PublishedVerifiedEvery 6 weeksSources6 namedAuthored bySquareRank Team
ChatGPT · § 1.1.4 · How-to
Measure ChatGPT Referral Traffic on Squarespace
An analysis of 446,000 sessions across AI-traffic-detection-equipped sites found that roughly 70% of AI referral traffic arrived without a referrer header and was categorised as Direct in GA45. ChatGPT's utm_source=chatgpt.com tag1 catches the rest, but conversational inline links remain untagged and mobile-app traffic strips referrers entirely. Squarespace Analytics, GA4, the AI Visibility panel, and a manual tracking spreadsheet each cover a different slice of the dark.
This leaf is the measurement layer for a Squarespace site that is doing the other four steps right (crawlers, schema, passage content, entity wiring). The honest framing: no single tool catches all ChatGPT traffic. The 2026 stack triangulates — the regex catches the tagged subset, the AI Visibility panel catches mention frequency, and the manual log catches the conversational citations that leave no trail.
§01The problem
The dark-traffic problem in 2026
ChatGPT visits arrive at your Squarespace site in three forms — tagged with utm_source=chatgpt.com (the visible third), untagged from a desktop browser but with a chatgpt.com referrer (a partially-visible third), and completely referrerless from mobile-app or privacy-stripped sources (the dark third). The shape changes month to month as OpenAI extends UTM coverage. The June 2025 update extended tagging to the 'More sources' panel, which closed part of the gap, but the conversational inline links — the largest single source of ChatGPT citations — remain untagged at time of writing.
The Clickport analysis5 of 446,000 sessions in early 2026 found that approximately 70% of AI-referral traffic landed as Direct in GA4 because the referrer header was missing or stripped. Mobile-app visits are the largest contributor — the ChatGPT iOS and Android apps open external links in a webview without passing a referrer header, so even tagged URLs lose half their attribution context. Ad blockers strip headers similarly, and so does the Brave browser's built-in AI summariser.
The honest 2026 framing for Squarespace owners: GA4 referral numbers are the floor, not the ceiling. A Lawrence Hitches analysis1 of 100 brands across July 2024 to March 2026 tracked 340,000 ChatGPT sessions and $690,000 in attributed revenue — but flagged explicitly that the dataset captured only the tagged subset. The full picture requires triangulation across three more surfaces below.
The dark third
~70%
of AI-referral traffic arrives as Direct in GA4 because the referrer header is missing.
OpenAI adds utm_source=chatgpt.com to a subset of citation links. The Source Cards above a ChatGPT Search answer have always carried the tag. The 'More sources' panel that expands below the answer received the tag on 13 June 2025. The conversational inline links inside the body of an answer — links that say something like 'as Squarespace's help center notes...' embedded in prose — remain untagged as of mid-2026. Mobile-app visits strip referrer headers regardless of tagging, which means even the tagged URLs lose attribution context once they open in the iOS or Android app webview.
The PPC Land documentation2 covers the timing of the change cleanly. Before June 2025, the only ChatGPT links that carried UTMs were the explicit source cards. After June 2025, the More sources expansion got tagged too, which roughly doubled the visible share of tagged traffic for sites whose citations appeared in that panel. Conversational inline links remain the largest single category of untagged citation traffic, and OpenAI has not publicly indicated when (or whether) those will be tagged.
§03GA4
The GA4 custom channel grouping for AI search
GA4 does not have a built-in AI Search channel as of mid-2026 — every visit from ChatGPT, Perplexity, Claude, or Gemini lands under Referral or Direct unless you build a custom grouping. The setup takes about five minutes: Admin > Data display > Channel groups > New channel group. Add a single channel named AI Search with a regex condition on Session source that matches the major AI engine hostnames. Save and apply. New traffic will route into the AI Search channel from that point forward; historical traffic stays in Referral until a comparable retroactive segment is built.
The regex is the load-bearing detail. The pattern below catches the major AI engines as of mid-2026. Add or remove engines as their hostnames change; new entrants from 2026 onwards (you.com, Pi, etc.) can be appended with a pipe character.
GA4 regexSession source regex for the AI Search custom channel grouping
Pair the channel grouping with a second GA4 segment for "Direct visits to AI-target pages" — a filter for landing-page paths that include your top citation-target URLs combined with a session source of Direct. The combined view catches some of the dark third: visitors who arrived from ChatGPT, lost the referrer in transit, but landed on a page that is heavily promoted in ChatGPT citations specifically. It is not perfect attribution, but it shifts a useful slice of Direct traffic into a "probably AI-search" segment for trend tracking.
§04Squarespace Analytics
The Squarespace Analytics gotchas
Squarespace's built-in Analytics surfaces Traffic Sources with a simpler model than GA4. Referrers must be present in the request header — Squarespace does not parse UTM parameters into a session source the way GA4 does. The practical implication: chatgpt.com appears in your Squarespace Traffic Sources only when the visit carried a referrer header, which is a smaller subset than the GA4 regex catches. Squarespace Analytics is useful for the trend signal but underreports compared to a properly-configured GA4 channel.
Squarespace's own help docs on Analytics4 describe the model: traffic sources are derived from the HTTP referrer header. There is no equivalent to GA4's "Source / Medium" pair, no UTM-parameter parsing into a session-level dimension, and no custom channel groupings. What you see in Squarespace Analytics is the direct read of referrer data, which the dark-traffic problem makes systematically incomplete for AI-search traffic.
For Squarespace owners on Personal plan, Squarespace Analytics is the only built-in measurement layer. Adding GA4 requires Code Injection (Business plan and above) and a Measurement ID. The honest recommendation: if AI-search measurement matters and you are on Personal, upgrade to at least Business to unlock Code Injection, install GA4, and run the regex channel grouping. The plan cost over a year is materially smaller than the missed-attribution cost on a site getting meaningful ChatGPT traffic.
§05AI Visibility panel
What Squarespace's AI Visibility tool actually tells you
The AI Visibility panel inside Squarespace's SEO settings runs a fixed cadence of branded and non-branded prompts against ChatGPT and reports whether your site is mentioned. The signal is mention frequency and Share of Voice, not click traffic. Core and Business plans test prompts every 14 days; Plus, Advanced, Commerce Basic, and Commerce Advanced plans test every 7 days. English-language sites only. The tool is useful as a leading indicator — mentions tend to precede measurable referral traffic by 2-4 weeks — but it is not a substitute for GA4 click attribution.
Squarespace's documentation3 describes the workflow: add 5-10 branded prompts ("What is [your brand] known for?", "Tell me about [your brand]") and 5-10 non-branded prompts ("Best wedding photographers in Asheville", "How do I optimize a Squarespace site for AI"). The panel runs them on cadence and surfaces whether your site appeared in the answer and where it ranked relative to competitors who also appeared.
§06The manual log
The manual tracking spreadsheet that catches what the tools miss
Build a tracking spreadsheet with 10-15 high-priority queries — branded prompts, non-branded category prompts, and the specific long-tail queries you have designed citation-target content for. Run each query against ChatGPT every Friday, log a yes/no for citation appearance, and capture a screenshot. The 30-minute weekly investment fills the gap GA4 and the AI Visibility panel both leave: it covers conversational citations the AI Visibility panel does not test for, and it gives you the qualitative shape of how ChatGPT is using your content (full quote, named mention, source card).
The template lives in a Google Sheet or Airtable with five columns: query, date, appeared (Y/N), citation format (source card / inline quote / named mention), notes. After six weeks the spreadsheet shows a clear trend line per query, and the qualitative column reveals which content types ChatGPT is treating as most authoritative. Pages quoted in source cards typically have stronger named-source hygiene; pages quoted inline typically have stronger passage shape; pages mentioned by name typically have stronger entity wiring.
