GA4 channels: give ChatGPT and Perplexity referrals their own group or not?

Referral reports in GA4 are getting noisy, with chatgpt.com and perplexity.ai mixed in with random blogs and directories. I'd build a custom channel group for AI assistant sources, using a regex on session source, and keep those sessions out of Organic Search so SEO numbers stay honest. Two caveats. A lot of answer-engine clicks arrive with no referrer and land in Direct, so this bucket will always undercount. And for most sites the volume is still small, so compare conversion rate per session instead of raw counts. Is anyone putting these under Organic, and how do you explain it when someone asks what drove the demo requests?

I wouldn't keep these out of Organic, at least not for the SEO view. Someone who asks ChatGPT for the best accounting software for freelancers and clicks a source is doing the same job as someone typing that into Google. And answer engines mostly pull from pages that already rank well or get cited elsewhere, so those sessions are often a direct result of the content work you're measuring. Pull them out and your SEO numbers look cleaner, but you lose the evidence that your pages are being picked up at all.

A better setup is a separate AI Assistants channel for reporting, plus a combined "Organic + AI answers" view for deciding what to write next. Your per-session conversion comparison is the right call given the volume. Just don't use the split to conclude that SEO is doing less than it is.

I'd keep these separate from Google clicks. Search Console is the only outside check you have on your organic numbers, and it doesn't show clicks from ChatGPT or Perplexity. If those sessions land in Organic, GA4 stops reconciling with GSC, and you can't tell whether a jump came from rankings or from citations. You're right that citations depend on content work, but a separate channel is what shows you the traffic exists and whether it converts.

I don't buy the reconciliation argument. GA4 Organic Search never matched GSC cleanly anyway: GSC only counts Google, GA4 organic also picks up Bing and DuckDuckGo, and sessions and clicks don't line up in the first place. So pulling AI sessions out of Organic wouldn't break something that was already loose.

What matters more is whether your regex catches the sources at all. ChatGPT can show up as chatgpt.com or chat.openai.com, Perplexity as perplexity.ai, and Gemini as gemini.google.com, which ends up in Referral unless you add it. Look at the session source report before you trust any AI channel number.

Agreed on checking the source report first. I'd also be clear about where the regex stops. Google's own AI surfaces, AI Overviews and AI Mode, send clicks as ordinary Google Search traffic, so those sessions already sit in Organic and GA4 won't flag them as AI. Your AI channel will only cover the standalone apps: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com.

Put that boundary in the channel description, or someone will compare the AI channel to total AI influence and decide it's tiny. It also means Mira's combined Organic plus AI view is the only place where AI answers show up in full, which is a good reason to keep both views.

Agreed on the boundary. I'd also label the channel by what it actually measures, something like "Standalone AI apps," so nobody reads it as total AI influence. Whether a separate channel is worth the setup depends on volume. If AI sessions are only a few per week, a combined view is enough to decide anything.

Volume matters, but I'd start with what decision the number feeds. If you just need the report to be honest, a combined view works even at a few hundred sessions a month. If you're deciding whether to fund citation-friendly content or a new page type, a separate channel is worth the setup even when it's small. A small line you can track over time tells you more than a bigger number buried in Organic where you can't read it.