How Earned Media Drives AI Search Visibility

Summary: Earned media drives AI search visibility by giving answer engines a third-party account of a company that they can retrieve and cite. But on the questions buyers ask before they spend, the third-party sources engines cite most are not newspapers. In the Machine Relations Index release of October 13, 2026, Reddit, YouTube and LinkedIn each out-cite every editorial publication. Forbes, the most-cited publication, is fourth.

Last updated: October 13, 2026. Release mri_score_v2.1+2026-10-13+6fc79cb02bbf.

The Short Answer #

Earned media is coverage, discussion and reference a company does not control: an article, a review listing, a forum thread, a video someone else made. Answer engines retrieve a small set of sources for each question and write from them. A third-party source lets the engine state something about a company without quoting the company about itself. That is the mechanism. It is real, and it is narrower than "get press".

The Index measured which sources six engines (ChatGPT, Claude, Gemini, Perplexity, Google AI Mode and Google AI Overviews) cite across 24,496 observed answer runs between May 10 and October 13, 2026. On the 18,915 runs that ask buyer questions (best tools, top lists, how to choose, is it worth it, comparisons and problem-first research), three findings change what earned media should mean in a plan:

1. The most-cited third-party sources are platforms, not publications. Reddit is cited in 10.94% of buyer-question runs, YouTube in 9.17% and LinkedIn in 4.69%. Forbes, the top editorial publication, is at 4.47%.
2. Publications are still the largest classified source class. Editorial publications carry 33.3% of the buyer-question citations the Index can classify, against 13.3% for community platforms.
3. Earned is not the only route. Brand-owned sites carry 24.6% of those classified citations. Microsoft, Palo Alto Networks and HubSpot sit in the top 25.

What Earned Media Means to an Engine #

The fifteen domains cited most often on buyer questions, with the class the Index assigns each one:

Selection rule: every domain in the release, ranked by the number of buyer-question runs in which it was cited, summed across all categories. News-topic questions are excluded.

Widen the list to the top 25 and it holds 8 editorial publications, 6 community platforms, 5 vendor-owned sites, 2 academic sources, 2 sources not yet classified, 1 video platform and 1 review database. A plan that counts only press placements is planning for one class of the earned sources engines reuse.

The engines also disagree about which earned source counts. Reddit, the most-cited domain on buyer questions, was cited by Gemini, Perplexity, Google AI Mode and Google AI Overviews and by neither ChatGPT nor Claude in this window (Reddit's Index profile). YouTube, LinkedIn and Forbes were cited by all six. The right earned source depends on which engine your buyer uses.

Earned Versus Owned on Buyer Questions #

Of the 119,050 domain-run citations on buyer questions, the Index classifies 36,066. Among those:

Every third-party class together carries 27,194 of the 36,066, or 75.4%. That is the measured version of "engines prefer earned media": three in four classified citations go to a source the company does not own. The other quarter goes to company sites, which is why the earned-versus-owned framing that gives owned pages almost nothing does not survive this data. The full class split by category and question shape, and why the wire class is ten domains, is on How PR Affects AI Search Visibility. Which company sites lead the head of the Index is on Company Websites Lead the Head of the AI Citation Index.

The caveat that matters most: 82,984 buyer-question citations, 69.7%, go to domains the Index has not yet classified. They are spread across 20,013 domains with a median of 2 buyer-question runs each, and the largest are retailers such as amazon.com and alibaba.com, finance sites such as bankrate.com and company sites such as apple.com. Every class share above describes the classified 30%, not the whole citation surface.

The Mechanism, and Where Studies Disagree #

Answer engines are retrieval systems. Google says its AI features draw on the same index and eligibility rules as the rest of Search (Google Search Central). OpenAI runs separate crawlers for search, user-triggered fetches and training (OpenAI crawler documentation), and Perplexity describes answers written from sources it retrieves and cites (Perplexity Help Center). Earned media drives visibility through that pipeline: an independent page that answers the question is retrievable, and it lets the engine attribute the claim to someone other than the company.

Controlled experiments by Chen and colleagues found AI search showed "a systematic and overwhelming bias towards Earned media (third-party, authoritative sources) over Brand-owned and Social content" (Chen et al., 2025). The Index agrees on the first half: three in four classified citations are third-party. It disagrees on social. On buyer questions in this release, community platforms and YouTube are the three most-cited domains. The two instruments use different prompts, engines and classifications, so the disagreement is a measured difference to test, not a refutation.

What to Plan For #

1. Count community and video as earned media. Candid practitioner discussion, explainer video and professional posts by people who are not you are where engines reach first on buyer questions. They cannot be bought like a placement; they are earned by being worth discussing.
2. Pursue publications for the questions they win. Editorial is the largest classified class, and its share is highest on best-tool questions and lowest on how-to-choose questions. Place coverage where the buyer question is answered, not only where the news is.
3. Keep your own pages answerable. A quarter of classified buyer-question citations go to company sites. Documentation and product pages that answer the question directly are cited alongside coverage, not instead of it.
4. Plan by engine. Reddit reaches four engines and not ChatGPT or Claude. Check which engines your buyers use before choosing which earned source to work on.
5. Measure at the page, not the domain. A domain's citation rate is not proof that your placement on it is cited. Whether a placement is cited depends on whether the question names what it is about.

For the ranking of publications alone, on B2B questions only, see Top Publications Cited by AI Search.

Frequently Asked Questions #

Does earned media drive AI search visibility? #

Yes, in the measured sense: on buyer questions in the October 13, 2026 Index release, 75.4% of classified citations went to sources the company does not own. Whether a specific placement caused a specific citation needs page-level evidence.

Is social media useless for AI visibility? #

No. Reddit, YouTube and LinkedIn are the three most-cited domains on buyer questions in this release, ahead of every publication.

Do brand-owned pages still get cited? #

Yes. Vendor-owned sites carry 24.6% of classified buyer-question citations, and Microsoft, Palo Alto Networks and HubSpot are among the 25 most-cited domains.

Which publication is cited most on buyer questions? #

Forbes, in 4.47% of buyer-question runs across all categories, cited by all six engines. Rankings differ by category and by B2B or consumer questions.

Does a press release count as earned media to an engine? #

Wire and press-release distribution carries 0.7% of classified buyer-question citations. A release is a distribution channel; the coverage and discussion it produces are what engines cite.

Methodology and Sources #

All Index figures come from the public Machine Relations Index dataset, release mri_score_v2.1+2026-10-13+6fc79cb02bbf: 30,093 cited domains across 24,496 answer runs from six engines, observed May 10 to October 13, 2026. Buyer questions are the six non-news question shapes, 18,915 runs. A domain's buyer-question count sums its runs cited across every category stratum of those six shapes; its share divides that count by 18,915. Class shares sum the same counts by the Index's declared source role. To reproduce, sum domains[].mri_score_v2.strata[].runs_cited where question_shape is not news_topic, grouped by source_role, in the public JSON at machinerelations.ai/data/machine-relations-index.json.

A citation records that an engine linked a source in an answer. It is not evidence that the answer was supported by the source, or that a company's coverage caused a recommendation. Machine Relations is the discipline this site documents (definition); AuthorityTech, a related party, is disclosed on the about page.

• Machine Relations Index
• Chen et al., 2025, Generative Engine Optimization: How to Dominate AI Search
• Google Search Central, AI features and your website
• OpenAI crawler documentation
• Perplexity Help Center, how Perplexity works