The AI customer journey, and why it does not end at the answer
54% of UK adults now use AI tools and 24% have tried a new local business because an assistant recommended it. But only 2% buy without checking first. AI is a shortlisting layer, not a closing channel.
There are two stories told about AI and buying. One says search is dead and AI answers are the new front door. The other says AI referral traffic is a rounding error and nothing has changed. The measured data supports neither, and the shape it does support changes what is worth doing about it.
What has actually changed
| Measure | Value | Source |
|---|---|---|
| UK adults using AI tools | 54%, up from 31% the previous year | Ofcom, 7,533 UK adults |
| UK 16–24s / 25–34s using AI tools | 79% / 74% | Ofcom |
| UK online adults reading AI search summaries at least sometimes | 75% | Ofcom |
| UK consumers who used AI for local search in the past month | 36.7% | Yext, 600 UK consumers |
| UK consumers who tried a new local business on an AI recommendation | 24% | Yext |
Alongside that, the click is disappearing from ordinary search. The share of Google searches ending with no click at all reached 68% in early 2026, up from 60% in 2024 and 49% in 2019.
Pew Research measured roughly 70,000 searches by 900 US adults. When an AI summary was present, 8% of searches produced a click to a website, against 15% when it was absent. Just 1% of users clicked a source inside the AI summary. Session abandonment rose from 16% to 26%.
That Pew study is the highest-grade evidence in this area: an independent, non-commercial institution measuring real browsing behaviour with no product to sell. Nearly everything else in this field comes from companies selling tools.
The part almost everyone gets wrong
If you stop reading at the numbers above, the obvious conclusion is that AI answers are now the destination and the job is to be the answer. The verification data says otherwise, and it is the single most commercially important finding here.
After receiving an AI recommendation, UK consumers report that they:
- 49% search Google or Bing to verify it
- 43% go directly to the business's own website
- 37% check the sources the AI cited
Separately, 86% of US consumers who used AI for product research verified through another source before buying — and only 2% bought from an AI-recommended brand without checking it first.
AI recommendation is currently an influence and shortlisting layer, not a closing channel. Being named gets you into the consideration set. Your website, your reviews and your Google profile still have to survive the verification step that follows.
This is a meaningfully different proposition from 'get AI traffic', and it has a practical consequence. Winning a mention while having a thin website, no reviews and an unclaimed Google profile converts the mention into a lost customer at the verification step. The two halves are not separate projects.
Why the referral traffic numbers look so small
AI referral traffic is genuinely tiny as a share of visits, and honest providers should say so. Two things explain why the number understates the shift.
First, the zero-click effect means the influence happens without a visit by design. A person who reads an answer, then searches your name directly, arrives as branded search traffic rather than as an AI referral. The channel that gets the credit is not the channel that did the work.
Second, referrals from native apps often carry no referrer header at all, so a meaningful share of AI-driven visits are invisible in analytics by construction.
| Platform | Pages crawled per human referred back | Date |
|---|---|---|
| 5:1 | Jun 2026 | |
| Perplexity | 186:1 | Jun 2026 |
| OpenAI | 848:1 | Jun 2026 |
| Anthropic | 1,917:1, improved from 70,900:1 a year earlier | Jul 2026 |
Cloudflare's crawl-to-refer ratios show AI platforms consuming content far faster than they return visitors, though the gap is closing sharply. Cloudflare's own caveat applies: missing referrer headers overstate the gap by an unknown amount.
The trust picture is genuinely mixed
Two findings sit awkwardly together and both are worth knowing. Among UK AI users, 74% rate their trust in AI local recommendations at four or five out of five, and 74% say their trust has grown against 4% saying it declined. At the same time, 57% of UK adults say they trust AI-generated content less than human-written content.
The resolution is probably that people trust AI to shortlist and not to decide, which is exactly what the verification behaviour shows them doing. Age splits it sharply too: trust in AI recommendations runs at 53% among 18 to 29s against 15% among the over-61s.
What to actually do with this
- Measure whether you are in the consideration set, not whether you rank. Being named at all is the thing that matters at this stage of the market.
- Assume the verification step will happen. Anyone who hears about you from an assistant will check you, and the check happens on your website, your reviews and your Google profile.
- Do not judge this by AI referral traffic in analytics. The mechanism is designed not to produce a click, and native app referrals often carry no referrer at all.
- Weight your effort by audience age. If your customers are predominantly over 60, this is a smaller and slower shift for you than the headline numbers suggest.
Common questions
- Do people buy from AI recommendations?
- Rarely without checking first. Only about 2% of consumers bought from an AI-recommended brand without verifying it elsewhere. After a recommendation, roughly half search Google or Bing to verify, and 43% go directly to the business's own website.
- Is AI traffic worth chasing if the referral numbers are tiny?
- The referral number understates the effect by design. Most AI-influenced visits arrive as branded search or direct traffic because the person verifies you separately, and many native-app referrals carry no referrer header at all, so they are invisible in analytics.
- How many UK adults actually use AI tools?
- 54% of UK adults, up from 31% the previous year, according to Ofcom's survey of 7,533 adults. Among 16 to 24 year olds it is 79%. Among the over-65s it is closer to 18%, so the size of this shift depends heavily on who your customers are.
Sources
Every claim above should be checkable. Where a study has limits, they are stated rather than left out.
- Pew Research Center — 900 US adults, approximately 70,000 searches, March 2025Grade A. Independent, non-commercial, measured browsing behaviour.
- Ofcom — Adults' Media Use and Attitudes, 7,533 UK adults aged 16+, published April 2026Grade A. Official UK regulator research.
- SparkToro and Similarweb — zero-click search panel, January to April 2026Grade A.
- Yext — Consumer Search Behaviours, 600 UK consumers, June 2026Grade B. Vendor-published consumer survey.
- Cloudflare — crawl-to-refer ratios by AI platformGrade A, with Cloudflare's own caveat on missing referrer headers.
- Ahrefs — traffic share tracker across 75,000 domains, June 2025 to May 2026Grade A.
Want to know where you stand?
We ask five AI assistants for a company like yours and send you a free report showing how often you came up and who came up instead.
Get my free report