Gener8's July 2026 panel data (US, mobile) shows the prompt-to-visit rate rising 9x to 10.3x between 1 hour and 72 hours after the prompt. A same-session click report looks at the wrong window. Here is what that changes for how consumer brands measure AI discovery, and what to do about it.
AI Discovery Is a Delayed Channel: Most Visits From an AI Prompt Land Days Later, Not Minutes
By Stephen Honight, Founder of Lmo7
A brand opens its analytics, filters to AI referrals, sees a thin line and concludes that AI sends it nothing. It is a conclusion we hear from clients, and I think probably it is wrong. The line is not hiding anything. The report is looking at the wrong window.
Gener8’s July 2026 AI Landscape report is the best panel data I have seen on this. In its US mobile panel, the prompt-to-visit rate multiplies 9x to 10.3x between 1 hour and 72 hours after the prompt (Gener8, AI Landscape July 2026, US mobile panel). Gener8’s reading is that most vertical-related site visits take days, not minutes, to materialise.
A same-session report sees the 1-hour end of that range. A good share of the visits sits at the other end.
The short version
- What the data says. In Gener8’s US mobile panel, the share of AI prompts followed by a visit to a site in the same vertical rises 9x to 10.3x between the 1-hour and 72-hour marks.
- What it changes. A report that counts only same-session clicks from AI platforms measures the early end of a lag. “AI isn’t sending us traffic” is usually a statement about the window.
- Where the checklist is. The “What to do” section near the end, followed by the Lmo7 routes for measuring this properly.
What Gener8 measured, and how
Gener8 runs a GenAI and clickstream panel. Participants consent to having their activity metered, and Gener8 weights the panel to census by gender, age (18 to 64) and region each month. The details that matter for reading the numbers:
- Volume. 100m+ prompts and responses collected, with 15m+ new ones each month, across 7+ LLMs. The ones named are ChatGPT, Gemini, Google AI Mode, Claude and Perplexity.
- Markets and dates. The US and UK, August 2025 to June 2026.
- Devices. Mobile-first collection: Android in-app and mobile web. iOS in-app tracking is excluded because of Apple policy.
- What is left out. Google AI Overviews are deliberately excluded. Gener8’s reason is that they occur inside standard Google search and would distort the AI adoption metrics.
The prompt-to-visit method works like this. Gener8 defined four verticals upfront: Apparel & Fashion, Consumer Electronics, Home Improvement and B2B SaaS. It tagged each prompt to a vertical using the full conversation as context. A visit by the same user to a site in the same vertical within a set window counted as a linked visit, assigned across the top 20 to 40 sites in each vertical.
Now the scope, stated plainly. The prompt-to-visit figures in this post are US and mobile. The adoption and search figures further down cover the US and UK. Nothing here covers desktop or iOS in-app sessions. If your customers buy mostly on desktop, your own numbers may sit differently, and Gener8’s data cannot tell you which way.
A linked visit is also a match on user, vertical and timing. It does not prove the prompt caused the visit. I read these figures as strong correlation, not attribution.
The numbers
Here is the prompt-to-visit rate at three time marks after the prompt, by vertical.
| Vertical | 1 hour | 24 hours | 72 hours |
|---|---|---|---|
| Apparel & Fashion | 0.7% | 3.5% | 7.2% |
| Consumer Electronics | 2.2% | 10.2% | 19.9% |
| Home Improvement | 1.1% | 5.0% | 10.0% |
| B2B SaaS | 2.7% | 14.1% | 25.6% |
Source: Gener8, AI Landscape July 2026, prompt-to-visit rate by vertical, US mobile panel, report window August 2025 to June 2026.
Two patterns from Gener8’s own read. B2B SaaS leads at every time mark. Apparel & Fashion is lowest throughout and is still under 10% at 72 hours. Gener8 notes that B2B SaaS buyer intent looks research-led rather than typical shopping behaviour, which fits a longer, more considered path.
For a consumer brand the useful row is your own category. Consumer Electronics goes from 2.2% at 1 hour to 19.9% at 72 hours (Gener8, US mobile panel). A report anchored to the first column would record the smallest figure in that row and call it the channel.
The mechanism
Gener8’s data is about timing. The route a shopper takes back to a brand is my inference from how attribution works, so read this section as my view.
A shopper asks a model which headphones to buy, or what to wear to a wedding. The model gives a shortlist. The shopper does not click through. They sit on it and compare prices. Payday is a week away. A day or three later they search for one of the names on the shortlist, or type the URL, or open the app.
Your analytics sees a branded search or a direct visit. Last-click gives the credit to branded search or direct. The prompt that built the shortlist appears nowhere in the path.
So the AI influence is real and invisible in a same-session report. Gener8 can see both ends because it meters the same user across the prompt and the later visit. Your analytics cannot.
This is why “AI isn’t sending us traffic” is usually a measurement-window statement, not a demand statement.
One honest limit. Gener8’s marks stop at 72 hours, and I have not seen data beyond that, so I do not know how far the curve runs. For a considered purchase my guess is weeks. Treat that as my judgement to test, not a Gener8 finding.
Three more layers from the same report
All of this is Gener8’s data, with its scope attached to each figure.
Which platform: Google AI Mode, then ChatGPT, then Gemini
Here is the 24-hour prompt-to-visit rate by platform.
| Vertical | Google AI Mode | ChatGPT | Gemini |
|---|---|---|---|
| Apparel & Fashion | 5.0% | 3.9% | 2.7% |
| Consumer Electronics | 13.2% | 10.7% | 8.6% |
| Home Improvement | 7.5% | 5.9% | 3.6% |
| B2B SaaS | 20.4% | 16.2% | 11.0% |
Source: Gener8, AI Landscape July 2026, 24-hour prompt-to-visit rate by platform, US mobile panel. The chart covers Google AI Mode, ChatGPT and Gemini only.
Google AI Mode is highest in all four verticals. ChatGPT is second in every one and Gemini third. The gap changes size and the order does not. Gener8 puts AI Mode at 1.5x Gemini in Consumer Electronics and 2.1x in Home Improvement.
Gener8 adds its own caveat, and it matters. AI Mode is typically reached inside Google search rather than opened as a destination, so its traffic likely skews more navigational and ready-to-buy than a standalone ChatGPT or Gemini session. That may explain part of its lead. I would not read the table as proof that AI Mode is the best platform for your brand. I would read it as showing that all three produce linked visits and the ordering is stable.
Adoption is jumping, and search is not shrinking
Gener8 reports that AI Mode going default at Google I/O 2026 sparked the sharpest adoption gains in its series. US AI 1+ day reach rose from 32% in April to 42% in June. The UK went from 21% to 38% over the same two months (Gener8, US and UK panel, AI 1+ day reach). Gener8 describes this as a structural product change.
More people reaching AI means the delayed visits sit behind a bigger audience. That inference is mine.
Search is not being cannibalised either. Searches per user are up 15% in the US and 20% in the UK since August 2025 (Gener8, US and UK panel). Gener8 reads AI and search as growing together, additive rather than substitutive.
Heavier searchers visit more, heavier AI users do not
Gener8 finds that heavier searching predicts more visits. The top search decile has a prompt-to-visit rate 2 to 4.3x that of the lightest decile across all four verticals (Gener8, AI Landscape July 2026, US mobile prompt-to-visit analysis). By GenAI-use decile, 3 of 4 verticals stay flat or decline. Apparel is the only climber and it fades by the top decile.
My read is that the people who act on an AI answer are people who already search a lot. That points back to branded search as the place the visit shows up.
Where Lmo7 stands on this
Everything above is Gener8’s data. This section is Lmo7’s view.
We have argued all year that AI visibility should be read as a trend and not a rank, and that sales evidence arrives later through proxies. Our piece on the proxy chain from AI visibility to revenue sets out that chain. Our read of the IAB measurement standard for AI visibility covers where the industry is heading on standards.
Gener8’s panel is a named, large-sample reinforcement of that position, and it puts a number on the lag. AI visibility is a leading indicator. The visit and the sale follow it, days later.
One caution. A Gener8 percentage is a panel rate for a vertical, not your conversion rate. I would use it to set the shape of the expectation and not to forecast revenue.
What to do
My recommendation would be to change four things in how you read AI, in this order.
- Widen the window. Judge AI on at least 72 hours, and ideally weeks. A same-session view will keep telling you the channel is empty.
- Watch branded and direct lift next to your AI visibility trend. Do not hunt for a referral line. Put branded search and direct traffic on the same chart as your AI visibility trend, week by week, and look for movement that follows.
- Set expectations by vertical. A fashion brand seeing 3.5% at 24 hours is in line with Gener8’s US mobile figure for Apparel & Fashion. That is normal, not failing. Consumer Electronics at 10.2% is the same kind of normal for that category.
- Keep a clean baseline month. Take a before and after baseline that has no promotions in it, so a promo spike does not get read as AI influence or hide it.
There is live proof that reading AI over a proper window works. At Trip Drinks, a 60-day Lmo7 engagement moved average position across the major LLMs from 7th to 3rd, and AI referral traffic rose 33% (source: Lmo7 client engagement with Trip Drinks, measured across the 60-day window). That was measured over a proper window, not a same-session report. It is one client and one engagement, and it is an Lmo7 result, not a Gener8 one and not a promise for your brand.
What to do next
To be clear, everything in the checklist above is work your own team can run this month with the analytics you already have. If you would rather do it with help, this is how we would sequence it, with the prices attached so you can judge the trade. Three steps, in priority order.
Fix the measurement window first. Owner: your head of ecommerce or Amazon. Our DaaS tier gives you direct API data access at £250 + VAT per month per marketplace. It is the measurement layer that reads AI-influenced visits over the right window. The commercial reason is simple. Every budget decision made on a 1-hour view is made on the smallest number in Gener8’s table.
Turn the trend into content work. Owner: your head of marketplaces or brand content. Challenger, Agentic Tracking & Content Optimisation is £1,000 per month. It is the track, diagnose and fix loop that converts a visibility trend into specific content changes. The reason to do it after step 1 is that you will know which changes moved branded and direct lift, and which only moved a rank.
Align the business on one window. Owner: the CEO or founder, with finance and marketing in the room. Enterprise workshops start from £5,000 and build multi-brand measurement frameworks and stakeholder alignment. This earns its cost when several teams or brands are reading AI on different windows and arguing about budget from different numbers.
Start with step 1 this week. Put your last full month of AI visibility data beside branded search and direct traffic, week by week, and see whether they move together.
Stephen Honight is the Founder of Lmo7, an agency working in AI search and agentic commerce. Lmo7’s current clients are Trip Drinks, Veloforte, Brown-Forman, Haleon, Pelotan and Symprove.