# AI Search Traffic Analytics and Logs: Track Conversions in GA4

> Find AI search traffic analytics logs data in GA4: separate AI referrals, missing referrers and Google Search, then tie visits to conversions.

- Author: Simran Rawat, Growth Executive, Nurdd
- Published: 2026-11-16
- Publisher: Noma by Nurdd (https://nomahq.app)
- Original page: https://nomahq.app/blog/ai-search-traffic-analytics-logs

To track conversions from AI search in GA4, split your traffic into three buckets: tagged AI referrals (such as utm_source=chatgpt.com), Google Search referrals, and visits with no referrer at all. Then match each bucket's landing pages to conversion events and cross-check against server logs. Your AI search traffic analytics and logs will always undercount, so treat the totals as a floor.

## Key takeaways

- Google's new help page says a google.com referrer proves Google sent the visit, but not whether it came from Search, Discover or Images.
- Google's page does not cover AI Mode or AI Overviews, and Search Console has no appearance type that separates those clicks from standard ones.
- ChatGPT is the one engine that reliably tags links with a utm_source. Everything else depends on the referrer header, which can be missing.
- Build a custom GA4 channel group for AI referrals, and keep a separate, labelled bucket for direct visits that land on pages an AI engine is likely to cite.
- Server logs catch visits that blocked scripts never report, but you must count only landing-page requests, not images, CSS or JavaScript.
- Analytics shows who clicked. It cannot show whether an AI answer named you. That needs prompt-level measurement.

## What did Google's referrer help page change for AI traffic measurement?

In November 2026 Google added a Search Central page documenting the HTTP Referer values that Search traffic sends. [Search Engine Journal, November 2026](https://www.searchenginejournal.com/google-search-traffic-analytics-logs/597382/) reported that Google's update log describes it as documenting "how site owners can identify and track traffic from Google Search" in analytics. The [Google Search Central page](https://developers.google.com/search/docs/monitor-debug/identify-google-search-traffic) is the first official reference you can hold your channel rules against.

Three points from it matter for AI measurement:

- **Origin-only referrer.** Google sends only the scheme and host, such as https://www.google.com/ or a local domain like https://www.google.co.uk/. No path, no query, no search terms.
- **Shared values.** Discover and Google Images send the same values, so a google.com referrer cannot tell you which surface produced the visit.
- **Referrers can vanish.** Private mode, some new-tab behaviour and switching from a mobile search widget to a browser can strip the header, according to the page as summarised by Search Engine Journal.

The same report notes the page was last updated on 12 November and does not cover AI Mode or AI Overviews. That gap is the reason this guide exists. Google Analytics Help states that its Organic Search channel includes [Google's AI Overviews and AI Mode, per Google Analytics Help, as checked November 2026](https://support.google.com/analytics/answer/9756891?hl=en). So in GA4 those clicks are folded into Organic Search, and you cannot pull them out with a channel report alone.

## Why do my Search Console clicks differ from analytics?

Search Console counts a click when someone selects a link in Google results. GA4 counts a session only after the page loads and the tracking script runs. Google's help page, as summarised by [Search Engine Journal, November 2026](https://www.searchenginejournal.com/google-search-traffic-analytics-logs/597382/), lists ad blockers, consent choices and privacy settings as reasons the script may never fire.

Use this table to work out which cause applies to your gap.

| Symptom | Likely cause | How to check |
| --- | --- | --- |
| Search Console clicks higher than GA4 Organic Search sessions | Script blocked, consent declined, or user left before load | Compare against server-log landing-page requests with a google referrer |
| Google visits sitting in Direct | Referrer stripped by browser, app or redirect with a strict Referrer-Policy | Trace redirects on top landing pages and check the Referrer-Policy header |
| Google visits sitting in Referral | Custom channels, referral exclusions or filters reclassified them | Review channel group rules and unwanted-referrals list in GA4 admin |
| GA4 higher than Search Console | GA4 may include Google Ads traffic, Discover or Images; Search Console counts organic clicks only | Split by medium, and use the Discover report for Discover |
| Lost gclid on paid landing pages | Redirect drops the query string | Test the redirect and confirm it passes parameters |

A mismatch of a few percent is normal. A large and growing gap usually points to a consent banner change or a redirect, not to a Google reporting fault.

## How do I find AI search traffic in GA4?

AI visits arrive in three forms, and each needs its own method. Do them in order.

### Step 1: Find tagged referrals

OpenAI's publisher FAQ says ChatGPT [automatically includes utm_source=chatgpt.com in referral URLs (OpenAI Help Center, as checked November 2026)](https://help.openai.com/en/articles/12627856), and that publishers who allow OAI-SearchBot can see this in analytics such as Google Analytics. In GA4, open Reports, Acquisition, Traffic acquisition, switch the primary dimension to Session source / medium and search for chatgpt.

### Step 2: Build a custom channel group

Google maintains the default channel group and does not let you edit it, according to [Google Analytics Help, as checked November 2026](https://support.google.com/analytics/answer/9756891?hl=en). So create a custom one: Admin, Data display, Channel groups, Create new channel group. Add a channel called AI referrals with a condition where Source matches regex, then list the engines you see in your own data, for example chatgpt.com, perplexity, gemini.google.com, copilot.microsoft.com, claude.ai. Place it above Referral so the rules fire first. Validate the pattern against your actual source list rather than copying one from a blog, because domains change.

Some properties also show an AI Assistant default channel; if yours does, add a condition that Default channel group matches it, so your custom group and Google's stay aligned.

### Step 3: Isolate Google Search from AI Mode inside Organic Search

You cannot separate AI Mode clicks in GA4. What you can do is watch landing pages. Query-style, long-form informational pages that gain Organic Search sessions while Search Console impressions for classic queries stay flat are candidates for AI-driven growth. Label that as inference, not measurement. Our guide on the [Search Console AI report and why it has no clicks](https://nomahq.app/blog/search-console-ai-performance-report-clicks) covers the reporting limits in more detail, and the [free Google Rank vs AI Mode Checker](https://nomahq.app/tools/google-rank-vs-ai-mode) shows what AI Mode cites for a single query next to your organic rank.

### Step 4: Hold a separate bucket for missing referrers

Create an Exploration with Session default channel group as a row, Landing page as a second row and Key events as values. Filter to Direct. Then flag landing pages that are deep, specific articles or comparison pages. Few people type those URLs by hand, so Direct sessions on them are probably untagged clicks from apps, AI browsers or stripped referrers. Report this bucket as "unattributed, likely referred", never as AI traffic.

## Why is AI traffic hard to attribute in the first place?

Tags can break. On 15 October 2026, [Search Engine Land, October 2026](https://searchengineland.com/chatgpt-appears-to-drop-utm-tracking-making-ai-referral-traffic-harder-to-measure-496552) reported that ChatGPT links appeared to have lost utm_source=chatgpt.com after the GPT-6 release. It said OpenAI confirmed a known bug and began rolling out a fix. Because referral data was also missing from most ChatGPT clicks, those visits could land in Direct, so a fall in ChatGPT referrals alongside a rise in Direct may be a tracking change rather than a real drop.

> Search Engine Land advised teams to keep using UTM tracking, since referral information is still missing from most ChatGPT clicks.

Concentration is the other issue. [BrightEdge, September 2026](https://www.brightedge.com/news/press-releases/brightedge-data-chatgpt-referral-traffic-doubles-in-2026) reported that ChatGPT referral traffic rose 101% from January to August 2026 and reached a 95.1% share of AI-generated referrals, roughly 22 times the combined share of Gemini, Claude and Perplexity. This is BrightEdge's own dataset of brand sites, so your mix may differ, but it explains why a ChatGPT-shaped hole in your data matters more than any other engine.

## How do server logs complement GA4 for AI search traffic?

Logs record the request whether or not a script ran. They give you a second count that ad blockers and consent choices do not touch. Google's guidance, as summarised by [Search Engine Journal, November 2026](https://www.searchenginejournal.com/google-search-traffic-analytics-logs/597382/), says to count only requests to landing pages, because images, CSS and JavaScript requests carry your own landing page URL as their referrer.

1. **Filter to page requests.** Keep GET requests with HTML content type and a 200 status. Drop assets, bots and health checks.
2. **Extract the referrer and query string.** In a combined-format log, the referrer is the quoted field after the status and size. Keep the request URL too, so you can search it for utm_source=chatgpt.com.
3. **Group by referrer host.** Count google domains separately from chatgpt.com, perplexity.ai and similar hosts. Treat android-app://com.google.android.googlequicksearchbox as Google, per the Google page.
4. **Count empty referrers by landing page.** This is your log-side version of the Direct bucket.
5. **Compare weekly with GA4.** Log sessions should exceed GA4 sessions by a stable ratio. If the ratio jumps, a consent or tag change is the likely cause.

Separate crawlers from people. AI bots fetching your pages are not visits, and mixing them in inflates the numbers. Our guide to [tracking AI crawlers in logs and analytics](https://nomahq.app/blog/ai-crawler-analytics-track-bots) shows how to split them out.

## How do I connect AI visits to conversions?

Once the buckets exist, attach conversions. In GA4, mark your lead form, trial start and purchase as key events, then build one table: rows are your channels (AI referrals, Organic Search, Direct on deep pages), columns are sessions, key events and key event rate.

A worked example, using made-up numbers to show the arithmetic:

| Bucket (illustrative) | Sessions | Trial starts | Rate |
| --- | --- | --- | --- |
| AI referrals (tagged) | 400 | 16 | 4.0% |
| Organic Search | 10,000 | 200 | 2.0% |
| Direct, deep pages only | 600 | 12 | 2.0% |

If the Direct deep-page bucket converts like AI referrals, you might estimate that part of it is untagged AI traffic. Do not add it to the AI total. Show the range: 16 trial starts confirmed, up to 28 if every deep-page Direct visit were AI. That range is honest and still useful for a budget discussion. For a fuller framework, see our [AI search attribution guide](https://nomahq.app/blog/ai-search-attribution-guide) and the explainer on [what AI search attribution is](https://nomahq.app/blog/what-is-ai-search-attribution).

## What mistakes distort AI search traffic analytics?

- **Reading a Direct spike as brand growth.** Check the landing pages first. Lost UTMs make a steady channel look like it vanished.
- **Counting asset requests in logs.** They inflate Google referral counts, as Google's page warns.
- **Putting the AI channel below Referral.** GA4 evaluates rules in order, so the AI traffic is classified as Referral before your rule runs.
- **Assuming one UTM covers every engine.** Only ChatGPT is documented as adding one. Check each engine's behaviour in your own data. Our post on [Gemini UTM parameters](https://nomahq.app/blog/gemini-utm-parameters-track-referrals) shows how to test one engine.
- **Judging AI search by clicks alone.** Many answers name a brand and never send a click.

## What can analytics not tell you about AI search?

Analytics is blind before the click. It cannot show whether ChatGPT named your brand for a buyer question, which competitor it named instead, or which of your pages the answer drew on. A visit with no referrer says nothing about what the AI said. As we argue in [measuring AI influence beyond referrals](https://nomahq.app/blog/geo-referral-traffic-measurement), a referral count is the thin end of the signal.

Noma by Nurdd is an answer engine optimisation platform that asks 13 AI engines, including ChatGPT, Gemini, Perplexity and Claude, the questions a brand's buyers ask, every day. It shows where the brand is named, where competitors are named instead, and which pages the answers draw on. That is the pre-click view GA4 and server logs cannot give you. See how it works on [AI share of voice](https://nomahq.app/blog/measure-ai-share-of-voice) and choose questions with [how to choose prompts to track](https://nomahq.app/blog/choose-prompts-to-track). To see it on your own brand, [book a demo](https://nomahq.app/demo) or [start the 3-day free trial](https://platform.nomahq.app/signup). A card is taken at signup so billing can start on day 3; cancel before then and nothing is charged. Plans are listed on the [pricing page](https://nomahq.app/pricing).

## A weekly checklist for AI search conversions

1. Confirm the AI referrals channel sits above Referral and still matches current source names.
2. Check ChatGPT tagged sessions week on week. A sudden fall with a Direct rise means check the tag, not demand.
3. Compare GA4 Organic Search sessions with Search Console clicks and log landing-page requests. Note the ratio.
4. List Direct sessions on deep pages and their key events.
5. Test the top landing-page redirects for dropped query strings and strict Referrer-Policy headers.
6. Record which prompts name your brand, using prompt-level tracking, and compare movements with the referral line.

## How we researched this

We read Google's Search Central page on identifying Search traffic, Google Analytics Help on channel groups, OpenAI's publisher FAQ, trade reporting from Search Engine Journal and Search Engine Land, and BrightEdge's published referral data. We did not run our own traffic experiments, and the worked example above uses invented numbers for illustration. Facts, dates and figures were checked on 2026-11-15.

## What to do next

Build the AI referrals channel today, run the three-way comparison between GA4, Search Console and logs this week, and add a view of what AI engines actually say about you. [Book a demo](https://nomahq.app/demo) or [start the free trial](https://platform.nomahq.app/signup) to see where your brand is named and where it is not. You can also run a [free brand audit](https://nomahq.app/brand-audit) first.
