An analysis of 1,000 queries across ChatGPT, Perplexity, Google AI Overviews, and Claude — tracking how brands appear, disappear, and compete in AI-generated search results.
Between November 2025 and April 2026, ZapTap Labs tracked 1,000 queries across four major LLM platforms to measure how brands gain and lose visibility in AI-generated search results. The study reveals a rapidly maturing AI search ecosystem where traditional SEO signals are necessary but insufficient for visibility.
Of AI search visibility is uncaptured by the average brand
We tracked 1,000 queries across four LLM platforms — ChatGPT (GPT-4o), Perplexity Pro, Google AI Overviews, and Claude (Opus) — weekly for six months from November 2025 through April 2026. Queries were selected to represent high-intent commercial and informational searches across 12 industries: B2B SaaS, healthcare, financial services, e-commerce, legal, real estate, education, manufacturing, professional services, cybersecurity, HR tech, and marketing technology.
Each query was run three times per platform per week to account for response variability. Responses were scored on three dimensions: brand citation presence (binary — was the brand named?), recommendation strength (1-5 scale measuring how favorably the brand was positioned), and entity accuracy (whether factual claims about the brand were correct).
We excluded navigational queries (branded searches) and focused exclusively on non-branded queries where multiple brands could legitimately appear. The 1,000 queries were split evenly between informational intent (500) and commercial/transactional intent (500).
Across the 500 commercial-intent queries in our dataset, we found that 34% of high-intent B2B queries now start on an LLM platform rather than a traditional search engine. This represents a 79% increase from our November 2024 baseline measurement of 19%.
The distribution is not uniform across industries. Cybersecurity and HR tech show the highest LLM query origination rates at 47% and 43% respectively, driven by technical buyers who use LLMs as research assistants. Real estate and manufacturing lag at 18% and 21%, where local and visual search still dominate the discovery process.
Platform share among LLM queries breaks down as follows: ChatGPT captures 41% of AI-originated queries, Perplexity takes 28%, Google AI Overviews account for 22%, and Claude holds 9%. Perplexity showed the fastest growth rate over the study period, nearly doubling its share from 15% in November 2025.
Of high-intent B2B queries now originate outside Google — up from 19% in late 2024
LLM citation rate for brands with entity authority scores above 0.50
We calculated entity authority scores for 340 brands that appeared across our query set using ZapTap's proprietary scoring methodology (0.0–1.0 scale measuring cross-platform entity consistency, knowledge graph presence, and topical association strength).
Brands with entity authority scores above 0.50 appeared in 78% of relevant LLM queries. Brands scoring between 0.30 and 0.50 appeared in 34% of relevant queries. Brands below 0.30 appeared in only 12%. The correlation between entity authority and citation frequency was 0.81 — stronger than any other single variable we measured, including domain authority (0.52 correlation), backlink count (0.47), or content volume (0.39).
This finding has significant implications for SEO strategy. Domain authority and backlink profiles — the traditional currency of search optimization — are weaker predictors of LLM visibility than entity consistency and structured data coverage. Brands investing in entity architecture are building a durable competitive advantage that compounds across platforms.
We analyzed the publication and last-modified dates of every page cited by an LLM across our 312,000 data points. Articles updated within 90 days of the query date were 3.4x more likely to be cited than articles last updated more than 90 days prior.
For comparison, we ran the same analysis against Google organic rankings for the same queries. In Google results, freshness showed a 1.6x advantage for recently updated content — meaningful but less than half the impact we observed in LLM responses. This suggests that LLMs weight recency signals more heavily than Google's core algorithm, likely because LLM providers are actively working to reduce the "stale training data" criticism their platforms have faced.
The freshness effect was strongest on Perplexity (4.1x advantage for content under 90 days old), followed by Google AI Overviews (3.8x), ChatGPT (3.1x), and Claude (2.6x). Perplexity's real-time web access likely accounts for its heightened freshness sensitivity.
More likely to be cited when content is updated within 90 days
Higher citation rate for sites with 80%+ structured data coverage
We audited the structured data implementation of the top 200 most-cited domains in our dataset and compared them to a control group of 200 domains that rarely appeared in LLM responses despite ranking well in Google organic results.
Sites with comprehensive schema markup (Organization, Product, Service, FAQ, Article, and HowTo schema on 80% or more of relevant pages) were cited 2.7x more frequently than sites with minimal or no structured data. The effect was most pronounced for Product and FAQ schema, which showed 3.2x and 2.9x citation advantages respectively.
Our hypothesis is that structured data serves as a high-fidelity signal for LLMs during retrieval-augmented generation. When an LLM's retrieval system fetches pages to ground its response, structured data provides machine-readable entity relationships and factual claims that reduce the risk of hallucination — making those pages more likely to be selected as citation sources.
For each of the 340 brands in our analysis, we identified the 30 most relevant non-branded queries where a brand in their category should logically appear. We then measured how many of those 30 queries each brand actually appeared in across all four LLM platforms.
The average brand appeared in just 4 of 30 relevant queries — a 13% capture rate. The median was even lower at 3 queries. The top-performing brand in our dataset appeared in 27 of 30 queries. The bottom quartile appeared in 0-1 queries despite having established Google organic presence.
This represents a massive untapped opportunity. For most brands, 87% of the queries where they should be recommended by LLMs are queries where they are completely absent. The gap between AI search leaders and laggards is wider than anything we have observed in traditional search, where even poorly optimized sites typically appear for a meaningful percentage of branded and long-tail queries.
Average number of relevant LLM queries where a brand appears
Five evidence-based actions based on what the data tells us.
Entity authority was the single strongest predictor of LLM citation rates in our study — stronger than domain authority, backlinks, or content volume. Before publishing more content, ensure your brand entity is consistently defined across knowledge graphs, structured data, authoritative directories, and your own website. Fix entity inconsistencies first.
The 3.4x freshness advantage in LLM citations means content refresh is no longer optional. Implement a systematic refresh cadence: update high-value pages at least quarterly, add new data points, and ensure last-modified dates reflect genuine substantive updates rather than cosmetic changes.
The 2.7x citation advantage for sites with strong schema coverage is actionable today. Prioritize Organization, Product, FAQ, and Article schema. Aim for 80%+ coverage across all indexable pages. Structured data is the bridge between your content and the machines deciding whether to cite it.
If you are only tracking Google Search Console, you are blind to 34% of high-intent queries. Establish baseline LLM visibility by running 20-30 relevant queries across ChatGPT, Perplexity, and Claude. Track monthly. Correlate changes with your content and authority-building actions.
LLMs cite content that contains original data, specific claims, and structured arguments. Blog posts written solely to match keyword intent will rank on Google but get ignored by LLMs. Include proprietary statistics, named methodologies, and expert attribution in every piece of content you want cited.
The complete State of AI Search Visibility 2026 report includes industry-by-industry breakdowns, platform-specific citation patterns, the full list of entity authority variables, and quarter-by-quarter trend data. Available as a downloadable PDF.
Usman is the CEO and founder of ZapTap. He leads ZapTap Labs, the agency's research division, and personally oversees the methodology and analysis for every published study. Before founding ZapTap, Usman spent eight years in enterprise SEO and data analytics, working with brands across SaaS, healthcare, and financial services.
His research has been cited in Search Engine Journal, Moz, and Ahrefs. He speaks regularly at SMX, BrightonSEO, and MozCon on the intersection of AI and search visibility.
CEO & Founder, ZapTap
Head of ZapTap Labs
8+ years in enterprise SEO
Speaker: SMX, BrightonSEO, MozCon