The Ask Gap: Why visitors who talk convert more than visitors who type
Fala Research Technical Report FR-2026-01, v1.0, published 2026-09-15. Authors: Fala Research (Elav8). A quantitative review of voice versus text modality effects on website conversion, with a model for estimating lift. The PDF is available from the report page.
Summary of findings
- In the only like-for-like production comparison found (the same AI agent on live B2B websites, voice mode vs text mode, 13 weeks; Docket 2026), attention was a tie (185 s vs 195 s in conversation; CTA click-through 10.3% vs 10.4%) and conversion was not: voice captured 1.8x the emails and booked 7.4x the meetings.
- The gap appears at the ask and scales with the commitment asked: +22% for orders in a 1.5-million-user voice-shopping pilot (Meesho, 2026), 1.8-2.0x for an email address, 7.4x for a calendar slot. Fala Research calls this the Ask Gap.
- Chat itself works, but less than its headline number: the widely quoted 2.8x compares visitors who chose to chat with those who did not; the selection-corrected causal estimate is +16% (Tan, Wang & Tan, Information Systems Research, 2019).
- Speaking is 2.9x faster than typing on a phone (153 vs 52 words per minute) with a 20% lower error rate (Ruan, Wobbrock, Liou, Ng & Landay, 2018); people type at 36 WPM in the field (Palin et al., 2019).
- Words delivered as narration alongside visuals outperform on-screen text alongside the same visuals, d = 0.72 across 43 effects (Ginns, 2005).
- Hearing the same words makes the speaker seem more competent, thoughtful and intelligent than reading them, including to professional recruiters (Schroeder & Epley, Psychological Science, 2015).
- Humans reply within 0-200 ms in every language tested (Stivers et al., PNAS, 2009); production voice agents reached 500-900 ms in 2026.
- Modelled lift (Monte Carlo over stated priors): about 1.2x for purchases, about 1.8x for lead capture (80% interval 1.25-2.45x), and a wide, low-confidence estimate for booked meetings. Voice is cheaper per conversion once its lift exceeds its per-interaction cost ratio.
- Boundary conditions: 79-85% of consumers prefer a human to an AI agent and 78% have tried to bypass one; that is a preference for humans over AI, not for typing over talking, and it sets the design constraints: tap-to-talk, text one tap away, disclosed AI, human handoff, voice in and visual out, grounded answers, sub-second latency.
Definition
The Ask Gap is the divergence between voice and text conversion that appears only at the moment of commitment - not in whether a visitor engages or how long they stay, but in whether they say yes when asked.
How to cite
Fala Research (2026). The Ask Gap: Why Visitors Who Talk Convert More Than Visitors Who Type. Technical Report FR-2026-01, v1.0. Elav8. https://fala.elav8.ai/research/voice-vs-chat
Related: Fala Research, Features, Pricing.