5 min to read
Something strange started happening with our clients' analytics early this year. Traffic was holding steady, conversion rates were decent, but when we dug into the sources, something kept appearing: direct traffic was spiking. Not a little — significantly. And the pattern was consistent across multiple accounts in different industries.
At first, we did what any good analytics team does. We looked for tagging issues. Broken UTM parameters. Referral exclusions gone wrong. But the technical setup was clean. What we were seeing wasn't a bug. It was a blind spot — and it's one that's now affecting virtually every business that depends on digital marketing.
Here's what's happening, why it matters for your Q2 budget decisions, and what I'm telling every Marketing Director we work with right now.
Why Your Analytics Dashboard Is Giving You a False Sense of Security
Google Analytics 4 was built for a world where people discover things, click links, and land on your site. The attribution model — last click, first click, data-driven — all assumes a human making a deliberate journey from point A to point B.
That world is being dismantled. Right now, a growing percentage of your potential customers aren't clicking links to find you. They're asking ChatGPT a question, reading the answer, and then — if your brand was mentioned — typing your URL directly into their browser. Or they're using Perplexity to research options, getting a synthesized response, and navigating to the top-cited source.
That traffic lands as "direct" in GA4. No source. No medium. No campaign. Just a visitor who appeared from nowhere.
Research tracking the dark traffic phenomenon suggests LLM-driven referral traffic can represent 15–35% of a site's direct traffic depending on the industry. For B2B brands in technology, marketing, legal, and finance — sectors where people actively use AI research tools — the number skews higher.
The problem isn't just an analytics inconvenience. It's a budget problem. If your performance marketing reports show paid search driving 40% of conversions and direct driving 25%, your team is making spend decisions based on those numbers. But if 30% of that "direct" traffic is actually coming from AI citations you've earned, you're under-valuing content and over-crediting paid campaigns. You might be about to cut the wrong budget line.
The Hidden Traffic: How AI Tools Are Sending Visitors You Can't Track
Let me explain the technical reality, because it matters for how you fix it.
When a user reads an AI-generated answer in ChatGPT, Perplexity, or Google's AI Overview and then visits your site, the request typically arrives without a referrer header. In some cases the AI interface doesn't pass referrer data at all. In others, users manually type or copy-paste URLs. Either way, GA4 sees no source and defaults to "direct."
There's also a secondary issue: branded search. When ChatGPT mentions your brand and a user then searches your brand name on Google, that traffic shows up as branded organic — organic search, your brand keyword. That looks great on paper, but it's actually AI-assisted discovery, not traditional SEO performance. The two are very different in terms of what drives them and how you'd optimise for them.
This is something I've been sitting with a lot lately, especially after working through the implications with our own data team here at Codedesign and reading through some excellent research from Sparktoro on dark traffic patterns. The core insight is this: attribution models were built to track human-to-human digital journeys. AI is inserting a non-human intermediary into that journey, and our measurement infrastructure hasn't caught up.
Platforms like Perplexity have started to introduce referral tracking for some publishers, but this is patchwork. ChatGPT, Gemini, and others don't systematically pass referrer data. The gap is structural, not accidental, and it's not going to be solved by a platform update in the next 90 days.
What We Found When We Audited a Client's GA4
A few months ago, one of our clients — a B2B professional services firm targeting mid-market companies — came to us with a familiar concern: their paid search costs were rising but conversions weren't keeping pace. The instinct from their internal team was to cut content investment and double down on paid.
We ran a full attribution audit before making any recommendations. What we found reframed the entire conversation.
Their direct traffic had grown 41% year-over-year. When we cross-referenced this with their brand mention data across AI platforms — running manual queries across ChatGPT, Perplexity, and Gemini, plus third-party monitoring tools — we found their brand was being cited regularly in responses to queries like "best [industry] consultants for mid-market companies." We also noticed their branded search volume had jumped 28% in the same period, without any new paid brand campaigns.
Their content was their most valuable AI visibility asset. Cutting content investment would have been exactly the wrong move. The attribution model was hiding this entirely — making the content look like a cost centre when it was actually one of the highest-performing channels in the stack.
At Codedesign, this is now a standard part of how we approach analytics audits. You can't make good channel investment decisions if your measurement framework was designed for a world that no longer exists.
How to Start Measuring AI-Driven Traffic in 2026
Here's the practical side, because theory without action is useless. This is what I'm recommending right now.
Segment your direct traffic properly. Stop treating "direct" as a mystery bucket. Break it down by landing page (AI citations tend to land on deep content pages, not your homepage), device, and time-to-convert. AI-primed visitors often arrive pre-sold and convert faster. The patterns are visible if you look.
Monitor your AI mentions manually and with tools. Run weekly queries across the major AI platforms for your most important search terms and questions. Track whether and how your brand appears. Tools like Brandwatch and emerging AI visibility trackers are beginning to capture this data — it's not perfect, but it's directionally critical.
Track branded search volume separately. Create a dedicated segment in GA4 for branded organic. If branded search is growing without paid brand campaigns, that delta is almost certainly AI-assisted discovery. Quantify it and put it on the leadership dashboard.
Add UTM parameters everywhere you control external links. Guest articles, podcast show notes, LinkedIn posts, PR features — anywhere your link might be scraped by AI training data, make sure you have tracking in place so at least part of the journey is visible.
Audit your content for AI citation potential. The content most likely to be cited by AI systems answers specific questions with clear, authoritative information. Identify your highest-performing content by this measure and build more with those characteristics. This is simultaneously a GEO strategy and a dark traffic generation strategy.
I've been sharing more thoughts on the intersection of measurement and AI strategy over at VoiceofExperts.com — worth exploring if you're trying to build the right analytics foundation for the rest of 2026.
What This Means for Your Q2 Budget Right Now
If your Q2 budget review is coming up — and for most organisations it's either just happened or is imminent — here's the clear-eyed take.
Your paid channel performance is probably being over-stated in your attribution reports. Not because paid isn't working, but because some of the conversions it's being credited with were influenced by AI-driven brand discovery that your model can't see. This doesn't mean cut paid — it means reinterpret the data before you optimise purely on what the model tells you.
Your content and thought leadership investment is probably being under-stated. In an AI-first world, strong content is what earns you citations in AI responses, which drives the dark traffic that converts. This is a measurable business impact — you just need the right framework to see it.
And your brand-building investment — PR, podcasts, bylines, authority-building — is more valuable than it's looked in a decade. AI systems cite sources they trust. Building a brand that is recognised as authoritative by AI systems is one of the highest-leverage things you can do right now, and it compounds over time.
The businesses that figure this out in Q2 2026 are going to look very prescient by Q4. The ones that keep optimising purely on what their GA4 dashboard tells them risk making expensive decisions based on fundamentally incomplete data.
At Codedesign, AI attribution analysis is now part of every performance marketing engagement we run — because the measurement gap is real, and the budget consequences of ignoring it are significant.
I'd genuinely like to know: have you noticed unusual growth in your direct traffic or branded search that you can't explain through traditional attribution? And is your team currently tracking AI-driven brand mentions in any systematic way? Drop a comment below or reach out directly — this is one of those conversations where hearing what others are seeing in their data is genuinely useful.
— Bruno Gavino, Founder, Codedesign
Add comment ×