Last updated September 2026. Added Adobe’s July 2026 reading and the Digital Commerce 360 report, confirmed GA4’s AI Assistant channel excludes Perplexity, and added the Loamly and Opollo benchmark figures.
On Aug. 19, 2026, Digital Commerce 360 published Adobe’s latest read on AI-driven shopping traffic: AI referrals to U.S. retail sites grew 62% year over year in July 2026, and those visitors converted 60% higher than non-AI traffic. It was the 11th straight month Adobe reported AI traffic beating every other channel on conversion.
That single data point answers the question most marketers actually have: does AI traffic convert, or does it just browse? Adobe’s answer, repeated for nearly a year now, is that it converts.
But the traffic-and-conversion story is only half the picture. The other half is measurement, and most teams are getting that half wrong. This page collects the AI referral traffic numbers worth citing in 2026: Adobe’s retail series, conversion benchmarks outside retail, engine-level share of traffic, and the large share of AI sessions that never show up in your analytics at all. Every figure names its source and its date. Where two credible studies disagree, you see both.
The latest Adobe reading: July 2026
Adobe Analytics tracks direct online transactions across more than 1 trillion visits to U.S. retail sites, and it publishes a fresh read most months. Here is the July 2026 snapshot, as reported by Digital Commerce 360 on Aug. 19, 2026, sourced to Adobe.
| Metric | July 2026 |
|---|---|
| AI-referred traffic, year over year | +62% |
| Conversion rate vs. non-AI traffic | +60% |
| Consecutive months AI traffic has out-converted other channels | 11 |
| Time on site vs. non-AI traffic | +59% |
| Bounce rate vs. non-AI traffic | -33% |
| Add-to-cart rate vs. other shoppers | +28% |
| Revenue per visit vs. other shoppers | +53% |
Source: Adobe Analytics, via Digital Commerce 360, Aug. 19, 2026.
This table updates when Adobe publishes its next monthly read. Treat every figure in it as retail-specific and U.S.-specific. Adobe’s panel comes from its own analytics customers’ transaction data, not a random sample of the internet, and it says nothing about B2B, SaaS, or any market outside retail.
Which AI engine sends the most referral traffic
ChatGPT sends the most. By how much depends entirely on what a study measures, and that gap is the clearest illustration of why a single “AI share” number is hard to trust on its own.
| Study | What it measures | ChatGPT’s share | Date |
|---|---|---|---|
| SE Ranking (101,574 sites) | Referral clicks sent out to external websites | 75.96%, down from 79.74% a year earlier | July 2026 |
| Similarweb | Visits to the AI platforms themselves, not referral clicks out | About 53%, down from about 76% a year earlier | July 2026 |
| Grips Intelligence, via eMarketer | Share of AI-attributed purchase paths | 92.2% (Gemini 5%, Perplexity 1.9%) | 2026 |
None of these numbers are wrong. They answer three different questions: how many clicks ChatGPT sent out, how many people used ChatGPT versus a rival chatbot, and how many purchases traced back to ChatGPT specifically. Pick the definition that matches what you are trying to report, and say which one you used. Most articles that cite a single “AI share” percentage do not.
The trend matters more than any single snapshot. SE Ranking’s own two data points, 79.74% in 2025 and 75.96% in 2026, show ChatGPT’s share sliding even as its absolute referral volume keeps growing. A shrinking share of a fast-growing category is not a warning sign. It is what fragmentation looks like once Gemini and Perplexity both start sending meaningful traffic of their own, which is exactly why tracking a single engine gives you an incomplete read on your own site’s numbers too.
AI-referred traffic converts better outside retail too
Adobe’s conversion numbers come from retail. The pattern holds in B2B as well.
Opollo’s 2026 AI Search Benchmark Report, published Feb. 22, 2026, tracked 312 B2B technology firms from January 2025 through January 2026. It found AI-referred visitors converting at a mean rate of 14.2%, versus 2.8% for Google organic traffic, a gap of roughly five times. Opollo excluded firms with fewer than 10 AI sessions a month and removed any conversion rate above 35% to keep outliers from skewing the average.
Two studies, two industries, two independent methodologies, and the same direction: AI-referred visitors show up ready to act. That consistency across unrelated datasets is worth more than either number alone.
The undercount: why your analytics show less AI traffic than you actually have
Every number above assumes your analytics can see the traffic in the first place. Most cannot.
Loamly’s State of AI Traffic 2026 report, updated in February 2026 from a database of 446,405 visits, found that 70.6% of AI-driven traffic arrives with no referrer header at all. Without a referrer, GA4 has nothing to classify the session by, so it defaults to Direct, indistinguishable from someone who typed your URL from memory.
This is not low-value traffic hiding in the numbers. Loamly’s data shows this no-referrer traffic (also called dark AI traffic) converting at 10.21%, against 2.46% for ordinary non-AI direct traffic, a gap of more than four times. The sessions your dashboard cannot name are often the best ones you have.
The mechanism is mostly technical, not deliberate. Mobile AI apps frequently open links inside an in-app browser that strips the referrer. Someone who copies a link out of an AI answer and pastes it into a new tab carries no referrer either way. No-referrer traffic is the name for this gap: real visits driven by AI research that arrive with no way to trace them back to their source.
GA4’s AI Assistant channel: what it catches, and what it still misses
Google shipped a partial fix. According to Google’s own release notes, GA4 added a native AI Assistant channel to its default channel groups on May 13, 2026, automatically sorting sessions from a short list of named chatbot referrers, including ChatGPT, Gemini, and Microsoft Copilot, into one labeled bucket.
It is a real improvement, and it needs no setup. It is also incomplete in two specific ways.
First, Perplexity is not on the list. Its sessions still land in generic Referral, not AI Assistant. A GA4 channel audit from Digital Applied confirmed the gap and noted you cannot add Perplexity to the native channel yourself, only work around it with a custom channel group.
Second, Google’s own AI Overviews and AI Mode do not count as AI Assistant traffic either. GA4 classifies clicks from both as Organic Search, which for most sites is likely the single highest-volume source of AI-driven visits, hiding inside a metric nobody thinks to break apart.
Neither gap touches the 70.6% of sessions that arrive with no referrer at all. A channel group can only sort traffic it can see.
What each measurement method misses
Put every method side by side, and the pattern holds: nothing sees the whole picture alone.
| Method | What it captures | What it misses |
|---|---|---|
| Adobe Analytics (retail panel) | Aggregate trend and conversion lift across a huge retail sample | Individual-brand detail; anything outside retail; your own site’s number |
| GA4 default channel groups | AI Assistant sessions from named chatbot referrers, automatically | Perplexity; Google AI Overviews and AI Mode; every no-referrer session |
| GA4 custom channel group | Any referrer domain you add yourself, including Perplexity | Every no-referrer session; needs manual setup and upkeep |
| Server-side or CDN log analysis | Bot and crawler visits; some no-referrer sessions via behavioral matching | A confirmed source for the match; a documented history of over-attribution |
| AI visibility monitoring tools | Whether your brand gets mentioned upstream, including zero-click answers with no visit at all | Actual session-level traffic and conversion; that stays a job for analytics |
| Similarweb or Semrush traffic estimates | Directional share of AI platform usage and competitor benchmarking | Your own site’s exact referral sessions and conversions |
Tools for closing the gap
No single tool answers both halves of this problem: what your AI traffic converts at, and whether your brand shows up in AI answers in the first place. You need one tool for each job.
For referral analytics, start with a custom GA4 channel group that explicitly adds Perplexity and any other referrer the native AI Assistant channel misses. This walkthrough covers the exact hostnames and regex to use. Similarweb and Semrush both layer directional, competitor-level traffic estimates on top of your own GA4 data, useful for benchmarking share even though neither replaces your first-party numbers.
For prompt and citation monitoring, the job is different: knowing whether your brand gets named at all, including in the zero-click answers that never produce a session. Otterly.AI is a straightforward entry point for prompt-level tracking on a tight budget. Profound goes deeper on citation-level attribution for teams reporting to a board. Temso combines both jobs in one subscription, tracking share of voice and citations across engines alongside a prioritized fix queue, from $89 a month.
Running only one half of this stack tells half the story. A brand with solid analytics and no monitoring sees which AI sessions convert but misses the answers where a competitor gets named instead. A brand with monitoring and no analytics setup knows it gets cited but cannot prove the revenue behind it. The two jobs answer different questions, and a complete report needs both.
Full comparisons, pricing, and engine coverage for every tool in this category sit at /rankings/ai-visibility-tools, and the scoring approach behind that ranking is documented at /methodology. For the mention-side benchmarks behind these tools (share of voice, sentiment, citation rate), see the AI brand visibility statistics roundup and the AI visibility glossary.
What this means for your reporting
Adobe’s own framing is the useful part. For 11 straight months, AI-referred traffic has out-converted every other channel Adobe tracks, including paid search. That is not a brand-awareness footnote anymore. It is a revenue channel with a measurable return.
Report it that way. Put AI referral traffic in the same attribution model you already use for paid search: sessions, conversion rate, revenue per visit, and the cost of acquiring the traffic, even if that cost today is mostly content and citations rather than media spend. Then separately track whether your brand gets mentioned upstream at all, since a meaningful share of your best AI traffic is currently invisible to the model you just built.
Budget conversations get easier once this channel sits next to paid search in the same spreadsheet, instead of sitting in a vague “AI visibility” line item nobody can defend in a planning meeting.
This is also the argument to bring into that meeting. A channel that has converted better than every alternative for 11 straight months, in a dataset as large as Adobe’s, does not need a separate justification process from paid search. It needs the same one, with the same rigor applied to both the win rate and the gaps in how your team is counting it.
Start with your own numbers. Pull your GA4 AI Assistant channel this week, cross-check it against a custom channel group that adds Perplexity, and estimate what a 70.6% no-referrer gap is hiding inside your Direct traffic. That baseline is what turns “AI is probably growing” into a number you can defend in a budget meeting.
Sources cited
- Adobe Analytics, via Digital Commerce 360: “Adobe: AI-referred retail traffic up 62% YoY in July, converting 60% better than other channels,” Aug. 19, 2026.
- Opollo: “The 2026 AI Search Benchmark Report” (312 B2B technology firms), Feb. 22, 2026.
- SE Ranking: “Referral Traffic from ChatGPT Hit an All-Time High in May 2026” (101,574 sites), July 9, 2026.
- Similarweb: “AI Search Stats 2026: Market Share, Referral, and Citation Data,” July 2026.
- Grips Intelligence, via eMarketer: AI referral traffic and purchase-path data, 2026.
- Loamly: “State of AI Traffic 2026: Industry Benchmark Report” (446,405 visits), updated Feb. 2026.
- Google Analytics: “What’s new in Google Analytics” release notes, May 13, 2026.
Monthly refresh cadence: the Adobe reading in this post is checked against Digital Commerce 360’s latest coverage each month. Check the updatedDate in the frontmatter for the most recent revision.