Engines tend to discount sponsored or syndicated content. The signal that matters is independent editorial coverage with stable URLs.
Reference
Chen, J., Wang, Y., Chen, L., & Koudas, N. (2025). Generative Engine Optimization: How to Dominate AI Search. arXiv preprint arXiv:2509.08919v1. The study performs a large-scale comparative analysis of AI Search engines and Google across verticals, languages, and paraphrased queries, documenting a systematic preference in AI Search for earned media over brand-owned and social sources.
Final CTA block
See which earned sources AI engines cite for your brand.
Run the earned media audit
Read the methodology
Certification: early access — applications open.