Written by: Michał Sobieraj, Operations Manager, Digital Colliers
Reddit's Q2 letter to investors did something most brands haven't done yet. It said out loud that traffic from search is being pressured by AI answer engines, and it gave shareholders a rough sense of the shape of the hit. If a platform the size of Reddit can measure that, your brand can too. The problem is your analytics stack almost certainly isn't set up to see it.
The column your GA report doesn't have
Open your channel grouping report. You'll see Organic Search, Direct, Referral, Paid, Email, Social. There is no row called AI Answer Engines. So when ChatGPT, Perplexity, Gemini, or Google's AI Overviews send a shopper your way, that visit lands somewhere else in the report. Sometimes it's Direct because the referrer got stripped. Sometimes it's Organic because Google's AI surface still passes through as google.com. Sometimes it's a Referral row you've never looked at.
The practical effect: your Organic line looks softer than last year and you can't tell whether that's SEO decay, seasonality, or answer engines quietly intermediating your top-of-funnel. In a market where UK eCommerce grew about 3% in 2024 versus 2023, a couple of missing points of attribution is the difference between a plan that works and one that doesn't.
What to instrument, concretely
You don't need a new platform. You need four things wired up in the next two sprints.
- Referrer inspection at the edge. Log the raw Referer header and the Sec-Fetch headers on every landing hit. Bot fingerprints from OpenAI, Anthropic, Perplexity, and Google-Extended are documented. Tag them.
- A custom channel group. Add rows for AI Assistants (human clicks arriving from chat.openai.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and known wrappers). Separate that from AI Crawlers (bots doing the retrieval, not the click).
- UTM discipline on anywhere you can control. If you're seeding content that answer engines chew on (PR, affiliate, YouTube descriptions, PDP schema), tag it. You can't tag what an LLM decides to cite, but you can tag every human-authored surface that feeds it.
- A one-question survey on the order confirmation page. "Where did you first hear about us?" with an option for "ChatGPT or another AI assistant." This is the only source of truth for zero-click discovery. Run it for a quarter and cross-reference against SKU.
The SKU cross-reference is where it gets interesting. Answer engines don't recommend evenly. They lean toward products with rich third-party review coverage and clean structured data. You'll usually find a small cluster of SKUs punching above their paid-media weight in the survey data. That's your signal.
Why this matters for FY planning
Acquisition cost is not getting cheaper. Meta CPMs kept rising through 2024 and 2025, and CAC across DTC has climbed roughly 40% since 2023. If a meaningful slice of what used to be free organic discovery is now being routed through an AI layer you can't see, your paid budget is silently backfilling a channel loss you haven't named.
Then look at the margin side. At a typical multi-channel brand, around 30% of SKUs lose money per order once returns and ad spend are counted. Online return rates sit around 19-20% of gross sales, higher for apparel. If your buying team is doubling down on hero SKUs based on last year's organic mix, and that mix was partly synthetic (AI-mediated discovery you attributed to Organic or Direct), you're going to over-order the wrong things.
What the operators getting ahead of this actually do
The pattern I keep seeing on brands who take this seriously in 2026 planning:
- They separate AI crawler traffic from AI-referred human traffic in reporting. Different meanings, different actions.
- They run the "how did you hear about us" question year-round, not as a one-off.
- They monitor which of their own SKUs get cited in answer engines. Manual sampling weekly is fine to start.
- They rebuild PDP structured data and review coverage for the SKUs that show up in AI answers, because that's the flywheel.
- They treat the AI channel as a real line in the FY plan with a target, even if the target is qualitative for the first two quarters.
The brands that skip this step will spend 2026 blaming SEO, blaming Meta, and reallocating budget away from channels that were actually working. The instrumentation is cheap. Not doing it is the expensive option.