How Much Traffic Comes From ChatGPT in 2026? The Attrifast 200-Site Cohort Benchmark

Part of the AI Search Hub — browse all 35 AI Search guides.

For the last six months the most common inbound message I have received looks like this: a screenshot of a GA4 channel report, a Direct/(none) row growing faster than every other channel, and one sentence underneath: "How much of this is ChatGPT?" By March of 2026 I was getting that screenshot multiple times per week. So I sat down with the dataset I actually have access to, the roughly 200 Stripe-connected sites instrumented through Attrifast, and cut the numbers.

This is the result. It is the article I wished existed when I started getting those screenshots. It is also, by design, a backlink magnet: there is no other public dataset I have found that breaks ChatGPT traffic share down by vertical, by site size, by geography, and against the actual server-side recovered figures rather than the GA4 default that misses two-thirds of the channel. If you cite this article in your own writing, please link back. The numbers are ours, the methodology is below, and we welcome challenge.

Two caveats before the data starts. First, this is our cohort, not an industry-wide truth. The 200 sites are bootstrapped SMBs in the $5k to $250k MRR range, skewed Stripe-native and skewed toward English-speaking US and EU markets. Where our numbers diverge from public benchmarks like Similarweb's chatbot tracker, Cloudflare Radar's AI insights, or Profound's benchmark reports, I flag it. Second, the methodology section explains exactly what is in and out of the dataset. If you only read one section, read that one, because every other number depends on it.

Quick facts

Two numbers anchor everything else here. The first is 800 million weekly actives on ChatGPT in Q1 2026 [1], up from 400 million in Q4 2025. That is the demand side, and at this scale even a single-digit referral rate produces material site traffic. The second is 71% of ChatGPT-referred sessions arrive with no referrer at all in our cohort. That is the measurement-error side, and it is the reason most operators reading their own GA4 dashboards have no idea how much of their traffic actually comes from ChatGPT. The rest of this article is what falls out when you take those two numbers seriously.

Why I ran this benchmark

The honest answer is that I got tired of the same conversation. Someone messages me a screenshot of an inflating Direct bucket, asks how much of it is ChatGPT, and I would walk through the referrer-stripping mechanics, point at dark AI traffic in GA4, and explain that they need server-side recovery. They would nod and ask for a number. And the public datasets did not give a number that answered the question.

The public datasets are good at what they actually measure. Similarweb's chatbot tracker tells you how many people visit chatgpt.com itself, which crossed 5.2 billion monthly visits by March 2026. Cloudflare Radar's AI insights tells you that AI bots are roughly 5 to 6% of all bot traffic on the Cloudflare edge. Profound and Otterly tell you whether your brand appears in AI answers. None of these answer "how many real humans does ChatGPT send to a real third-party website."

That last gap is the one Attrifast's data shape is built to fill. Every site in the cohort runs a server-side script that captures the Referer header before any client-side processing strips it, applies behavioral heuristics to the no-referrer fraction, joins the session to a Stripe payment event by webhook, and stores the result for the operator. Cutting across 200 of those sites gives a defensible cohort number for "what percentage of measured sessions actually came from ChatGPT." That number is what this article is about.

I also want to be specific about what I am not claiming. I am not claiming this dataset is representative of the entire web. It is biased toward Stripe-native businesses, bootstrapped SMBs, US and EU markets, and sites that suspected they had un-attributed AI traffic in the first place (self-selection). The numbers below are cohort numbers, not industry truth. Where I have cross-checked against public benchmarks like Similarweb, Cloudflare Radar, and the Profound 2026 benchmark report, the directional pattern agrees. Where the magnitudes diverge, I flag it.

Methodology, in brief

The full version lives in the 2026 AI search revenue benchmark methodology section. Here is the short form.

ChatGPT-attributed sessions are detected via four signals: referrer match (chatgpt.com, chat.openai.com, oai.com), UTM (utm_source=chatgpt), User-Agent match for the OAI-SearchBot citation-fetch leg, and behavioral fingerprinting on no-referrer deep-page entries that cluster with known ChatGPT traffic by geography and time of day. The behavioral layer validates at roughly 80% precision against a post-purchase survey ground truth on the subset of converted sessions where we can ask. The 20% noise floor is a known limit and is flagged where it changes a headline number materially.

The cohort vertical mix matters for reading the table that follows. Here is the breakdown.

The ecommerce slice is further broken into apparel (18 sites), beauty and personal care (12), supplements (11), home goods (7), and consumer electronics (6). The SaaS slice is broken into developer tools (28), security and infrastructure (19), marketing and analytics (24), HR and payroll (12), and general productivity (35). I cite these sub-slices below when the data warrants.

The headline number: 3.4% of sessions come from ChatGPT

The single number that matters: the median Attrifast cohort site received 3.4% of its measured sessions from ChatGPT during the rolling 30 days ending May 15, 2026. Interquartile range 1.1% to 6.8%. The distribution is wider than I expected going in.

This is the recovered number, meaning it includes the no-referrer fraction reconstructed by server-side fingerprinting. If you took the same cohort and read only GA4 default attribution, the median would sit at 1.0%, because GA4 catches only the 29% of ChatGPT sessions that survive the referrer-stripping mechanics described in our chatgpt referral analytics guide.

Read the histogram carefully. Roughly 20% of cohort sites sit below 1% ChatGPT share, mostly local services and high-impulse retail. Roughly 30% sit between 2 and 5%, which is the mainstream SaaS and ecommerce zone. About 10% sit above 8%, almost all of them developer-tools, security, or analytics-buyer SaaS. The long right tail is where the operators who are most aggressive about ChatGPT visibility have already concentrated their efforts, and the curve I expect to see flattening upward over the next 12 to 24 months as the rest of the distribution shifts right.

For comparison, here is the same cut split by what GA4 would have reported on its own versus what server-side recovers. The gap is the entire point.

The recovery gap is remarkably stable across percentiles, sitting between 71 and 75%. That stability is what convinced me the no-referrer mechanic is mostly a single underlying behavior (AI clients strip the Referer header on outbound clicks) rather than a per-site oddity. The implication: if you are reading GA4 and seeing ChatGPT at 1%, your real number is almost certainly between 3 and 4%. The multiplier is roughly 3 to 4x on the median, and it is stable enough across our cohort to be useful as a back-of-envelope correction. This is the same mechanic that powers the ga4 missing traffic write-up.

ChatGPT traffic by vertical, ranked

This is the table that I think will be cited the most often, so I am putting it early. Median ChatGPT share of sessions, by vertical, May 2026, across the Attrifast cohort.

The spread is roughly 18 to 1 between the top of the table and the bottom. That spread is the single most important takeaway: there is no useful single answer to "how much traffic comes from ChatGPT" that is industry-blind. Anyone publishing an unsegmented average is averaging local plumbers and DataDog competitors into the same number.

The pattern within the table is consistent with two structural drivers. First, buyer research intensity: verticals where buyers compare options before purchasing (security tools, developer SaaS, consumer electronics) sit higher than verticals where purchases are habitual, impulse, or locally constrained. Second, content depth and citability: verticals with deep technical documentation and benchmark posts (developer tools especially) get cited more frequently in ChatGPT answers, because the engine pulls passages and citations from extractable, structured content. Local services have neither structural feature: their queries rarely surface an AI answer, and when they do, the citation pool is dominated by Yelp, Google Maps, and directory aggregators rather than the local business's own site.

The Profound 2026 benchmark report (tryprofound.com/profound-index) shows a similar directional vertical pattern when measured at the citation-share level rather than the session-share level, with developer tools and security at the top and local services at the bottom. Their absolute numbers are not directly comparable (they measure citation appearance, not click-through), but the rank order largely matches.

ChatGPT traffic by site size

The second-most-cited cut, I expect. Median ChatGPT share of sessions, by site monthly session volume, May 2026.

The pattern is the inverse of what most operators expect. Smaller sites in the cohort have higher ChatGPT share than larger sites, not lower. The reason is mathematical and worth stating explicitly: larger sites have proportionally more branded-search and direct type-in volume that dilutes the AI fraction, while smaller sites have a more discovery-heavy traffic mix where any new discovery channel shows up as a larger fraction. The absolute volume of ChatGPT sessions is much higher on larger sites (a 1M-session site at 1.7% is still 17,000 ChatGPT sessions, more than a 10K-session site at 5.1% which is 510 sessions). The share is just smaller because the denominator is bigger.

This pattern matters for how you interpret a single-site number. If you run a 20,000-session site and you see ChatGPT at 5%, you are at the cohort median for your size band, not above it. If you run a 600,000-session site and you see ChatGPT at 2%, same story. Comparing a small site's ChatGPT share against a large site's published ChatGPT share without normalizing for size will mislead you in both directions.

In absolute terms, here is the same data as session counts rather than shares, to make the volume story explicit.

The largest sites still pull the most absolute ChatGPT traffic by a wide margin. A 1M-session site averaging 24,000 ChatGPT visits per month is doing 90 to 100x the volume of a 5,000-session site averaging 270. The share number flips the ranking, but the absolute opportunity scales with site size. Both readings are correct; both matter for different decisions.

ChatGPT traffic growth from 2024 to 2026

The 24-month trend is the part most operators most want to see, because it answers the question "is this getting bigger" in a way the snapshot does not. Cohort-aggregated ChatGPT-attributed sessions, indexed to May 2024 = 100, across the subset of sites with continuous data through the window (n=87 sites).

A roughly 12x increase in ChatGPT-attributed traffic across the 24-month window, compounding at 11.2% monthly on average. The trend is not perfectly smooth: three step changes are visible in the data. The first is Q4 2024 around the broad rollout of ChatGPT Search (as covered by OpenAI in October 2024 and confirmed in subsequent Reuters reporting). The second is Q1-Q2 2025 around expanded mobile and desktop app distribution. The third is Q1 2026, the largest single quarter at +37%, which aligns with the 800-million-weekly-active milestone OpenAI confirmed in early 2026 (Reuters coverage tracks this through Q1 2026).

The same data as month-over-month percent growth, to make the rate visible.

The growth rate is reasonably stable in the 10 to 13% range across the window. There is no inflection toward slowdown yet in our cohort, though aggregate ChatGPT product growth is reportedly moderating from the 800M-WAU milestone according to reporting in WSJ and The Information during Q1 2026. Click-through from answers may be increasing even as total user growth slows, which would keep referral traffic compounding for some time even after the user base flattens.

For comparison, here is how other channels grew on the same cohort sites across the same window.

ChatGPT is third-fastest behind Perplexity and Claude, both of which start from much smaller absolute bases. Google organic on the same sites grew at 1.1% monthly, which means ChatGPT-attributed traffic is compounding at roughly 10x Google organic's rate on a smaller base. The gap between the rates is what closes the volume ratio over time.

What GA4 reports versus what we measured

This is the table I expect to be referenced most often in product marketing decks. Median ChatGPT-attributed sessions, GA4 default versus server-side recovery, May 2026.

Read this carefully. On the median cohort site, GA4 default attribution misses 2.4 percentage points of ChatGPT traffic, which is more than the ChatGPT share GA4 actually reports (1.0%). The 75th percentile site is missing 4.9 percentage points, the 95th percentile is missing 8.1. The gap is widest on sites that get cited most often, because they have the most volume to lose to the no-referrer mechanic.

The gap by detection method:

The "100%" in the last row is the defined recovery target against post-purchase survey ground truth; the actual precision sits around 80% on the behavioral fingerprinting layer, so the recovered number has roughly a 20% confidence band on the no-referrer fraction. The detail of that recovery is in the track ChatGPT traffic guide.

What does this mean if you run a site? Multiply your GA4-reported ChatGPT share by roughly 3.5x to estimate your actual share. If GA4 says ChatGPT is 0.6%, your real share is probably in the 2 to 2.5% range. If GA4 says 1.5%, you are probably looking at 5 to 5.5% actual. The multiplier is approximate but cohort-stable enough to be useful as a starting point.

ChatGPT traffic by geography

Geographic distribution matters because ChatGPT adoption is uneven across regions, and your traffic share will scale with where your audience lives. Across the Attrifast cohort, May 2026, share of total ChatGPT-attributed sessions by visitor region:

The US over-indexes mildly, EU under-indexes mildly, and the largest under-index is APAC excluding India. The Pew Research 2025 AI adoption survey (pewresearch.org) shows a similar US-over-EU gap on self-reported ChatGPT usage, and Similarweb's panel data confirms the US is the dominant geography for chatgpt.com itself.

Within the cohort, ChatGPT share of sessions also varies by region after controlling for vertical. Same cohort, median ChatGPT % of sessions by region of visitor:

The continental-EU figures are likely understated by 10 to 20% due to the combination of GDPR consent loss and the AI-client referrer-stripping issue both eroding the same metric. The detail on the GDPR side is in our gdpr analytics 2026 playbook. The US-EU gap in actual underlying behavior is probably narrower than the table suggests, but the directional pattern (US first, English-speaking second, continental EU third) is consistent across every cut we have run.

A couple of notable absences in the table: China is missing because ChatGPT is not accessible there in any meaningful way (the cohort sees a near-zero share from China-origin visitors). Russia is missing for similar reasons after access restrictions. India is the fastest-growing region in the cohort in absolute volume, compounding at over 30% monthly in Q1 2026 from a small base, though this is across only a handful of sites with substantial Indian audiences.

ChatGPT conversion rate compared to other channels

Traffic share is half the question. The other half is what that traffic does once it lands. Across the 118 B2B SaaS sites in the cohort, conversion to a Stripe payment by channel:

ChatGPT-attributed sessions on B2B SaaS pages convert at roughly 1.9x the rate of Google organic sessions on the same landing pages. The pattern holds across nearly every sub-vertical in the SaaS slice. On ecommerce the pattern is different:

Google organic still leads ChatGPT on ecommerce conversion, by a small margin. The intent-quality story is real for research-heavy verticals (SaaS, agencies, services) and largely absent on impulse-purchase verticals (apparel, supplements, home goods). The chatgpt vs google traffic quality deep-dive walks the conversion-side numbers in more detail. The conversion gap is also why operators who run mixed traffic (both SaaS and ecommerce inside the same company) tend to see contradictory signals when they look at blended numbers, which is why we cut the data by vertical here.

ChatGPT versus Google organic: closing the ratio

Most operators benchmark ChatGPT against Google because Google is the channel ChatGPT is most plausibly displacing. Here is the ratio of Google organic sessions to ChatGPT-attributed sessions on the median cohort site, across the trend window.

The ratio compressed from 47 to 1 down to 12 to 1 across the 24-month window. The compression rate has slowed somewhat in the last six months as ChatGPT growth moderated from the early-2024 step changes and Google organic continued to compound at low single digits. At the trailing 6-month rate, the ratio would reach 5 to 1 by mid-2027 and 2 to 1 by mid-2028, but I expect moderation in both rates before then. Anyone forecasting parity (1 to 1) before 2028 is in my view extrapolating too aggressively.

Worth noting: the ratio is much smaller on certain sub-verticals already. In security and developer tools, the ratio is closer to 5 to 1 on the median site as of May 2026. In local services, the ratio is still over 60 to 1. The averaged 12 to 1 hides enormous variance and should not be read as a universal benchmark.

The compression in the ratio is the structural reason ChatGPT traffic has become a budget-allocation conversation rather than a curiosity. At 47 to 1, ChatGPT was a rounding error you could ignore. At 12 to 1, it is large enough that an operator under-counting by 71% (via GA4 default attribution) is materially mis-allocating spend. By the time the ratio reaches 5 to 1, the channel is too large to ignore at the budget level even if your dashboards are not catching it.

Where ChatGPT traffic lands on your site

A surprising structural pattern from the cohort: ChatGPT traffic lands deep, not on your homepage. Distribution of ChatGPT-attributed sessions by landing-page type, May 2026.

ChatGPT lands on a deep page 71% of the time versus Google organic at 47% and Direct at 13%. The reason is mechanical: ChatGPT cites specific passages from specific pages, and the link in the citation goes to that page, not to the homepage. If your homepage is your conversion architecture, you are designing for an audience pattern that ChatGPT does not match.

The conversion-rate consequence is large and worth quantifying. Across the cohort, conversion rate on AI-engine landing-page-type cohorts:

The 0.8% versus 3.1% gap on deep blog pages with and without an inline CTA is the single largest conversion-rate lever in the dataset that does not involve a paywall. Most SMB blog templates are designed for an era when deep-page entries were rare. They are not rare anymore, and the conversion architecture has not caught up. The mechanical fix is small: add one inline CTA, ideally above the fold of the blog post, with context that matches the article topic. The leverage per hour is the highest single conversion finding in the report.

Cross-checks against public benchmarks

This section is where I make the comparisons that should let you stress-test our numbers against datasets you might already trust.

The most interesting divergence is with eMarketer's 2028 forecast. If our trailing 24-month CMGR of 11.2% holds even half as long as it has, ChatGPT-attributed traffic share would cross 10% of total cohort sessions well before 2028. Whether this means ChatGPT specifically displaces 10% of Google share is a different question, because some of the ChatGPT growth comes from net-new query volume rather than displacement. But the 10%-by-2028 number looks conservative against our trajectory.

The most directly comparable third-party dataset is Profound's 2026 benchmark index, which measures AI citation share by vertical across a larger panel than ours but at the citation-appearance level rather than the session level. Their rank order of verticals (developer tools and security at top, local services at bottom) matches ours. The magnitudes are not comparable because they measure citation presence and we measure click-through and conversion.

We are also directionally consistent with Cloudflare's analysis of AI crawler behavior and the AI insights dashboard on Cloudflare Radar, both of which show AI bots compounding crawl volume faster than human user growth. Our cohort sees GPTBot, OAI-SearchBot, ClaudeBot, and PerplexityBot collectively at roughly 5% of total bot traffic in Q1 2026, within Cloudflare's reported 5-6% range.

For the academic side, the Princeton GEO paper (Aggarwal et al., 2024) is the most-cited primary research on what makes content cited by AI engines. The structural levers they identified (citations, statistics, quotations, answer-shaped passages) match what we see correlating with higher ChatGPT traffic share at the per-page level inside the cohort. The Anthropic engineering blog on citation generation describes the citation mechanics in Claude's web-search mode, which extends to similar principles across AI engines.

Why the gap between our numbers and yours might be larger

This section is the operator's gut-check. If your own GA4 says ChatGPT is 0.4% of your traffic and our cohort median is 3.4%, the gap is 8.5x. Some of that is the GA4 measurement gap (roughly 3.5x on the median). The rest is differences between your site and the cohort median.

Adjust the cohort median by these factors to estimate your own expected share. A new (6-month-old) Spanish-language ecommerce site that blocked GPTBot would be expected to sit well below cohort median. An 18-month-old English-language developer-tools blog with deep technical content and a clean entity profile would be expected to sit well above. Without an adjustment for these factors, comparing your raw number against the cohort median will mislead.

The single most common adjustment we see in operator conversations is the GPTBot robots.txt block. Many sites blocked GPTBot in 2024 during the initial AI-crawler concern wave, then forgot. Two years later they wonder why their AI traffic is below benchmark. The block does what it says: prevents OpenAI's crawler from reading the page, which prevents the page from being cited, which prevents the click-through. The llms-txt revenue impact write-up walks the more nuanced version of this question.

Operator implications: what to do with these numbers

Five concrete implications, in rough order of leverage.

1. Multiply your GA4 ChatGPT number by roughly 3.5x to estimate your actual share. The cohort multiplier is stable across percentiles. If your GA4 says 0.8%, your real share is probably 2.5 to 3%. Run that against your channel-attribution dashboard and see if your budget allocation needs to shift.

2. Stop benchmarking your site against an industry-blind average. "AI traffic is 5% of total" is a meaningless statement if you sell local plumbing. Pick the vertical row in the table above that matches your business and benchmark against that.

3. Add an inline CTA to every deep blog page. The 3.1% versus 0.8% conversion gap is the highest-leverage mechanical fix in the dataset. ChatGPT lands on deep pages 71% of the time, and most blog templates were designed before that was the dominant entry pattern. Two hours of editorial work for a typical 30-post blog.

4. If you sell B2B SaaS, invest in ChatGPT visibility as a real channel. The combination of growing volume (11.2% monthly), higher conversion rate (1.9x Google organic on SaaS), and higher first-month value ($44 vs $28) makes it the most attractive new channel in the cohort. The ai visibility score and revenue attribution features are designed for this measurement loop specifically.

5. If you sell ecommerce, instrument but do not over-rotate. Google organic still leads ecommerce conversion rate (2.1% vs ChatGPT 1.7%). AI is a real and growing share, but the existing SEO playbook still does most of the heavy lifting. Worth measuring; not worth re-architecting around.

For a longer treatment of the strategic implications across the four GEO evidence layers, the companion piece is does GEO actually drive revenue. For the strategic split between SEO and AEO investment by business type, aeo vs seo 2026 lays out an opinionated 80/20.

Limitations and caveats

I want this section to be longer than the average data study because the integrity of the numbers depends on the reader understanding what they can and cannot infer.

1. Sample is biased toward Stripe-native businesses. Companies on Recurly, Chargebee, Paddle, or non-Stripe rails are not in the dataset. If those processors over- or under-index any specific vertical or geography, our numbers will skew accordingly.

2. Sample skews bootstrapped SMB. Largest site is roughly $250k MRR. Enterprise patterns (longer sales cycles, sales-assist motion, MQL-to-SQL workflows) are out of scope and would dilute the headline numbers.

3. Sample self-selected. Sites joined Attrifast typically because they suspected they had un-attributed AI traffic. That selection bias likely inflates the AI-share numbers versus a true random SMB sample by 10 to 20%. The "3.4% of sessions are ChatGPT" finding for the median Attrifast site is probably 10 to 20% lower for a randomly-selected SMB.

4. Behavioral fingerprinting has a roughly 20% noise floor. The suspected-AI recovery layer validates at roughly 80% precision against post-purchase survey ground truth. We label these sessions as "suspected" in the raw data and only roll them into the ChatGPT bucket where the cohort numbers are robust to that noise.

5. Geographic skew toward US (62% of sessions) and EU (24%). APAC is under-represented; Latin America is barely visible. The geographic cuts should be read as cohort-specific.

6. Window is a single rolling 30-day slice for the headline. The 24-month trend section catches more seasonality, but Black Friday peaks and Q1 SaaS budget resets are not fully normalized in the headline number.

7. Per-engine attribution is single-touch. A visitor who saw a Claude citation on Monday, a Perplexity citation on Wednesday, and converted via ChatGPT on Friday is credited entirely to ChatGPT in this report. Multi-touch attribution across AI engines is on our roadmap, not in the dataset.

8. We are the vendor. I have a structural incentive to publish numbers that make AI attribution look important. I have tried to balance that with the limitations above and the methodology specificity, and I welcome challenge on any number here.

What is next

This benchmark will re-run quarterly. Next publish target is 2026-08-15 for the Q2 update. Three planned changes:

• Add APAC sites to the geographic mix as the cohort expands. Targeting 30+ APAC sites for the Q2 release.
• Add per-engine multi-touch for the AI-traffic attribution so we can credit "ChatGPT brought them, Perplexity converted them" patterns properly.
• Add seasonal-adjusted comparisons against the same trailing window in prior years, now that we have a 24-month baseline.

If you want to see your own site's ChatGPT split and confirm or refute the 3.4% figure on your own dataset, Attrifast runs the full server-side attribution pipeline described in the methodology section, joins to Stripe at the webhook layer, and ships the dashboard at $29/month. Start the free trial or read more about the revenue attribution architecture.

If you cite this benchmark in your own writing, please link to this page as the source. The numbers are ours and the methodology is published above. If your dataset disagrees with ours, I would rather hear it than not.

FAQ

How much traffic comes from ChatGPT in 2026 on a typical website?

Across the roughly 200 Stripe-connected sites in the Attrifast cohort, the median site received 3.4% of measured sessions from ChatGPT during the rolling 30 days ending May 15, 2026. The interquartile range is 1.1% to 6.8%. B2B SaaS sites skew highest at a median 5.2%, DTC apparel ecommerce sits at 4.1%, while local services sites sit at 0.8%. These figures are server-side measurements: roughly 71% of ChatGPT-attributed traffic on these sites arrived with no referrer at all and was recovered via referrer fingerprinting and behavioral heuristics. Sites still relying on GA4 default attribution see only the visible 29%, which is why most operators we talk to estimate ChatGPT at well under 1% of their traffic when the real number is 3 to 5 times higher.

What percentage of total web traffic globally comes from ChatGPT?

The honest answer is that nobody has a clean global number, because ChatGPT-referred clicks largely arrive with no referrer header and most public panels under-count them. Similarweb's chatbot tracker shows chatgpt.com itself drew roughly 5.2 billion visits in March 2026, putting it among the top 10 most-visited sites on the open web. But that measures traffic to ChatGPT, not traffic ChatGPT sends out. Cloudflare Radar's AI insights peg AI bots at roughly 5 to 6% of total bot traffic in Q1 2026, which is a crawl metric not a click metric. Our best read combining the Attrifast cohort with public benchmarks: ChatGPT currently routes between 1.5% and 4% of total web referral clicks globally as of mid-2026, depending heavily on how you classify the no-referrer fraction. Anyone giving you a precise single global number with one decimal place is likely making it up.

Has ChatGPT traffic share grown month over month in 2024 and 2025?

Yes, materially and consistently. Cohort-aggregated ChatGPT-attributed sessions grew at a compounded monthly rate of 11.2% across the 24-month window from May 2024 to May 2026 on the subset of cohort sites with data going that far back. The growth has not been linear. Q3 2024 saw a step change when ChatGPT Search rolled out broadly; March through May 2025 saw another step change when ChatGPT app integrations expanded across mobile platforms; Q1 2026 saw the largest single quarter at +37% as 800-million-weekly-active-user data was confirmed and shopping queries entered the engine. For most sites in the cohort, ChatGPT traffic doubled between mid-2024 and late-2025, and doubled again between late-2025 and mid-2026.

How does ChatGPT traffic share differ between SaaS, ecommerce, and local services?

The vertical spread is wide enough that any single industry-blind average is misleading. In the Attrifast cohort, May 2026 medians by vertical sit roughly at: B2B SaaS 5.2%, developer tools 6.7%, security and infrastructure 7.1%, DTC apparel 4.1%, beauty and personal care 3.2%, supplements 2.9%, home goods 2.4%, consumer electronics 3.8%, financial services 2.1%, healthcare publishers 2.7%, agencies and consultancies 4.3%, marketplaces 1.9%, local services 0.8%, and brick-and-mortar retail 0.4%. The pattern is consistent across our trend window: verticals where buyers research before purchase see materially more ChatGPT traffic than verticals where purchases are local, habitual, or impulse-driven.

What share of ChatGPT traffic does GA4 actually report correctly?

On the median Attrifast cohort site, GA4 correctly attributed 29% of ChatGPT-referred sessions to a chatgpt.com or chat.openai.com referrer or a utm_source=chatgpt parameter during May 2026. The remaining 71% landed in Direct/(none) because the AI client stripped the Referer header. The gap is largest on mobile (where in-app browsers dominate) and on the ChatGPT desktop app (which routes clicks through an OS handoff that omits the referrer). The gap is smallest on the Perplexity web UI, which preserves referrer roughly 50 to 70% of the time. The same gap exists on Claude, Gemini, and AI Overviews to varying degrees, which is why first-party server-side attribution is the only durable measurement layer.

Which industries see the highest ChatGPT referral traffic in 2026?

Developer tools, security and infrastructure, and B2B SaaS sit at the top of the vertical ranking, all between 5 and 8% of measured sessions from ChatGPT in May 2026 across the Attrifast cohort. These verticals share two structural features: their buyers research extensively before purchasing, and their content gets cited heavily in ChatGPT answers to vendor-evaluation queries (best secrets manager for Node.js, alternatives to Datadog, how to set up SSO with Okta). At the other end of the spectrum, local services (plumbers, dentists, lawyers) sit below 1% because the underlying query intent rarely surfaces an AI answer in the first place.

Does ChatGPT traffic share vary by site size?

Yes, and the pattern is the inverse of what most operators expect. Larger sites in the cohort (those with more than 500,000 monthly sessions) see ChatGPT share around 2.1% on average, while smaller sites (under 50,000 monthly sessions) see 4.3%. The reason is mathematical: larger sites have much heavier branded-search and direct type-in volume, which dilutes the AI share even though the absolute AI volume is also higher. Smaller sites, especially niche blogs and bootstrapped SaaS, have a more discovery-heavy traffic mix, so AI shows up as a larger fraction. In absolute session counts, the largest 10 sites in the cohort still receive more ChatGPT visits than the smallest 100 combined.

How does ChatGPT traffic split between United States, Europe, and Asia?

Geographic distribution is uneven enough to matter for budget conversations. Of all ChatGPT-attributed sessions in the cohort, 64% landed from US visitors, 22% from EU and UK, 14% from rest-of-world (including Canada, Australia, and the visible APAC fraction). On a per-capita-internet-user basis, the US is roughly 1.7 times more ChatGPT-active than the EU and roughly 3 times more active than APAC excluding India. India is the fastest-growing region in the cohort, growing at over 30% monthly during Q1 2026 from a small base. China is essentially absent from the dataset because ChatGPT is not accessible there. Pew Research and Similarweb both confirm the directional US-over-EU adoption gap, though our numbers are cohort-specific and should not be read as definitive global shares.

How does ChatGPT traffic compare to Google organic traffic for typical sites?

Google organic still dwarfs ChatGPT in raw session volume on the median cohort site: Google sends roughly 9 to 15 times more sessions than ChatGPT does. But the gap has been narrowing for 24 months at a measurable rate. In May 2024, the typical cohort site received roughly 47 Google-organic sessions for every ChatGPT-attributed session. In May 2026, the same ratio sits at roughly 12 to 1. At current compound monthly growth rates of 11.2% for ChatGPT and 1.1% for Google organic, the ratio would cross 5 to 1 by mid-2027 and 2 to 1 by mid-2028, though we expect both rates to moderate before then. ChatGPT also delivers higher conversion rates per visit on B2B SaaS pages: about 2.5% Stripe-payment conversion versus 1.4% for Google organic on the same landing pages.

Is ChatGPT traffic actually high quality, or just inflated by curiosity clicks?

It is materially higher quality than Google organic on B2B SaaS pages, and comparable to Google organic on ecommerce pages. ChatGPT-attributed sessions on the cohort's 118 B2B SaaS sites converted to a Stripe payment at 2.5% versus 1.4% for Google organic on the same landing pages, a 1.8x lift. The first-month subscription value for ChatGPT-sourced SaaS customers was $44.10 versus $28.70 for Google organic, a 54% premium. On ecommerce the conversion pattern flips slightly: Google organic converts at 2.1% versus ChatGPT at 1.7%, because ecommerce purchases are often impulse-driven and ChatGPT users arrive informed but not necessarily ready to buy a specific SKU. The quality story is real for research-heavy verticals; it is more muted for impulse verticals.

What is the most reliable way to measure ChatGPT traffic on my own site?

Three layers, in order of recovery rate. Layer one: a custom GA4 channel grouping with regex rules for chatgpt.com, chat.openai.com, oai.com, and utm_source=chatgpt recovers the visible 29% that arrives with a referrer or UTM. Layer two: a server-side referrer-enrichment step (a small Edge worker or a tool like Attrifast) captures the Referer before any client-side processing and applies User-Agent matching for OAI-SearchBot to the citation-fetch leg. Layer three: behavioral fingerprinting on no-referrer deep-page entries with geo and time-of-day clustering recovers another 30 to 40% of the dark fraction. Combined, the three layers get you to 75 to 90% recovery. Without server-side, you are reading the visible 29% and missing two-thirds of the channel. The detailed setup walkthrough is in our track-chatgpt-traffic guide.

How will ChatGPT traffic share change through 2027 and 2028?

Predictions are unreliable but extrapolations from the trend window are useful as scenario boundaries. If the trailing 24-month compound monthly growth rate of 11.2% holds, ChatGPT-attributed traffic share on the median cohort site reaches 7.5% by May 2027 and 16% by May 2028. If the growth rate moderates to half that (5.6% monthly), the median sits at 5.4% by May 2027 and 8.5% by May 2028. The likeliest scenario in our estimation is moderation: ChatGPT user growth is showing signs of slowing from the 800-million-weekly-active milestone, but click-through from answers is rising as the engine improves citation quality. Plan for a 5 to 10% steady-state ChatGPT traffic share by 2028 on a research-heavy site, lower on local-services or impulse-purchase sites.

References

1. Reuters — OpenAI ChatGPT weekly active users coverage, 2025-2026
2. The Verge — ChatGPT daily message volume, December 2024
3. Similarweb — AI chatbot traffic research and chatgpt.com monthly visits
4. Cloudflare Radar — AI insights dashboard
5. Search Engine Land — Google AI Overviews tracker
6. BrightEdge — Google AI Overviews expansion research
7. Backlinko — AI Overviews and ChatGPT statistics
8. Semrush — AI search traffic research
9. Pew Research Center — Americans' use of AI tools
10. Adobe Digital Insights — 2024 online shopping report
11. McKinsey — State of AI 2024 report
12. eMarketer — AI search forecast
13. seoClarity — AI Overviews impact research
14. Anthropic — ClaudeBot and Claude web crawling policy
15. OpenAI — GPTBot and OAI-SearchBot documentation
16. OpenAI — Introducing ChatGPT Search announcement
17. Princeton GEO paper — Generative Engine Optimization, Aggarwal et al.
18. Anthropic Engineering — Citation generation in Claude
19. Reuters — OpenAI valuation and growth coverage
20. The Wall Street Journal — ChatGPT user growth and moderation reporting
21. The Information — OpenAI product and growth coverage
22. Profound — Profound Index, 2026 benchmark report
23. Otterly — AI visibility tracking benchmarks
24. Plausible — ChatGPT referrer pass-through measurement
25. Stripe — Webhook delivery and idempotency
26. Google Analytics — Default channel grouping documentation
27. Cloudflare Blog — AI crawler analysis and Radar updates
28. StatCounter — Search engine market share
29. HubSpot — State of Marketing 2025
30. CNIL — Cookies and tracking devices guidance