AI shoppers are worth more per visit. How many do you have?
Adobe says AI-referred shoppers are now worth far more per visit. Its report doesn't say how many of them there are.
In this article Alli Brand Equity

Figure 1
How AI-referred visits compared with other traffic
September 2026, US retail, percent higher than non-AI visits
Bar chart: AI-referred visits beat non-AI visits by 90% on conversion, 83% on revenue per visit, 73% on time on site, 53% on add-to-cart rate, and 25% on pages per visit
On this page
In August 2025, according to Adobe, visits from AI assistants converted 9% worse than the rest of a US retailer's traffic. By September 2026, Adobe found that they converted 90% better.
Adobe published that swing on October 29, with a headline figure: AI-referred visits earned 83% more revenue per visit than other traffic. But it doesn't say how many visits that is. Independent studies put AI referrals at roughly 0.2% to 1% of site traffic.
So the useful question is whether you can find these visitors in your own analytics, see what they're worth to you, and make sure your pages can be read by the systems sending them. Here is what Adobe reported, what it leaves out, and five checks to run first.
What Adobe reported
Adobe's data come from direct online transactions covering over 1 trillion visits to US retail sites. It counts AI traffic as users clicking a link in an AI answer, and began tracking in October 2024. Beyond the figures in the chart, it says AI traffic grew 71% year over year in September.
For the holidays, Adobe projects AI traffic to grow 159% year over year on Thanksgiving, 95% on Black Friday, and 88% on Cyber Monday. It also expects AI's revenue-per-visit advantage to double. Those are forecasts with no starting volume.
The trend is up, though not smooth. Across Adobe's 2026 reports, the pattern looks like this:
Figure 2
AI visits went from converting worse to converting better
Conversion of AI-referred visits versus non-AI visits, US retail, percent difference
Line chart: AI-referred visits converted 38% worse in March 2025 and 9% worse in August 2025, then 42%, 54%, 60%, 48%, and 90% better in March, May, July, August, and September 2026
| Month | Conversion vs non-AI | Revenue per visit vs non-AI |
|---|---|---|
| March 2025 | 38% lower | Not reported |
| August 2025 | 9% lower | Not reported |
| March 2026 | 42% higher | 37% higher |
| May 2026 | 54% higher | 53% higher |
| July 2026 | 60% higher | 53% higher |
| August 2026 | 48% higher | 43% higher |
| September 2026 | 90% higher | 83% higher |
In Adobe's data, conversion slipped from 60% higher in July to 48% in August, then jumped to 90%. Adobe's year-over-year growth rate for AI traffic also swung, from 62% in July to 127% in August to 71% in September. The reports we reviewed don't say why.
How big AI search is getting
OpenAI made its own claim on November 2. At a press event, Adam Fry, OpenAI's consumer product lead for ChatGPT, said the volume of searches run through ChatGPT now equals about 20% of the volume run through Google, The Verge reported. OpenAI counts a search as any query that leads ChatGPT to search the web. Defining a term or giving a simple how-to doesn't count.
Figure 3
ChatGPT now runs one search for every five on Google, OpenAI says
Search volume, with Google set to 100
Bar chart: Google 100, ChatGPT about 20, as reported by OpenAI in November 2026
Treat it as a company claim. It comes from OpenAI as it heads toward a possible IPO, and it can't be checked. But it points the same way as Adobe's growth rates.
It also measures something different from the rest of this article. A search inside ChatGPT is not a visit to your site. An answer can end in the chat window, and the visits that do arrive are what Adobe and the studies below count. So more AI searches make the question of how many reach you, and what they're worth, more urgent, but they don't answer it.
What the number leaves out
The first thing missing is size. Adobe's charts are indexed to October 2024, so Retail TouchPoints' reported rise of 1,772% since then could start from almost nothing. Independent measurements of AI's share of traffic are small.
Figure 4
Independent estimates of AI referrals as a share of traffic
Share of all website visits, different samples and periods
Bar chart: Kaiser and Schulze under 0.2%, Contentsquare 0.2%, SE Ranking 0.32%, Conductor 1.08%
Do the sum, and the scale of the point shows. If AI referrals were 0.32% of visits (the SE Ranking figure) and each earned 83% more than other visits, they would account for about 0.6% of revenue. That is our arithmetic, not Adobe's, and Adobe's retailers differ from SE Ranking's sample. Some retailers told Digital Commerce 360 AI is 5% or more of their traffic. Still, a big percentage on a small base is not yet a big business.
The second is cause. The 83% compares AI visits with non-AI traffic. A shopper who clicks a link in an AI answer may already be close to buying, and a click is a visit, not proof the answer caused the sale. That is a question for incrementality testing, which a report like this can't answer.
Third, the independent evidence on conversion is mixed. A peer-reviewed study of 973 e-commerce sites found organic search about 13% more likely to convert than ChatGPT, though on data through August 2025, when Adobe also had AI converting worse. Shopify data reported by CMSWire show AI visitors who reached a product page converting about 80% better than organic search referrals, while organic still sent more sessions than all AI platforms combined.
Last, Adobe sells things in this space. Adobe Analytics is the source of its data, so the sample is Adobe's customers. It launched LLM Optimizer in October 2025 and agreed to buy Semrush for $1.9 billion in November 2025, which reportedly closed in spring 2026. That doesn't make the figures wrong. Read them as a signal, and use your own data as the benchmark.
Can you see these visitors in your own analytics?
Not completely. Google Analytics 4 added an AI Assistant default channel on May 13, 2026, for visits whose referrer matches a recognized assistant such as ChatGPT, Gemini, Deepseek, Copilot, or Grok. Google hasn't published the full list, and visits with no referrer still land in Direct. AI Overviews and AI Mode sit in Organic Search, and Search Console reports them within the Web search type, with no separate breakout.
Bing Webmaster Tools covers the other side, which is citations. Its AI Performance report, released in preview, shows how often your site is cited in Microsoft Copilot, Bing's AI summaries, and select partner integrations. You can see which pages are cited, the grounding queries (the phrases the AI used to retrieve your content), the intent and topic of each query, and your citation share for it. There are no clicks, no filter by channel, and the partners aren't named. So it shows whether AI is using your content, but not what that use is worth. Read it next to your referral data.
Tagging is fragile too. In October, ChatGPT links briefly lost their utm_source=chatgpt.com tag. Adobe's Cedric Huesler reported that attributable ChatGPT visits fell to about a third of normal from October 8, with tagged clicks back near normal by October 15, and said OpenAI told him it was a bug. OpenAI hasn't made a public statement. Gemini's tagging is undocumented.
That makes referral counts a floor. Similarweb says referrer-based counts understate true volume. No independent study we found gives a trustworthy share of AI visits that hide in Direct. Two vendors that sell analytics tools put it between about 36% (Clickport) and about 71% (Loamly) of identifiable AI visits. This is signal loss in a new place. Crawler hits such as GPTBot are software requests, not visitors, so keep them out of conversion analysis.
What AI can read on your site
Adobe's other number is its Citation Readability Score, which it describes as how much of a page's content AI can understand and cite. The average US retail homepage scored 63% in September, so more than a third of homepage content isn't visible to AI systems. Electronics scored 79%, grocery 55%, and general merchandise homepages 40%.
Adobe doesn't publish how it calculates the score or which pages it sampled, and it sells the tool that addresses it. An Adobe description of a related extension suggests it compares a page's initial HTML with the fully rendered page, though we can't confirm that is the retail score. We found no independent study that isolates promotional layouts as a cause.
Three things hold up regardless. First, put product details, buying guides, FAQs, and return policies in the HTML your server sends. A December 2024 Vercel and MERJ analysis found that major AI crawlers fetch JavaScript files but don't execute them, and we found no newer primary statement from the major AI companies. Google says it can process JavaScript content that isn't blocked.
Second, allow the search bots if you want to appear in AI answers. OpenAI separates OAI-SearchBot, which surfaces sites in ChatGPT search, from GPTBot, which gathers training data. Third, Google's guidance is to write for your human audience, with no special markup or llms.txt required. That is the practical core of generative engine optimization.
Where this is heading
The direction isn't in doubt. Pew found 49% of US adults using AI chatbots in February 2026, and 42% using them to search for information. SE Ranking's measure of AI platforms' share of website traffic went from 0.02% in 2024 to 0.24% in 2025 to 0.32% in early 2026. Shopify reported AI-referred sessions up 197% year over year in the second quarter. Vendor or not, every source in this article points up.
What nobody can tell you yet is how fast, or how large. Today AI is a small share of traffic and a smaller share of revenue. We expect it to become a bigger part of how people find, compare, and choose brands, though that is our forecast, and no study has measured it yet.
It is also the case for doing the plain work now. Tracking set up this month gives you a baseline to read next year's growth against. Pages that AI systems can read get cited before the volume arrives.
You may also hear about llms.txt, a plain-text file at the root of a site that lists and summarizes your key pages in clear text for AI systems. It is a proposal, not an agreed standard. Ahrefs calls it "a speculative idea with no official adoption," and found that 97% of about 38,000 domains with a valid file received zero requests for it in May 2026 (Ahrefs sells SEO tools, and its sample is its own users). Google says you don't need one. A file like this is cheap to publish and could help tools that choose to read it. But we found no evidence that it improves AI retrieval, so it can't guarantee the right data gets found.
Not every change is in the words you write. How the site is built shapes how well AI systems can read and trust it: whether content is in the server-sent HTML, how headings and pages are organized, whether crawlers are allowed in, and whether structured data (schema) describes your products, prices, and FAQs accurately and matches what shoppers see. Microsoft's Fabrice Canel said at SMX Munich in March 2025 that schema markup helps Microsoft's LLMs understand content, as reported by Search Engine Roundtable. Google says structured data isn't required for its generative AI features. So treat schema as clarity for machines, with no guaranteed boost.
That is why an AEO and GEO audit is worth running. Look at your key pages the way an AI crawler would, check your robots rules for search bots, compare your schema with the visible page, confirm your product feeds are current, and record which questions your brand gets cited for. Audit tools score these differently, Adobe's included, so use the scores to find problems, and don't use them to compare yourself with others.
Why this is worth preparing for
There is little debate that AI search is growing. Pew found that about half of US adults (49%) use AI chatbots, in a survey of 5,119 adults fielded in February 2026. Adobe expects its AI traffic to keep climbing through the holidays, though that is a forecast. It is reasonable to expect AI to take a larger share of how new customers find and compare brands.
How much it matters will still vary by brand. It depends on your category, how your customers research a purchase, and how much AI traffic you already see, which is why the numbers above are worth pulling. Gartner's January 2026 survey found only about a third of consumers say AI chatbots are as effective as search engines, and it advises marketers to optimize for both. Much of this work (clear pages, accessible content, accurate data) helps in traditional search too, costs little, and prepares your brand as search habits keep changing.
Where Alli fits
PMG's teams, including its Forward Deployed Engineering team, do the AEO and GEO work for clients: readable pages, crawler access, structured data, and product feeds. Alli Brand Equity measures whether it's working. Its Share of LLM Presence measure sits alongside the brand's other signals, so teams can see how often AI answers name and cite the brand, next to the traffic and revenue in their own reporting.
None of that replaces your own analytics. The first step is still the one above: find the AI Assistant channel and write down your numbers.
Before Thanksgiving, pull the AI Assistant channel for the last 60 days and write down three numbers: sessions, orders, and revenue per session. Then the argument about Adobe's 83% becomes an argument about yours.
Frequently asked questions
Does AI traffic convert better than other traffic?
Adobe reports that AI-referred visits to US retail sites converted 90% better than non-AI visits in September 2026, after converting 9% worse in August 2025. Independent evidence is mixed. A peer-reviewed study on data through August 2025 found organic search converting better than ChatGPT.
How much traffic comes from AI assistants?
Independent studies put AI referrals at roughly 0.2% to about 1% of site traffic, depending on the sample and period. Adobe does not publish a share. Some retailers report 5% or more of their own traffic.
How do I track AI traffic in Google Analytics 4?
GA4 has had an AI Assistant default channel since May 13, 2026, for visits whose referrer matches a recognized assistant. Visits with no referrer land in Direct, so counts are a floor. Google places AI Overviews and AI Mode traffic in Organic Search.
Do I need an llms.txt file?
No platform has said it needs one. llms.txt is a proposal, not an agreed standard, and Google says no special files are needed. Ahrefs found 97% of valid files received no requests in May 2026. It is cheap to publish but not a proven fix.
Can I see which of my pages AI cites?
Partly. Bing Webmaster Tools has an AI Performance report in preview that shows citations in Microsoft Copilot, Bing's AI summaries, and select partners, by page and by grounding query. It does not show clicks or name the partners.
What is Adobe's Citation Readability Score?
It is Adobe's own metric for how much of a page's content AI can understand and cite. Adobe reported an average of 63% for US retail homepages in September 2026. The calculation method is not published.
How do I make a retail site readable to AI?
Put product details, buying guides, FAQs, and return policies in the HTML the server sends, allow search-oriented AI bots in robots.txt, and keep product data current. Google says to write for people and that no special markup is required.
Sources
- Adobe, AI traffic drives 83% higher revenue per visit, October 29, 2026. Figures as printed in the report; Adobe sells Adobe Analytics and LLM Optimizer.
- Retail TouchPoints, Adobe: AI traffic driving higher revenue per visit for retailers this holiday season, October 29, 2026.
- Adobe Digital Insights, 2026 holiday forecast (September 2026).
- Adobe Digital Insights, AI traffic trends report, August 2026.
- Adobe Digital Insights, Q3 AI traffic trends report (June 2026).
- TechCrunch, AI traffic to US retailers rose 393% in Q1, April 16, 2026.
- The Verge, OpenAI says ChatGPT is catching up to Google in search, November 3, 2026. OpenAI did not say where its Google figure came from.
- Kaiser and Schulze, Marketing Science, online April 21, 2026.
- SE Ranking, AI traffic study, June 18, 2026. SE Ranking sells SEO and AI visibility tools.
- Contentsquare, AI-referred traffic, April 3, 2026. Contentsquare sells digital experience analytics.
- Conductor, 2026 AEO/GEO benchmarks report. Conductor sells AEO and SEO software.
- CMSWire, Does AI search really beat organic? Not yet, Shopify says, August 26, 2026. Shopify sells commerce software.
- Digital Commerce 360, How AI referrals impact traffic at four online retailers, September 16, 2026.
- Google Analytics Help, Traffic-source dimensions, and What's new (May 13, 2026).
- Search Engine Journal, Google Analytics adds AI Assistant as a default channel group, May 14, 2026.
- Google Search Central, AI features and your website, and Optimizing your website for generative AI features (updated July 10, 2026).
- Pew Research Center, Americans and AI 2026, June 17, 2026. Survey of 5,119 US adults, February 17 to 23, 2026.
- Search Engine Land, Bing Webmaster Tools AI Performance report, 2026, and Practical Ecommerce, Bing's improved AI performance section, June 22, 2026.
- Ahrefs, What is llms.txt?, updated June 15, 2026. Ahrefs sells SEO tools.
- Search Engine Roundtable, Microsoft confirms schema helps its LLMs, March 20, 2025.
- Gartner, Only one-third of consumers say GenAI rivals search engines, January 20, 2026.
- Relevant Audience, ChatGPT UTM tag loss was a bug, October 16, 2026, and Search Engine Roundtable, OpenAI drops UTM parameters from ChatGPT results, October 14, 2026.
- Similarweb, AI referral traffic by industry, September 3, 2026. Similarweb sells digital intelligence tools.
- Clickport, ChatGPT direct traffic in GA4 (sells analytics), and Loamly via Ecommerce Fastlane, Your AI traffic is filed as Direct, August 3, 2026 (sells detection software).
- Adobe, LLM Optimizer launch (October 14, 2025), and Semrush acquisition announcement (November 19, 2025).
- Vercel and MERJ, The rise of the AI crawler, December 17, 2024, and OpenAI crawler documentation.
Last updated November 17, 2026. Statistics are quoted from the sources above and were checked on that date. Survey results are self-reported unless noted.
