CapstonAI is an AI visibility platform that helps brands track, audit, and optimize their presence in AI-generated answers. It monitors how your brand is mentioned across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews, then provides actionable fixes to improve your share of voice in AI search results.

How does CapstonAI track AI visibility?

CapstonAI sends structured brand queries to multiple large language models (LLMs) on a recurring basis and records whether your brand is mentioned, cited, or recommended. It aggregates these results into a GEO score, a share-of-voice metric, and a citation report.

Which AI models does CapstonAI monitor?

CapstonAI currently monitors ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Perplexity, and Google AI Overviews. Coverage is continuously expanded as new AI search surfaces emerge.

CapstonAI offers a free WordPress plugin that provides basic AI visibility scanning for your site. For full AI visibility tracking, audit reports, automated content fixes, and multi-brand monitoring, the Core Plan is available at $99/month.

How is CapstonAI different from Semrush or Ahrefs?

Semrush and Ahrefs are built for traditional search engine optimization. CapstonAI is purpose-built for the AI search era: it tracks your brand presence inside AI-generated answers from ChatGPT, Claude, Gemini, and Perplexity.

Earned Media Bias In AI Search: What The Toronto Study Reveals (Chen Et Al., 2025)

Earned Media Bias in AI Search: What the Toronto Study Reveals (Chen et al., 2025)

The single most replicated finding in academic GEO research is that AI Search engines systematically prefer earned media (third-party reviews, news, government sources) over brand-owned content. The University of Toronto study by Chen, Wang, Chen & Koudas (arXiv:2509.08919, 2025) quantifies this bias across multiple verticals and shows it holds across Claude, ChatGPT, Perplexity, and Gemini, in sharp contrast to Google’s more balanced source mix. Below: the exact percentages from the Toronto experiments, what they mean for brand strategy, and the 8-step playbook to win earned coverage that AI engines will cite.

TL;DR: Across automotive, consumer electronics, software, and other verticals, AI Search returned 67% to 92% earned-media citations, with Social platforms (Reddit, YouTube) nearly absent from Claude and ChatGPT outputs. Google by contrast retained a balanced Brand/Earned/Social mix. The strategic implication: shift content investment from brand-owned to earned-media outreach (PR, expert reviews, vertical publications) if you want AI citations.

Free CapstonAI scan → GEO Research hub

The exact numbers from the Toronto experiments

Chen et al. ran ranking-style prompts (e.g., “Top 10 brands…”) across web-enabled engines and classified every cited URL as Brand (official manufacturer or retailer), Earned (independent reviews, media, government), or Social (Reddit, YouTube, Quora, Facebook). The breakdowns by vertical are stark.

Vertical / Region	Google Brand	Google Earned	Google Social	AI Search Brand	AI Search Earned	AI Search Social
Automotive / Canada	36.6%	40.6%	22.8%	30.9%	69.1%	0%
Automotive / USA	39.5%	45.1%	15.4%	18.1%	81.9%	0%
Consumer Electronics / Canada	22.8%	54.1%	23.1%	22.1%	77.6%	0.3%
Consumer Electronics / USA	32.9%	~51%	15.4%	~8%	92.1%	~0%
Software Products / Canada	53.8%	~30%	~16%	~18%	~80%	~2%

Numbers from Chen et al., 2025, §4.2.1. Some cells rounded from chart approximations in the published figures.

The 8-step playbook for earned-media GEO

Step 1: Audit your current earned/brand/social ratio. Run 20-30 representative prompts across ChatGPT, Perplexity, Claude, Gemini. Classify every cited domain as Brand, Earned, or Social. If your brand domain captures more than 30% of citations, you are over-relying on owned content and under-investing in earned media.
Step 2: Map your top-cited earned domains by vertical. The Toronto data shows engine-specific patterns. ChatGPT cites Wikipedia, AP News, Reuters heavily in automotive. Perplexity surfaces YouTube and Car and Driver. Claude favors Consumer Reports, Car and Driver, US News. Identify your vertical’s top 10 cited domains per engine before spending on PR.
Step 3: Prioritize tier-1 publications cited by Claude and ChatGPT. These two engines are the most earned-heavy and the strictest gatekeepers. A single placement in a domain they cite repeatedly (e.g., Reuters, Wikipedia-eligible sourcing, Consumer Reports for autos) compounds across hundreds of generated answers.
Step 4: For Perplexity, add YouTube to the strategy. Perplexity is the only major engine that meaningfully cites Social sources (YouTube notably). A review on a relevant YouTube channel with strong watch time can be cited where it would never appear in Claude or ChatGPT outputs.
Step 5: Build a Wikipedia and Wikidata footprint. Encyclopedia sources outperform almost all other domain types in the Zhang, He & Yao (2026) absorption analysis (0.2144 mean influence vs. 0.0726 for news_media). A notable Wikipedia article with sourced citations is one of the highest-leverage GEO assets a brand can build.
Step 6: Engineer brand-owned pages to be quotable, not promotional. Even though earned media dominates, brand pages still appear in 8-30% of citations. Make those slots count: include extractable specifications, comparison tables, definitions, numerical data, and structured FAQs. Marketing fluff is filtered out by AI synthesis.
Step 7: Run quarterly cross-engine drift audits. Engine sourcing changes. Re-run your prompt panel quarterly and watch for: new earned domains rising in citation frequency, your earned-share trend (target: increasing), and engine-specific shifts (Gemini in particular sometimes increases brand share over time).
Step 8: Report earned-share to the C-suite as a leading KPI. Earned-share of citations leads revenue impact by 60-90 days in the CapstonAI Q1 2026 cohort. CFOs and CMOs accept earned-share growth as a defensible interim KPI while waiting for attributable-revenue signals to stabilize.

What “earned media” actually means in 2026

The Toronto classification was deliberately conservative. “Earned” included independent review sites (RTINGS, CNET, Car and Driver, Consumer Reports, US News), media outlets (Reuters, AP, NYT, WSJ, Forbes), government and institutional sources (energy.gov, ftc.gov, weforum.org), and encyclopedic sources (Wikipedia, Investopedia). “Brand” was strictly defined as manufacturer or retailer domains (e.g., ford.com, samsung.com). “Social” covered Reddit, YouTube, Quora, Facebook, TikTok, LinkedIn.

This matters for prioritization. Building a press placement in Forbes is high-leverage. Building a guest post on a low-traffic vertical blog is low-leverage. Earning a Consumer Reports review is one of the highest-leverage moves possible in automotive. Earning a Wikipedia citation (where notability allows) is the highest-leverage move possible in many categories.

Common errors when chasing earned-media GEO

Chasing low-DA placements. AI engines weight authority signals when selecting candidate sources. A guest post on a DR-30 blog rarely appears in citation pools. Prioritize DR-60+ outlets in your vertical.
Ignoring engine-specific patterns. Treating Claude and Perplexity identically wastes budget. Claude is earned-heavy and English-leaning; Perplexity includes YouTube. Different earned-media strategies apply.
Treating earned-share as a vanity metric. Earned-share without revenue attribution is incomplete. Pair it with the GEO attribution model (custom dimension ai_source in GA4) to close the loop.
Skipping Wikipedia eligibility. Brands often dismiss Wikipedia because of notability rules. If you are notable, Wikipedia is the single highest-influence domain type in the Zhang et al. dataset. Do not skip the eligibility analysis.
Over-investing in Social for ChatGPT. Reddit and YouTube barely appear in ChatGPT outputs. If ChatGPT is your priority engine, Social investment is misallocated.

FAQ — Earned media bias in AI Search

Does the earned-media bias hold across all verticals?

The Toronto study tested automotive, consumer electronics, software products, and several services categories. Earned dominance held across all of them, but the exact percentages varied. Consumer electronics in the US showed the most extreme bias (92.1% earned in AI Search). Service categories like airlines and streaming showed more overlap with Google because both systems converged on authoritative providers.

Why is Social media nearly absent from AI Search outputs?

Chen et al. interpret this as a structural shift: AI engines deprioritize community-driven and user-generated content in favor of third-party editorial and government sources. The only exception is Perplexity, which meaningfully includes YouTube and (to a lesser extent) Reddit. For Claude, ChatGPT, and Gemini, Social citations are below 5% in most verticals tested.

What about local search?

Local categories show much lower cross-engine overlap (often below 5% Jaccard) and fragmented domain ecosystems. Earned media still dominates, but the specific earned domains are highly localized (city dentist directories, regional service aggregators). Build local earned-media coverage in vertical-specific aggregators (Yelp, HomeAdvisor for services; OpenTable for restaurants; BookingHolidays for accommodations).

Tools and related reading

Ready to shift from brand-owned to earned-media GEO?

Free CapstonAI scan →

Last updated: May 2026. Primary source: Chen, M., Wang, X., Chen, K., & Koudas, N. (2025). Generative Engine Optimization: How to Dominate AI Search. University of Toronto. arXiv:2509.08919. https://arxiv.org/abs/2509.08919