AI Search & AI Overviews

Your traffic is dropping but profit is up? Time to change your metrics

By Eduard Solomko·2026-08-12
Your traffic is dropping but profit is up? Time to change your metrics

You open Google Analytics and see organic traffic dropping. You panic. And you might be panicking for nothing — maybe everything's fine, you're just looking in the wrong place.

Here's the thing. AI search engines — ChatGPT, Perplexity, AI Overviews in Google, Copilot — increasingly make the decision for the user BEFORE they ever open your site. Someone asks a question, gets a detailed answer that mentions your brand, forms an opinion — and either comes to you already warmed up, or doesn't come at all, because the question is already answered for them. The classic "organic traffic" metric sees less than half of what's actually happening here.

This isn't theory. It's already the reality for people working with long sales cycles and high AOV — B2B, YMYL niches, expensive services. Let's go through what to measure instead of good old "sessions from search."

Layer one: access

The most basic question — can AI bots even read your content? Sounds like a formality, but this is exactly where people trip up. OpenAI, Google, Perplexity, Claude — each has its own crawler, and if your robots.txt blocks it "just in case," or because someone in 2019 decided to lock down everything unnecessary — you've cut yourself off from a channel that's now shaping demand.

If I were you, I'd start right here: open your server logs and check who's actually showing up. User-agent is easy to fake, so you need to verify through reverse DNS — that's how you see the real picture, not what bots claim to be.

Layer two: visibility in answers

Next — are you actually showing up in AI-generated answers? This isn't about rankings, it's about your brand getting mentioned in the text the model generates. This takes discipline: build a stable list of queries your actual customers ask, and monitor it regularly. Not once, not "checked it and forgot" — treat it as an ongoing process, because AI answers shift faster than organic rankings ever did ten years ago.

Google Search Console already shows data on AI Overviews, Bing gives you a similar picture for its AI mode. Neither is a perfect tool, but ignoring them means measuring only part of reality.

Layer three: AI referrals

A separate kind of traffic — direct clicks from AI interfaces themselves. Volume is still low, but the quality of this audience is usually higher: the person already has context, already understands why they're coming to you. Don't lump this in with regular organic traffic — track it as its own line, or you simply won't see it against the noise of everything else.

Layer four: post-AI demand

Now this is where it gets interesting. Someone might see your brand mentioned in an AI answer, close the chat — and two days later come straight to you, typing your name into the address bar, or searching for you by brand through regular search. On paper, this looks like organic or direct traffic. In reality, it's a conversion from AI visibility that just never got attributed properly.

Tracking this directly is hard, but you can look at it indirectly: growth in branded search demand, an unexplained bump in direct traffic, spikes right after publishing content that clearly made it into test AI answers.

Layer five: money

And the most important one — tie all of this back to business metrics. Demo requests, leads, pipeline, revenue. If AI visibility is climbing but your downstream numbers aren't moving — either you're measuring the wrong thing, or your content is showing up in answers but failing to convert attention into actual interest. These are two different problems, and they need two different fixes.

What to do right now

Don't block AI bots "on principle" — that's not a threat to your traffic, it's a potential revenue channel that just hasn't learned to attribute itself properly in your usual reports yet. Check your robots.txt, check your logs, build a list of test queries, and start tracking brand mentions in AI answers as their own metric — even in a spreadsheet by hand, if you don't have a tool for it yet.

Old SEO metrics aren't dead. They're just incomplete. And whoever starts measuring all five layers first, instead of just the one they're used to, will see the real picture before their competitors do — which means they'll start reacting to it sooner too. If you need help figuring out what this looks like for your specific project, we offer SEO audit and strategy services.

ES
Eduard Solomko
iGaming SEO · Project Lead / Head of SEO. 13 years in IT, affiliate networks across 12 countries, own 9-tool stack. @neo_raketa