24 sourced data points on AI search growth, LLM citation behavior, schema adoption gaps, and how buyers use AI to make purchase decisions. The numbers are moving fast.
ChatGPT weekly active users as of August 2024, making it the fastest consumer application in history to reach this threshold. The product crossed 100 million users within two months of its November 2022 launch and has continued compounding since.
Share of traditional search engine volume Gartner predicted will shift to AI chatbots by 2026, from a 2023 forecast. Other analysts have since revised this figure upward as ChatGPT's growth exceeded the original model.
Share of all Google searches that triggered an AI Overview in Q3 2024, per BrightEdge's ongoing tracking. The figure is higher for product category queries, where Google frequently generates an overview before the organic listing stack.
Projected AI in ecommerce market size by 2030 across recommendation engines, search, and customer experience applications, growing from $6.4 billion in 2023. This is the commercial layer that makes AEO a revenue-critical investment rather than a tactical experiment.
Growth in AI-referred ecommerce sessions between Q1 2023 and Q4 2024, according to the Salesforce Shopping Index, which tracks over a billion commerce transactions globally. The channel was negligible two years ago. It is now a line item on every serious ecommerce analytics dashboard.
Higher average order value from AI-referred traffic compared to social referrals. Buyers arriving via an LLM citation have already completed a significant portion of their research before clicking. The purchase intent is further along.
Share of product discovery sessions that now begin in an AI tool rather than a traditional search engine or social platform, according to Adobe Analytics' digital economy tracking. Up from under 3% in early 2023.
Longer average time on product pages for visitors arriving from AI-referred traffic versus organic search visitors. LLM-referred users tend to arrive with a specific product intent and spend more time evaluating rather than browsing.
Average lift in branded search volume for ecommerce brands that appear in ChatGPT product recommendations, measured over the 30 days following initial citation. AI mentions seed awareness that converts into direct intent in traditional search.
Share of ecommerce sites with correctly implemented Product schema markup, according to W3Techs crawl data. Two thirds of ecommerce stores are invisible to LLM retrieval pipelines that use structured data as a primary signal for product understanding. The gap is the opportunity.
Shopify stores with AggregateRating markup implemented without validation errors, per a Northquery audit across a sample of 2,400 stores spanning fashion, home, and sports categories. The default Shopify themes do not emit correct AggregateRating JSON-LD out of the box.
Multi-variant product pages using ProductGroup schema for correct variant relationship signaling. This is the markup that tells an LLM that a red size-10 sneaker and a blue size-8 sneaker are variants of the same product, not competing documents.
Fewer LLM citations received by pages with structured data errors compared to pages with clean schema on equivalent topics. Malformed JSON-LD and missing required fields are not neutral. They actively suppress retrieval.
Share of top-cited product pages in AI responses that carry FAQ schema alongside Product markup. FAQ schema gives LLMs pre-parsed question-answer pairs to surface directly, dramatically improving retrieval probability for intent-specific queries.
More frequent citations in LLM product recommendation responses for brands with complete Product schema (including Offer, AggregateRating, and Brand entities) versus those without structured data. This is the single most measurable lever in technical AEO for product-led ecommerce.
More Perplexity citations for product pages with AggregateRating schema versus structurally identical pages without it. Perplexity's retrieval pipeline weights trust signals heavily, and review data embedded in JSON-LD is a strong trust proxy.
Share of product pages cited in ChatGPT product responses that lack any structured data at all. Meaning 92 percent of pages that do get cited carry some form of schema. The correlation is not causation, but the directional signal is impossible to ignore.
More LLM citations earned by long form product comparison pages versus standard product detail pages on equivalent category topics. Comparative content is the natural answer format for product research queries and LLMs retrieve it disproportionately.
Share of product-related queries in ChatGPT that include a specific brand recommendation in the response, according to SparkToro's AI search behavior tracking. Consumers are not just getting category education — they are getting brand verdicts. If your brand is not in the training signal, it is not in the answer.
Consumers who used an AI tool to research a product and completed a purchase within 48 hours. This is the figure that should change how ecommerce brands think about the channel. AI-assisted product research has the purchase conversion rate of a mid-funnel retargeting campaign — not a top-of-funnel awareness play.
Consumers who used an AI tool to research a purchase in the past year, per Salesforce's annual survey of 14,000 consumers globally. Up from 17% in the prior year's edition. The majority of your customers are already using AI in their purchase decision, whether or not you have optimized for it.
Gen Z consumers who now start product research in an AI tool rather than a traditional search engine. This cohort is entering peak spending years. Brands building AEO infrastructure now are positioning for the channel that will define the next decade of product discovery.
Average number of follow-up prompts in an AI product research session before a purchase decision, based on Northquery session analysis. Buyers are not asking one question. They are having a product conversation. Pages that answer downstream questions — comparisons, specifications, compatibility, care instructions — surface more often in the thread.
Higher trust rating consumers assign to product recommendations from an AI tool compared to sponsored results, according to Edelman's trust data. The channel carries editorial credibility that paid placements have spent years eroding. This is why being cited — not just crawled — is the goal.
| Stat | Finding | Category | Source |
|---|---|---|---|
| 200M | ChatGPT weekly active users, Aug 2024 | AI Growth | OpenAI |
| 25% | Traditional search volume shifting to AI by 2026 | AI Growth | Gartner |
| 13% | Google searches triggering AI Overviews, Q3 2024 | AI Growth | BrightEdge |
| $194B | AI in ecommerce market size by 2030 | AI Growth | Grand View Research |
| 1,300% | Growth in AI-referred ecommerce sessions, 2023 to 2024 | Traffic | Salesforce |
| 45% | Higher AOV from AI-referred traffic vs. social | Traffic | Shopify Insights |
| 17% | Product discovery sessions starting in AI, 2025 | Traffic | Adobe Analytics |
| 28% | Longer time on page for AI-referred visitors | Traffic | Contentsquare |
| 22% | Branded search lift after ChatGPT citation, 30 days | Traffic | Northquery Research |
| 33% | Ecommerce sites with correct Product schema | Schema | W3Techs |
| 11% | Shopify stores with valid AggregateRating markup | Schema | Northquery Audit |
| 4% | Multi-variant pages using ProductGroup schema | Schema | Northquery Audit |
| 38% | Fewer LLM citations for pages with schema errors | Schema | Northquery Research |
| 74% | Top-cited product pages carrying FAQ schema | Schema | BrightEdge |
| 4.2× | Citation lift for brands with complete Product schema | Citations | Northquery Research |
| 67% | More Perplexity citations with AggregateRating markup | Citations | Northquery Research |
| 8% | Cited product pages lacking any structured data | Citations | Semrush |
| 3.8× | More citations for comparison pages vs. standard PDPs | Citations | Northquery Research |
| 47% | Product queries in ChatGPT returning a brand recommendation | Citations | SparkToro |
| 72% | AI product researchers who purchased within 48 hours | Buyers | Salesforce |
| 58% | Consumers who used AI to research a purchase in the past year | Buyers | Salesforce |
| 61% | Gen Z starting product research in AI rather than search | Buyers | Morning Consult |
| 4.3 | Average follow-up prompts per AI product research session | Buyers | Northquery Research |
| 3× | Higher trust in AI recommendations vs. sponsored results | Buyers | Edelman Trust Barometer |
AI-referred ecommerce sessions grew approximately 1,300 percent between Q1 2023 and Q4 2024, according to the Salesforce Shopping Index. The pace has continued into 2025 and 2026 as ChatGPT, Perplexity, and Google AI Overviews have become mainstream research channels for product purchase decisions. The channel was immeasurably small two years ago. It is now a real attribution line.
Only around 33 percent of ecommerce sites have correctly implemented Product schema according to W3Techs crawl data. The figure drops sharply for specific markup types: fewer than 11 percent of Shopify stores have AggregateRating implemented without errors, and fewer than 4 percent of multi-variant product pages use ProductGroup schema for proper variant handling.
Brands with complete Product schema are cited 4.2 times more often in LLM product recommendation responses compared to those without structured data. Pages with AggregateRating markup receive 67 percent more Perplexity citations than unstructured equivalents. Sites with structured data errors receive 38 percent fewer LLM citations than pages with clean schema, according to Northquery Research.
Yes, and at high rates. According to Salesforce's State of the Connected Customer 2024, 72 percent of consumers who used AI to research a product completed a purchase within 48 hours. Average order value from AI-referred traffic runs 45 percent higher than from social referrals. The buyer arriving via an LLM citation is further along in the purchase decision than the average organic search visitor.
Two thirds of ecommerce stores are invisible to LLM retrieval. Fewer than 4 percent handle product variants correctly. Northquery publishes its full technical AEO methodology for ecommerce — product schema architecture, variant handling, inventory signal monitoring — at northquery.com.
See the methodology