Conversational commerce numbers that survive checking

An earlier version of this post promised that every number in it was cited. Every number was cited. Several of the citations pointed at pages that no longer contain the figure attributed to them, and one pointed at a forecast whose method is not published anywhere.

So this version is the audit instead. We opened every source, kept what states its sample and its period, and wrote down why the rest did not make it. The result is shorter than a statistics roundup usually is, which is the honest outcome. If you are here to size a build rather than a market, what the whole thing is made of is the more useful page.

What held up#

Adoption of AI assistants generally#

Pew Research Center surveyed 5,119 US adults between 17 and 23 February 2026 and found that about half now use AI chatbots, up from a third in 2024, with about a quarter using them daily. Around 42 percent use them for information searching, and 35 percent of Americans say they have a smart speaker (Pew Research Center, 17 June 2026).

This is the strongest number in this post, because it is a probability sample with published field dates and a stated question. It is also about assistants in general rather than shopping, and it should not be stretched into a commerce claim.

Globally, Kepios reports 6.12 billion internet users at the start of April 2026, or 73.8 percent of the world, and generative AI platforms reaching 2.42 billion monthly active users, close to 30 percent of the world’s population, after adding 1.4 billion users in twelve months (DataReportal, April 2026).

AI-referred shopping traffic, where two panels disagree#

Adobe Analytics reported that AI-referral traffic to US retail sites rose 62 percent year over year in July 2026, that those visitors generated 53 percent more revenue per visit than other shoppers, and that AI-sourced visits converted at a rate 60 percent higher than non-AI traffic. Its basis is over one trillion visits to US retail sites, drawn from more than 200 of the top 2,000 online retailers (reported 19 August 2026).

Contentsquare, analysing 99 billion sessions across more than 6,000 sites from Q4 2024 to Q4 2025, put AI-referred traffic at a 1.3 percent conversion rate, up 55 percent year over year, against 2.8 percent for paid search (2026 Digital Experience Benchmarks, 9 March 2026).

Those are not the same claim and they are not contradictory. Adobe compares AI-referred traffic against everything else on retail sites and finds it better. Contentsquare compares it against paid search across a broader site mix and finds it worse. Both are vendor panels of self-selected customers rather than random samples of the web. Quoting either one alone is how a real finding becomes a marketing line.

Friction, which is the durable part#

Baymard’s cart abandonment figure of 70.22 percent is quoted constantly and almost always wrongly. Their own page is explicit: “This value is an average calculated based on 50 different studies containing statistics on ecommerce shopping cart abandonment”, spanning 2006 to 2025, last updated 22 September 2025 (Baymard). It is a nineteen-year meta-average and it is not a measurement of anything happening now.

The more actionable Baymard figure is that checkout flows carried an average of 11.3 form fields in 2024 against the 8 a checkout needs, with 17 percent of users reporting abandonment because the process was too long or complicated (Baymard, 26 June 2024). Field count is a thing you can change this quarter, which is more than the abandonment average offers, and it is the lever behind getting a first-time user to something real.

On the support side, Gartner surveyed 5,728 customers and found only 14 percent of service issues fully resolved in a company’s self-service channel, with 36 percent for issues customers described as “very simple” (reported 19 August 2024). A later survey of 265 executives found “nearly 9 in 10 customer service journeys beginning in self-service are ultimately resolved through multiple channels” (CX Dive, 3 September 2025).

Where adoption is fastest#

The Stanford AI Index reports that in 2025, 58 percent of employees globally used AI at work on a semiregular or regular basis, and that in India, China, Nigeria, the United Arab Emirates, Egypt and Saudi Arabia the share exceeded 80 percent (2026 AI Index Report). That is a workplace figure rather than a consumer one, and the geography in it is the useful part, which where in-app agents are being adopted fastest takes further.

What did not survive#

Half of all searches would be voice by 2020. The archetype of the genre. It was not published by comScore, to whom it is almost universally attributed. It was a 2014 prediction by Andrew Ng, then at Baidu, that within five years at least half of searches would be “either through images or speech”, made about Baidu rather than global search (Brodie Clark, updated 19 September 2024). Voice-only, global, and 2020 were all added later by people repeating it.

A 72.8 billion dollar voice commerce market growing at 19.90 percent to 2040. The figures are on the page. The methodology is not: the public report page states no sample, no market definition, and no data sources behind the forecast (Roots Analysis). A forecast to 2040 with no stated definition of what is being counted cannot be checked, argued with, or used to size anything.

8.4 billion voice assistants in use worldwide. We opened the aggregator page the earlier version cited. The figure is not there now, and the page attributes no individual statistic to any individual source, listing four general references at the foot instead. Separately, a count of assistants installed is a count of capability rather than of use.

A 12 to 23 percent conversion lift from conversational AI. Also gone from the page it was cited to, which is published by a company selling conversational commerce software and which attributes its remaining figures to other aggregators without describing a study design.

Voice carts abandoned at 42 percent against the 70 percent average. This one is a comparison error rather than a missing source. The 42 percent describes one vendor’s client base; the 70.22 percent is an average of 50 studies over nineteen years. Two different populations measured by different methods over different periods do not form a comparison.

None of the five is necessarily false. Each one is unverifiable from the page it is cited to, which for a working purpose is the same thing.

How to check one yourself#

The whole audit is four steps and it takes about ten minutes per figure.

Open the actual page, not the article citing it. About half the time the number is not there.

Find the population. Who was measured, how many of them, over what period. A figure without a population is a slogan.

Find who paid. Vendor panels are useful and biased in a predictable direction, so quote them as what they are: “Adobe, measuring its own customers’ retail sites, found …”. That framing costs six words and makes the sentence defensible.

Check the arithmetic against a second source with a different method. Where two independent panels disagree, as Adobe and Contentsquare do here, report both. The disagreement is more informative than either figure.

What nobody has measured#

There is no credible public figure for the share of in-app requests an agent completes rather than a screen, for how often a spoken request works twice in a row in production, for how many users switch input mode mid-task, or for whether an agent moves retention after novelty decays.

That absence is not a gap in the literature waiting to be filled. Those numbers depend on your task set, your backend and your users, and a cross-industry average of them would not be actionable even if someone published one. Each is measurable from your own traffic within a few weeks, which is what the metric set worth keeping for an in-app agent is for, and what reading the first weeks of real usage produces.

The one honest thing a statistics post can do for a build decision is set the direction and then get out of the way. The direction here is clear enough: AI-mediated shopping traffic is growing quickly from a small base, form friction has not improved in twenty years, and self-service resolution is far worse than the industry claims. What none of it tells you is whether an agent belongs in your product, which is a question with real answers on both sides.

Common questions#

How big is the voice commerce market? No published figure we could verify answers this. The widely quoted forecast of 72.8 billion dollars growing at 19.90 percent to 2040 sits on a page that states no market definition, sample or data source, which makes it unusable as evidence.

Do AI-referred shoppers convert better? Two large panels disagree. Adobe, on over a trillion US retail visits, reports AI-sourced traffic converting 60 percent above non-AI traffic. Contentsquare, on 99 billion sessions across 6,000 sites, reports it at 1.3 percent against 2.8 percent for paid search. Both are vendor panels, and the honest answer quotes them together.

Is cart abandonment really 70 percent? That figure is Baymard’s average across 50 studies published between 2006 and 2025. It is a meta-average, not a current measurement, and Baymard says so on the page.

Why did the “50 percent of searches will be voice” prediction not happen? Because it was never the prediction. Andrew Ng said in 2014 that within five years half of Baidu searches would be through images or speech. The voice-only, global, dated-2020 version is an accumulation of misquotations.

What should I measure instead of quoting these? Task completion per task against your own backend, reliability as repeated clean runs, mode switching, and repeat use at seven and twenty-eight days. Those decide product questions, and no external statistic can.

Sources#

• Pew Research Center, Americans and AI 2026: Chatbots, Smart Devices and Views on Impact, 17 June 2026, n=5,119. Accessed 12 September 2026.
• DataReportal, Digital 2026 global overview, April 2026. Accessed 12 September 2026.
• Digital Commerce 360, Adobe: AI-referral traffic spending, converting more than counterparts, 19 August 2026. Accessed 12 September 2026.
• Contentsquare, 2026 Digital Experience Benchmarks: conversions, 9 March 2026. Accessed 12 September 2026.
• Baymard Institute, Cart abandonment rate statistics, updated 22 September 2025. Accessed 12 September 2026.
• Baymard Institute, Checkout flows average 11.3 form fields, 26 June 2024. Accessed 12 September 2026.
• CX Today, Only 1 in 7 customer service queries resolved with self-service, 19 August 2024, n=5,728. Accessed 12 September 2026.
• CX Dive, CX leaders say self-service and live chat will overtake phone and email, 3 September 2025, n=265. Accessed 12 September 2026.
• Stanford HAI, 2026 AI Index Report: Public Opinion. Accessed 12 September 2026.
• Brodie Clark, Stop using Comscore’s 2020 voice search statistic, updated 19 September 2024. Accessed 12 September 2026.
• Roots Analysis, Voice commerce market. Accessed 12 September 2026, no methodology published on the page.