For most websites, AI assistants send roughly 0.5–3% of total traffic as of mid-2026, with technology and ecommerce sites reaching 5–8% — but that published range is both an undercount of your real number and a poor predictor of it. The honest answer to "what percentage" is that there isn't a single one: it swings by an order of magnitude across niches, and standard analytics miss a large share of the AI referrals that do arrive, so any figure you read is a floor, not your number.
We build an AI-traffic counter for WordPress, and the pattern we see on real stores is consistent: after we fix the measurement gaps, the AI share is reliably higher than the site's default Google Analytics report showed — sometimes several times higher — because GA was filing those visits under "direct" or "organic." So the useful version of this question isn't "what's the average" — it's "what's mine, measured correctly, and is percentage-of-traffic even the metric that matters?"
What percentage of traffic comes from AI assistants right now?
Across most sites it's a low single-digit percentage in mid-2026 — commonly cited between 0.5% and 3% of total sessions, climbing to 5–8% on tech and ecommerce sites where users research purchases inside a chatbot. That range comes from 2026 traffic reports (Similarweb and referral analyses from SEO firms), and it's growing fast, but "growing fast from a small base" is the accurate framing, not "AI is a major traffic source" for the median site.
One number that is stable is which assistant sends the traffic: ChatGPT dominates AI referrals by a wide margin.
That chart is usage share — who people talk to. The split of actual clicks sent to external sites is even more lopsided: one July 2026 referral analysis put ChatGPT at over 90% of AI referral traffic, because platforms like Gemini answer more often without linking out. The practical takeaway is that if you're measuring AI traffic and only instrument one source, instrument ChatGPT first. The number you'll get is real but, as the next section explains, almost always understated — which is the same reason Google Analytics undercounts your AI traffic.
Why is the real percentage for your site higher than analytics shows?
Because a large share of AI-referred visits arrive with no usable referrer, so analytics files them as "direct" and the AI share silently vanishes into an unattributed bucket. Three mechanics cause this, and they stack:
- Stripped referrers — links opened from a native app (the ChatGPT or Perplexity mobile app) or an
https→httphop often carry noreferrerheader, so the visit looks like someone typing your URL from memory. - Cached and zero-click answers — when the assistant quotes you without a click, there's no session at all, yet your content did the work; conversely, a later "direct" visit from someone who saw you in a chat is AI-driven but untagged.
- Default channel grouping — GA4's out-of-the-box rules didn't have AI-referral channels for a long time, so ChatGPT and Perplexity visits landed in Organic or Direct rather than a distinct AI line.
This is the exact gap Contexta's AI Traffic tracking closes — it counts visitors arriving from ChatGPT, Perplexity, Gemini and Copilot using both the referrer and the utm_source tag, and it's built to survive caching so the visits GA drops still get counted. The point isn't a bigger vanity number; it's that you can't act on a channel you can't see.
Why does the percentage vary so much between sites?
Because AI traffic tracks query intent, and intent varies enormously by what you sell and who you serve. A site answering informational, how-to and comparison questions gets cited by assistants constantly; a site built on branded navigational traffic or transactional short-tail sees far less, because those journeys don't route through a chatbot. That's why a B2B SaaS blog and a local plumber can both be "average" sites and sit ten times apart on AI share.
The variation is also why chasing a benchmark is a trap — the honest move is to read your own referral and utm_source data rather than a report about someone else's niche. When ChatGPT began appending a tag to its outbound links, it made this measurable per-visit, which we covered in what utm_source=chatgpt.com means. Your mix of query types, not an industry average, decides your real percentage.
Is percentage of traffic even the right metric for AI?
Not on its own — percentage of sessions is a lagging, partial view of AI's actual impact, because the biggest effect happens in answers where nobody clicks. When an assistant recommends your product or quotes your guide and the user acts without visiting, you got the influence and zero sessions, so a sessions-based percentage scores that as nothing.
Why the percentage understates AI's real effect
An AI answer that mentions you but sends no click still shapes the buyer's decision.
Branded searches and "direct" visits often trace back to an earlier AI conversation.
The metric that matters is AI-influenced outcomes, not just AI-referred sessions.
This is where the AI era diverges from classic search analytics, and why measuring AI presence has to sit alongside measuring AI referrals — the reasoning we laid out in GEO versus SEO writing. Treat the referral percentage as a directional signal that AI knows you exist, then judge success by whether AI-influenced conversions and branded demand are rising, not by the referral line alone.
How do you measure your own AI-assistant traffic percentage?
Measure it from two signals — the HTTP referrer and the utm_source parameter — and count any visit whose source is a known AI host (chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com) as AI traffic. Doing this properly means adding an explicit AI channel rather than trusting the default grouping, and capturing the tag on entry before a client-side redirect or cache can drop it. The full setup for WordPress, including the referrer list and the channel definitions, is in how to track visitors from ChatGPT, Perplexity and Gemini.
Once you're counting correctly, compute the percentage as AI-sourced sessions over total sessions for a fixed window, and watch the trend rather than the absolute number — a share moving from 1% to 3% over two quarters tells you more than any single snapshot. Pair it with a note of which pages the AI visits land on, because that reveals what assistants are citing you for, and that is the part you can act on.
FAQ
What percentage of website traffic comes from AI assistants in 2026?
For most websites it's a low single-digit share — commonly cited between 0.5% and 3% of total sessions as of mid-2026, rising to 5–8% on technology and ecommerce sites. The figure is growing quickly from a small base, so it's directional rather than fixed, and it varies by an order of magnitude depending on your niche and query mix. Any published average is a floor for your own site, because standard analytics miss a share of the AI referrals that actually arrive.
Which AI assistant sends the most website traffic?
ChatGPT sends the overwhelming majority of AI referral traffic — one July 2026 referral analysis put it above 90% of clicks that AI tools send to external sites. Its lead in referrals is even larger than its lead in usage, because assistants like Gemini answer more often without linking out. If you can only instrument one AI source in your analytics, instrument ChatGPT first.
Why does my analytics show almost no AI traffic?
Because a large share of AI-referred visits arrive with no usable referrer and get filed as 'direct' or 'organic' instead of AI. Links opened from the ChatGPT or Perplexity mobile app often carry no referrer header, cached and zero-click answers create no session at all, and GA4's default channel grouping had no dedicated AI line for a long time. The traffic is real; it's landing in an unattributed bucket, so your true AI percentage is higher than the default report shows.
Is AI traffic worth optimizing for if it's only a few percent?
Yes, for two reasons — it's growing fast, and the sessions figure understates AI's real influence on buyers. Many AI answers mention or recommend you without sending a click, so they shape a decision while scoring as zero traffic, and later 'direct' visits often trace back to an earlier AI conversation. Judge the channel by whether AI-influenced demand is rising, not by the referral percentage alone.