AEO Case Studies and Evidence: What Actually Moves AI Visibility

The most rigorous public study of AI answer visibility is "GEO: Generative Engine Optimization" by Aggarwal et al., published at KDD 2024. The researchers built a benchmark of queries, rewrote source pages using different optimization methods, and measured how much of each source appeared in generative engine answers. Their headline finding: GEO methods can boost visibility by up to 40% in generative engine responses. The top-performing methods were Cite Sources, Quotation Addition and Statistics Addition, which achieved a relative improvement of 30-40% on their Position-Adjusted Word Count metric and 15-30% on their Subjective Impression metric. Keyword stuffing, a classic SEO tactic, offered little to no improvement. Two further results are worth knowing. First, the gains were uneven across sources: the Cite Sources method produced a 115.1% increase in visibility for websites ranked fifth in the search results, which suggests that lower-ranked pages have the most to gain from well-sourced content. Second, the effect held on a live product: when tested on Perplexity.ai, Quotation Addition performed best, with a 22% improvement in Position-Adjusted Word Count over the baseline. Note what this study measures: how much of a given source an AI answer uses, under controlled conditions. It does not measure brand mentions in ChatGPT, traffic or revenue, and the authors found that which methods work best varies by domain.

Official documentation is short but useful. Google's guidance on AI features says AI Overviews and AI Mode may use a "query fan-out" technique, issuing multiple related searches across subtopics and data sources to find supporting links. It also states there are no additional requirements and no special schema.org structured data needed to appear in these features. Its recommended practices include making sure structured data matches the visible text on the page and keeping Merchant Center and Business Profile information up to date. On the OpenAI side, OpenAI's crawler documentation explains that OAI-SearchBot is used to surface websites in ChatGPT's search features and recommends allowing it in robots.txt if you want your site to appear there. Finally, the same Pew Research analysis found that Wikipedia, YouTube and Reddit together accounted for 15% of the sources cited in Google's AI summaries, and government websites for 6%, compared with 2% in standard results. Widely referenced, authoritative sources are disproportionately what AI answers cite.

The best-known published case studies involving structured data come from Google's introduction to structured data. Rotten Tomatoes added structured data to 100,000 unique pages and measured a 25% higher click-through rate. Food Network converted 80% of its pages to enable search features and saw a 35% increase in visits. Nestlé measured that pages showing as rich results had an 82% higher click-through rate than non-rich result pages. These are real, company-documented results, but be precise about what they show: better performance in classic Google Search through rich results, not more citations in AI answers. Google itself says no special schema is required for its AI features. The honest takeaway is that structured data helps search engines understand and verify your pages, and keeping it accurate and consistent with visible content is good practice. It is not a guaranteed lever for AI mentions.

This is a recommended approach, not a reported result. A mid-market SaaS company that rarely appears when buyers ask AI assistants for tool recommendations would start by measuring its baseline across ChatGPT, Claude, Gemini and Perplexity for the questions its buyers actually ask. It would then make sure OAI-SearchBot and other search crawlers are not blocked, and publish accurate Organization and SoftwareApplication markup that matches what the pages say. The bigger lever, based on the GEO findings, is content: a cluster of pages around its core use case, each with question-based headings, a direct answer near the top, and claims backed by cited sources, specific figures and attributable quotes. Comparison pages that fairly describe alternatives give AI models extractable material for "best tool for X" questions. Finally, because AI answers lean on widely referenced sources, the company would pursue genuine coverage in industry publications, review platforms and community discussions so independent sources describe it consistently.

This is a recommended approach, not a reported result. A direct-to-consumer brand trying to appear in AI shopping answers would focus first on product data quality, a core part of e-commerce AEO. Google's product structured data documentation notes that product markup can surface price, availability, review ratings and shipping information, and that providing both on-page structured data and a Merchant Center feed maximizes eligibility and helps Google understand and verify the data. The brand would keep price, specifications and availability identical across its own site, marketplaces and feeds, so no AI system finds conflicting facts. It would write buying guides that answer the questions shoppers ask ("which model is best for small apartments?") with clear comparisons and cited specifications, and it would earn independent reviews from reputable publications rather than relying on its own claims.

This is a recommended approach, not a reported result. A multi-location dental practice pursuing a local AEO strategy would start with Google Business Profile. According to Google's explanation of local ranking, local results are mainly based on relevance, distance and prominence; businesses with complete and accurate information are more likely to show up, and more reviews and positive ratings can help local ranking. The practice would complete every profile field for each location, respond to reviews, and ask patients for honest reviews as a routine part of follow-up. On its website, each location page would carry LocalBusiness structured data with address, phone number, opening hours and coordinates matching the profile. It would also publish location-specific answers to common patient questions, such as typical treatment costs and what to expect at a first visit, so AI models have clear, local, extractable content to draw on.

Put the research and documentation together and a consistent picture emerges. Content that cites sources, includes specific figures and quotes experts is more likely to be used in AI answers, according to the GEO experiments, while keyword stuffing does little. AI systems cannot cite what they cannot reach, so crawler access and accurate, consistent business and product data are prerequisites. Structured data is worth doing well because it helps search engines understand and verify your pages, but Google is explicit that it is not a special requirement for AI features. Third-party authority matters: AI answers disproportionately cite widely referenced sources, so on-site work alone is rarely enough. These conclusions line up with the AEO ranking factors covered elsewhere on this blog. Because AI answers vary by engine and by query, measure before and after you make changes rather than assuming a tactic worked. Use the AEO Grader to benchmark your starting position and track your progress.