Machines on the server. Humans in GA4. One picture.
Viz sees every crawler on your server, and every purchase, booking and lead the visitors AI sends go on to make. GA4 sees the part in between: how long a human stayed and what they engaged with, and it is blind to every bot. Connect them and you get the full picture: what the machines read, what that reading sent back, and how those visitors behaved before they bought.
AI traffic happens in two places. You're measuring one.
When ChatGPT answers a question with your content, two measurable events occur. First, a crawler fetched your page - visible only at the server. Later, maybe, a human clicks through to you - visible only in browser analytics. Most sites measure neither; the rest measure one and call it the whole story. The interesting number is the ratio between the two: how much crawling does one real visitor cost you? That ratio is the economics of your content, and computing it requires both halves.
Machines read
GPTBot, ClaudeBot, and PerplexityBot fetch your pages. Viz logs and classifies every one of those requests inside WordPress - browse the bot directory to see who they are.
Answers get given
Somewhere on another company's servers, your content becomes part of an answer. Some answers cite you. Paid plans track how the major engines describe you - see AI Visibility.
Humans (sometimes) arrive
A reader clicks the citation. Viz logs that referral visit at the server the moment it lands and credits whatever they go on to buy, book or submit to the assistant that sent them - see AI Conversions. If GA4 is connected, the session and its engagement join the same dashboard as the crawl that started the chain.
The funnel, with sessions in the middle.
Viz reports the whole ladder as one funnel: crawl, answer, visit, convert, act - the AI crawls, the referral visits assistants send back, the conversions those visitors make and the agents that come to act, all captured server-side by the plugin itself, with the rate between each stage and no Google account at all. Linking GA4 fills in the layer between visit and conversion: sessions and engagement, so you can see how AI-referred visitors behave before they buy.
Sessions on top of visits
Viz already captures AI-assistant referral visits server-side, no Google connection needed. Linking GA4 adds the session layer: how long those visitors stayed and what they did, pulled from your own property.
Engagement, not just arrivals
AI referrals are often high-intent - the assistant already answered the easy part. See how long these sessions last and what they engage with, beside the conversions Viz already credits to each assistant on its own.
Search Console context
Classic search impressions and clicks alongside AI crawl volume, so you can watch whether AI answers are eating into queries you used to win in blue links.
One view, both halves
Machines on one side, humans on the other, in the same wp-admin dashboard. No more stitching a GA4 exploration to a hosting log by hand.
AI referrals undercount. Read them accordingly.
The web's referrer mechanism predates AI assistants, and it shows. A click from a browser tab on chatgpt.com usually announces itself; the same recommendation followed from a mobile app, a voice interface, or a copy-pasted link arrives with no referrer at all and books as direct traffic. Different assistants, different surfaces, different leakage - and none of it under your control. Any tool that hands you an “AI traffic” number without saying this is rounding a floor up to a total.
Attributed AI sessions are solid for direction and quality, soft on absolute volume. In practice that means watching three things together - attributed sessions trending in GA4, the engagement of those sessions against your site average, and retrieval-bot crawl volume on the server side as the channel's leading indicator. When crawls rise and attributed quality holds, the channel is growing even while part of it hides inside direct.
Read-only, aggregate, and reversible.
The integration reads reporting aggregates from Google's APIs - it does not add tracking to your pages, does not send your visitors' data anywhere, and can be disconnected at any time. Viz's own measurement stays where it always was: in your database, on your server, under your retention rules.
Integration FAQ.
Why do I need GA4 if Viz already logs every request?
You do not need it. Viz sees every machine that touches your server, captures the referral visits assistants send back, and records the purchases, bookings, leads and sign-ups those visitors make from your own commerce and form plugins - so the question "does being crawled by AI turn into customers?" is answered by the plugin alone. What a server cannot see is the middle of a visit: how long a human stayed, what they scrolled, which pages they engaged with. That is what GA4 is built for, and linking it adds that layer to the same picture.
What access does Viz get to my Google account?
Read-only API access to the GA4 property and Search Console site you choose. Viz reads aggregate reporting data; it does not write to your Google account, and Google OAuth tokens never touch your WordPress server.
How does Viz know a visit came from an AI assistant?
By referrer, at two layers. When someone clicks a link in ChatGPT or Perplexity, the browser typically reports chatgpt.com or perplexity.ai as the referring site. The plugin captures those referral visits server-side as they land - no Google connection needed - and, if GA4 is linked, Viz also segments the matching sessions out of your GA4 data. Some AI surfaces strip the referrer, so treat the number as a floor, not a ceiling - the limit is in the web platform itself.
Is the integration required to use Viz?
No. All server-side analytics - classification, dashboards, the request log, AI-assistant referral capture, AI Conversions attribution, and AI Control blocking - work with the plugin alone. The integration adds session depth (sessions and engagement) when you want it.
My AI-referred session count looks tiny. Is the channel not working?
Check the quality before judging the quantity. AI referral counts systematically undercount - apps and some assistant surfaces open links without a referrer, so a share of the channel books as direct. The sessions that do carry a referrer are your sample: if their engagement and conversion beat your site average, the channel is working and the true volume is higher than the chart shows. Compare the trend against your retrieval-bot crawl volume to see whether the pipeline that produces those referrals is growing.