AI search traffic: live data and measurement.
View daily panel data for ChatGPT, Claude, Gemini, Perplexity and other AI sources, or use the GA4 and PostHog workflow to measure identifiable AI traffic to your own site.
Index at 76% of peak, +47.8% over 30 days
Claude+400.6%ChatGPT+44.4%
Perplexity-22.4%Find AI referral traffic in your own GA4
Start with GA4's current AI Assistants channel, then audit the underlying sources and keep traffic that analytics cannot attribute in a separate evidence layer. The public panel's source mix is a panel benchmark, not global market share.
Reference: Google’s default channel definitions, Google’s GA4 landing page report and PostHog web analytics.
Traffic acquisition report
- Open Reports, then Acquisition, then Traffic acquisition. Check Session default channel group = AI Assistants first.
- Set the primary dimension to Session source / medium and compare the last complete month with the previous month.
- Add a Session source Matches regex filter to audit recognized assistant hostnames. Use the broad collector expression only when you can inspect aliases.
- Add Landing page + query string as a secondary dimension. Review Sessions, Engaged sessions, Key events, and Total revenue.
^(([a-z0-9-]+\.)*chatgpt\.com|chat\.openai\.com|openai\.com|chatgpt|openai|(www\.)?perplexity\.ai|perplexity|(www\.)?claude\.ai|claude|anthropic|gemini\.google\.com|gemini|bard|copilot\.microsoft\.com|copilot|bing-chat|deepseek\.com|deepseek|grok\.com|x\.ai|grok|you\.com|phind\.com|phind|poe\.com|poe|meta\.ai|meta-ai)$Use a Matches regex filter on Session source. Keep the anchors so tagged campaigns do not match by accident.
Use the conservative audit regex
^((www\.)?chatgpt\.com|chat\.openai\.com|(www\.)?perplexity\.ai|(www\.)?claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|(www\.)?grok\.com|(www\.)?deepseek\.com)$Use a Matches regex filter on Session source. Keep the anchors so tagged campaigns do not match by accident.
Free-form exploration
Views align with this index's GA4 screen-page-view measure. Sessions and key events answer a different question: what happened after the click.
Measure AI landing pages and conversions in PostHog
Use the same source domains, but measure the entrance page and the conversion event in one funnel. This keeps referral volume separate from what visitors did after the click.
Referral and landing-page report
- Open Web analytics, set the last 30 days, and review Acquisition or Referrers for the AI source domains in the table below.
- Compare visitors, sessions, views, and conversion goals by referring domain. Keep broad aliases separate until you have checked the raw values.
- Open Product analytics and create a funnel. Make the first step
$pageview, filtered where$referring_domainmatches the AI source regex. - Use your real signup, lead, checkout, or purchase event as step two. Break down the entrance
$pageviewby$pathnameto compare landing pages.
PostHog fields to use
Filter the entrance event, not every funnel step. A conversion event may not carry the original referrer property.
Build one auditable AI measurement system
The downloadable CSV turns the setup into an owner-by-owner checklist for GA4, product events, CRM fields, the monthly report, and dark-traffic limits. It is spreadsheet-ready and contains no customer data.
Download the measurement kitCSV, 22 implementation and reporting rows, updated 18 August 2026
1. Channel
Use GA4 Session default channel group = AI Assistants. Audit recognized hostnames with the conservative regex.
2. Landing report
Break out Session source / medium by Landing page + query string. Add sessions, engaged sessions, key events, and revenue.
3. Event contract
Join the entrance page to your real signup, lead, checkout, or purchase event. Test the event before it enters the report.
4. CRM contract
Store first and last AI source, landing page, self-reported discovery source, research theme, and attribution confidence.
5. Monthly report
Keep visible referrals, answer visibility, crawler activity, and sourced or assisted pipeline in separate evidence rows.
Four panels, four different claims
Do not force answer visibility, human referrals, bot requests, and revenue into one total. Give each panel its own source and confidence label.
Visible referrals
GA4 AI Assistants sessions, landing pages, key events, observed revenue.
Answer influence
Prompt presence, citations, owned and competitor sources. Directional unless joined to a journey.
Crawler access
OAI-SearchBot and other AI bot requests from server or CDN logs. Never human traffic.
Revenue ledger
Sourced, assisted, and directional rows kept separate with evidence links.
Illustrative monthly row
640 AI sessions → 32 lead events → 6 sourced opportunities
That is a 5.0% session-to-lead rate. If the six opportunities total £72,000, call it AI-sourced pipeline only when each record has a captured referral or matched event. The example is fictional and does not estimate dark influence.
Dark-traffic rule
Unknown is a valid result
Direct, branded search, and later organic visits can contain AI-influenced journeys, but a missing referrer does not prove one. Use self-reporting, CRM timing, and controlled tests as support, not as a relabeling rule.
| Platform | Known hostname or source | Current collector aliases | What to check |
|---|---|---|---|
| ChatGPT | chatgpt.com, chat.openai.com | chatgpt, openai, openai.com | Review broad OpenAI aliases before grouping them with ChatGPT. |
| Perplexity | perplexity.ai | perplexity | A missing referrer can move an app click into Direct. |
| Claude | claude.ai | claude, anthropic | The broad Anthropic alias can include tagged campaigns. |
| Gemini | gemini.google.com | gemini, bard | Google AI Overviews and some AI Mode clicks can remain google / organic. |
| Copilot | copilot.microsoft.com | copilot, bing-chat | Do not group ordinary bing / organic traffic into Copilot. |
| DeepSeek | deepseek.com | deepseek | Low-volume rows can be volatile over short periods. |
| Grok | grok.com, x.ai (frontend mapping) | grok | The panel collector currently recognizes values containing grok. Its exact browser hostname is not verified, and x.com is deliberately excluded from the filter. |
| Other supported sources | you.com, phind.com, poe.com, meta.ai | phind, poe, meta-ai | The collector recognizes you.com, phind, poe, and meta-ai. meta.ai exists in its normalize map but not its current collection regex, so review it. |
Calculate your AI traffic share
Use GA4 Views for an apples-to-apples comparison with this public index. The index does not expose an industry baseline for AI views as a share of all website views.
Enter your last 30 days of views to create a baseline before connecting anything.
Keep blind spots and bots in separate reports
- In-app browsers, app-to-browser handoffs, redirects, privacy settings, and referrer policies can strip the source. Those visits can appear as Direct or have an empty referrer.
- Gemini referrals can show as gemini.google.com, but Google AI Overviews and some AI Mode clicks can remain google / organic. Standard web analytics cannot reliably isolate them from conventional Search clicks.
- Treat dark AI traffic as an unmeasured remainder, not a number to relabel. Report identifiable AI referrals as the observed minimum and keep Direct and unassigned movement as a separate watchlist.
- A source domain identifies the sender when the referrer survives. It does not reveal the prompt, answer, citation, or citation position that caused the click.
- AI crawlers are bots, not human referrals. Keep crawler activity in server or CDN logs, and exclude known bots and internal traffic from visit and conversion reporting.
Connect the same workflow to Trakkr
Trakkr’s Visitors view connects Google Analytics, categorizes identifiable AI referrers, and reports landing pages and conversions alongside visibility data.
See Trakkr’s traffic analyticsAI referral traffic by source
Visible sources rescaled so the highest visible point = 100%.
Aggregated screen page views from 5,107 anonymized GA4 properties with qualifying AI source data. Visible chart lines are rescaled to the highest visible point. Source mix is the share of identified AI referral page views over 30 days, not global market share.
Who sends the most AI traffic?
Share of identified AI referral page views, past 30 days
ChatGPTChatGPTClaudeClaude
GeminiGeminiPerplexityPerplexity
GrokGrok
DeepSeekDeepSeekSingle-source deep dives
Which sectors changed fastest?
30-day AI referral index change across 6 broad groups
Provenance
What the current panel supports
The panel supports normalized movement and visible-referrer source mix. It does not support a claim about global AI market share, total AI clicks, conversions, or citation-to-click rate.
5,107
Qualifying GA4 properties
Distinct connected GA4 properties contributing at least one detected AI referral in the aggregation window.
89.8%
ChatGPT share of visible AI referrals
ChatGPT-attributed pageviews divided by pageviews attributed to all detected AI referral sources in the latest 30-day window.
47.8%
AI referral index, 30-day change
Change in smoothed AI-attributed pageviews across the panel between the latest 30-day period and the preceding 30-day period.
Methodology
Method and dataset coverage
AI Search Traffic Index
Verified 2026-08-18
- Records
- 4,928 qualifying GA4 properties and 443 daily observations
- Dates
- 2025-06-01 to 2026-08-17
- Sampling
- A changing panel of connected GA4 properties. Source rows require at least 3 properties and source-by-industry rows require at least 25.
- Model scope
- Traffic sources are detected from GA4 sessionSource values. This measures visible referrals, not model usage or total AI search volume.
- Geography
- Geography and device are not published.
- Privacy
- Public downloads contain normalized trends and aggregate source shares only. Property identifiers and absolute panel volumes are excluded.
- Not included
- Property identities, Absolute public pageview totals, Landing pages, Conversions, Geography, Device, Citation-to-click joins
Metric ledger
Open any metric to inspect its calculation, sample, scope, and limits.
- Definition
- Distinct connected GA4 properties contributing at least one detected AI referral in the aggregation window.
- Numerator
- Not applicable
- Denominator
- Not applicable
- Sample
- 458 daily observations
- Dates
- 2025-06-01 to 2026-09-01
- Model scope
- Traffic sources are detected from GA4 sessionSource values. This measures visible referrals, not model usage or total AI search volume.
- Geography
- Geography and device are not published.
- Limits and confidence
- The connected-property panel changes over time and is not representative of the whole web. Direct distinct count from the verified public aggregate.
- Review
- Verified 2026-09-01. Reviewed by Mack Grenfell.
- Source files
- Open source Download data
Download the AI traffic CSV files
Aggregate, privacy-filtered files. No customer identities, prompts, account IDs, raw URLs, or absolute traffic volumes.
Definitions
- Visible AI referral
- A GA4 sessionSource value matching a known AI assistant referrer. Visits without a visible referrer cannot be attributed.
- Traffic index
- A smoothed series normalized to a peak value. It shows movement, not the public panel’s absolute traffic volume.
- Source mix
- A source’s share of detected AI-attributed pageviews in the selected panel and period. It is not global AI market share.
- Verified snapshot
- A point-in-time value regenerated from an attributable public aggregate and checked against its numerator, denominator, and coverage metadata.
Update history
Flagship evidence refresh
Rebuilt the three public research surfaces from 48,422,016 citation appearances and 4,928 qualifying GA4 properties.
Markdown experiment added
Added the randomized 9,033-page crawler experiment as a reviewed State of AI Search dataset.
Longitudinal citation study added
Added citation persistence and volatility findings with an explicit 177-day citation observation window.
Corrections
Citation headline and Wikipedia share
Removed the stale 1.3M citation and 17% Wikipedia claims. The page now resolves current values from the verified Citation Index and records the snapshot date.
Traffic loading values
Removed invented placeholder growth and source-share values. Loading and failed requests now render as unavailable.
Live-data description
Replaced “every number is live” with dataset-level status because several reviewed studies are point-in-time snapshots.
What this dataset can answer
Latest 30-day share of pageviews attributed to detected AI referrers in the qualifying GA4 panel.
Smoothed daily aggregate and source indexes. Published values are normalized, not absolute traffic volume.
Available for six broad groups. Industry assignment and changing panel composition limit comparisons.
Distinct contributing property counts are published for the aggregate, each source, and each broad industry.
The public index deliberately withholds total pageview counts to reduce exposure of the underlying panel.
The aggregate does not publish engagement rate, revenue, leads, or conversion events.
Landing paths and page classifications are not included in the public panel.
Country, region, device, and browser breakdowns are not published.
There is no validated row-level join between Citation Index responses and GA4 referral sessions.
If a source is rising
Confirm the change in your own analytics, then inspect which landing pages and conversion events contributed.
If a source is declining
Check referrer classification and your own time series before treating a panel decline as a change in user demand.
If you are benchmarking
Track your own source mix weekly and compare it with the panel trend without treating the panel as global market share.
What is the AI search traffic index?
The AI Search Traffic Index is a daily-updated panel of identifiable AI referral page views. It tracks GA4 screen page views on sessions attributed to ChatGPT, Claude, Gemini, Perplexity, Copilot, DeepSeek, Grok, You.com, Poe, Phind and other supported source values across 5,107 anonymized properties.
Important: this index measures referral traffic, not citation share. A source can drive high referrals while your brand still has weak citation coverage across recommendation prompts.
The multi-source chart rescales the visible lines to the highest visible point. Source mix is calculated as each platform's percentage of identified AI referral page views over the trailing 30-day period. The page refreshes when the aggregate pipeline publishes a new snapshot.
Why track AI search traffic?
Referral traffic shows which identifiable AI sources sent people to a website. It does not measure prompts, citations shown without a click, or visits whose referrer was stripped. Marketers, SEOs, and growth teams can use it alongside citation and conversion data to monitor channel movement without treating one measure as the whole visibility picture.
Keep three planes separate: human referrals in analytics, possible no-click or dark influence, and crawler requests in server logs. OpenAI documents OAI-SearchBot as the crawler used to surface sites in ChatGPT search results, while ChatGPT referral URLs include utm_source=chatgpt.com. A crawler request is not a visitor, and an unattributed visit is not proof of AI influence.
Data sources and methodology
The pipeline filters GA4 sessionSource against the published source values and aliases, then aggregates the screenPageViews metric from matching sessions. All property-level data is anonymized before aggregation. The 5,107 count is the number of distinct GA4 properties with qualifying AI-source data in the current panel, not a denominator for every property connected to Trakkr.
Which AI search engine sends the most traffic to websites?
Use the current source ranking on this page. It orders the panel by each platform's share of identifiable AI referral page views over the latest 30-day period, so the answer can change as the data changes.
What does the AI Search Traffic Index measure?
The index aggregates GA4 screen page views on sessions whose source matches a current AI platform source or alias. It reports normalized trends and source mix across 4,928 properties with qualifying AI referral data, without exposing absolute panel volume.
Is AI referral traffic growing?
Read the current 7-day and 30-day changes shown above rather than assuming a permanent growth rate. The direction can change by period and platform, and short windows are sensitive to source mix and panel changes.
Does the index show AI traffic as a percentage of all website traffic?
No. The source ranking shows each platform's share of identified AI referral page views, not AI traffic as a share of all website traffic. Use the calculator and GA4 recipe on this page to establish that baseline for your own site.
How do you track traffic from different AI platforms?
The panel filters GA4 sessionSource using current platform domains and aliases, then aggregates screen page views from matching sessions across 4,928 anonymized properties. The source table on this page publishes the operational values and flags broad aliases that need review.
How is source mix calculated?
Source mix is each platform's percentage of all identified AI referral page views in the panel over the most recent 30-day period. It is not a share of all web traffic or AI usage.
Why can an AI click appear as Direct in GA4?
In-app browsers, app-to-browser handoffs, redirects, privacy settings, and referrer policies can remove the source before GA4 records the visit. Those clicks can appear as Direct, and standard GA4 data cannot reliably reconstruct the missing source.
Can GA4 separate Gemini and Google AI Overview traffic?
GA4 can identify a referral such as gemini.google.com when the referrer is preserved. Google AI Overviews and some AI Mode clicks can remain google / organic, so standard GA4 cannot reliably split those visits from conventional Google organic traffic.
Does an AI source row reveal the prompt or citation?
No. A source row identifies the sender when the referrer survives. It does not contain the prompt, answer, cited passage, or citation position that caused the click.
How can I check my AI search traffic in Google Analytics?
Open Reports, Acquisition, then Traffic acquisition. Use Session source / medium, add the anchored source regex published on this page, compare the last 30 days with the previous period, and break the result down by Landing page + query string.
How can I measure AI referral traffic in PostHog?
Start in Web analytics and review Acquisition or Referrers for known AI domains. Then create a funnel whose entrance pageview is filtered by $referring_domain, use a real conversion event as step two, and break the entrance pageview down by $pathname to compare landing pages.
How can I track AI traffic to my own website?
Start with the GA4 or PostHog workflow on this page. If you connect Google Analytics to Trakkr, the Visitors view categorizes identifiable AI referrers and reports landing pages and conversions alongside the public traffic panel.
Track AI search traffic to your website
Connect Google Analytics to see how much traffic ChatGPT, Claude, Gemini and Perplexity send to your site.