Research
Zero-click search & AI buying: the evidence
Every figure used in “The click is no longer the conversion”, with its source and caveats. Free to cite with attribution to AI Sandwich (ai-sandwich.com). Also available as JSON.
Last updated 2026-09-24.
Zero-click search
| Figure | Finding | Source |
|---|---|---|
| 68.01% | U.S. Google searches ending without a click, Jan–Apr 2026 | SparkToro / Similarweb (June 2026) |
| 58.5% | U.S. Google searches ending without a click, 2024 | SparkToro / Jumpshot (2019) and SparkToro–Datos (2024) |
| 50.33% | U.S. Google searches ending without a click, 2019 | SparkToro / Jumpshot (2019) and SparkToro–Datos (2024) |
| 32% | U.S. Google searches that still ended in a click, Jan–Apr 2026 | SparkToro / Similarweb (June 2026) |
| 18% | Google searches that produced an AI summary (March 2025 browsing data) | Pew Research Center (July 2025) |
| 8% | Visits where a traditional link was clicked when an AI summary appeared | Pew Research Center (July 2025) |
| 1% | Visits where a link inside the AI summary was clicked | Pew Research Center (July 2025) |
AI answer layer
| Figure | Finding | Source |
|---|---|---|
| >20% | Google searches showing an AI Overview (2026); CTR fell nearly 60% when one appeared | Search Engine Land (9 June 2026) |
| ~2% | AI Mode queries that send a visitor to a website (range ~1.6–2.5%) | Similarweb AI Mode referral analysis (2026) |
| 17–19% | Classic search queries that send a visitor to a website | Similarweb AI Mode referral analysis (2026) |
| 1B+ | Monthly users of Google AI Mode; queries more than doubled each quarter (company-reported) | Google, "A new era for AI Search" (19 May 2026) |
AI-first buying
| Figure | Finding | Source |
|---|---|---|
| 51% | B2B software buyers who begin purchase research in an AI chatbot | G2 (March 2026) |
| 69% | Chose a different vendor than initially planned based on AI chatbot guidance | G2 (March 2026) |
| 1 in 3 | Bought from a vendor they had never heard of before the AI conversation | G2 (March 2026) |
| 83% | AI-assisted buyers reporting more confidence in their final choice | G2 (March 2026) |
| 35% vs 13.6% | Consumers rating AI tools vs search engines as most useful at discovery | Similarweb, 2026 AI Brand Visibility Report (January 2026) |
AI citations
| Figure | Finding | Source |
|---|---|---|
| 57% | Citations across eight major AI models coming from brand-owned websites | Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026) |
| 69% | Google Gemini citations coming from brand-owned websites | Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026) |
| 32.8% | ChatGPT citations going to product pages in July 2026, nearly double March | Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026) |
| ~80% | Drop in Reddit's share of ChatGPT citations in a single week, August 2026 | Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026) |
| 25.6% | Overlap in cited domains between two ChatGPT reasoning modes for the same prompts | Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026) |
Measurement blind spots
| Figure | Finding | Source |
|---|---|---|
| 100% | TikTok, Slack, Discord and WhatsApp share visits reported as 'direct' in GA | SparkToro & Really Good Data (2023) |
| 75% | Facebook Messenger share visits reported as 'direct' | SparkToro & Really Good Data (2023) |
| 30% | Instagram DM share visits reported as 'direct' | SparkToro & Really Good Data (2023) |
Incrementality vs attribution
| Figure | Finding | Source |
|---|---|---|
| 1.53x → 0.70x | Mobile advertising: click-attributed ROAS vs incremental causal ROAS | IEEE Access (March 2026), "From Attribution to Causality in Digital Advertising" |
| ~2.00x → 0.92x | Search engine marketing: click-attributed ROAS vs incremental causal ROAS | IEEE Access (March 2026), "From Attribution to Causality in Digital Advertising" |
Sources
- Pew Research Center (July 2025)
- "Do people click on links in Google AI summaries?" Observational analysis of 900 consenting U.S. adults' browsing behavior in March 2025.
- SparkToro / Similarweb (June 2026)
- "In 2026, Less than One Third of Google Searches Still Send a Click" — Similarweb U.S. desktop + mobile web panel, Jan–Apr 2026. Excludes Google's mobile search app; "zero click" includes reformulated or abandoned searches.
- SparkToro / Jumpshot (2019) and SparkToro–Datos (2024)
- Earlier zero-click panels: 50.33% (2019) and 58.5% (2024).
- Similarweb AI Mode referral analysis (2026)
- Referral on ~1.6–2.5% of AI Mode queries vs 17–19% for classic search.
- Search Engine Land (9 June 2026)
- AI Overview prevalence and click-through context.
- Google, "A new era for AI Search" (19 May 2026)
- Company-reported AI Mode usage figures.
- G2 (March 2026)
- "The Answer Economy: How AI Search Is Rewiring B2B Software Buying" — survey of 1,076 B2B software buyers. Self-reported; directional for other B2B categories.
- Similarweb, 2026 AI Brand Visibility Report (January 2026)
- Consumer panel: AI tools rated most useful at discovery by 35% vs 13.6% for search engines.
- Promptwatch / Petra Labs (Aug 2026); Profound (Jul 2026); Semrush (Sep 2026)
- Reddit citation decline and product-page share; 11.8B cross-model citation analysis (57% brand-owned, 69% for Gemini); 25.6% reasoning-mode overlap. As cited by Amanda Natividad (August 2026).
- SparkToro & Really Good Data (2023)
- Dark social experiment: 1,100+ visits across 11 networks, tracked as 'direct' in Google Analytics. Plus WebKit Tracking Prevention and Google Analytics documentation.
- IEEE Access (March 2026), "From Attribution to Causality in Digital Advertising"
- Dropbox U.S. mobile and SEM market blackout test (April 2025). Corroborated by field experiments at Facebook (Marketing Science, 2019) and eBay (Econometrica, 2015).
- IAB/BWG "State of Data 2024"; DMA "Email Benchmarking Report 2025"
- Nielsen and Litmus guidance on shifting KPIs from platform proxies to outcome influence.
Methodological note: Much of the marketing industry's published data relies on observational panels, vendor-supplied metrics, or selected client cohorts rather than randomized causal trials. Vendor-reported figures should be treated as strong directional evidence for the shift in buyer behavior, not as universal baseline guarantees for all domains, audiences, or industries.