In this postCollapse -Open +
Your attribution model is becoming fiction. Not metaphorically. Actually.
Over the last 18 months, three colliding forces have made traditional marketing measurement basically unusable.
Zero-click search exploded. Google's AI Overviews now answer most queries without requiring a click. Your organic traffic dashboard shows declining CTR across the board, but not because your content is worse. Users simply stopped visiting websites to get the answer they needed.
Agentic AI tools hallucinate metrics. A July 2026 incident surfaced when a marketing team discovered their AI analytics layer had fabricated conversion data for three months straight. The VP had made territory decisions on fake numbers. This isn't rare anymore; it's becoming expected.
Attribution got fragmented across channels nobody can see into. ChatGPT Ads reports zero demographic data by design. TikTok's native metrics are opaque. WhatsApp Business APIs don't expose conversion paths. You're stitching together blind spots and calling it measurement.
The result: CMOs are reporting AI as their top strategic priority, yet spending only 8-10% of budgets on AI tools. That gap isn't ignorance. It's rational skepticism. You can't justify the spend when you can't prove the result.
In this postCollapse -Open +
What broke in 2026
Zero-click is real, and it's accelerating.
Bain research shows 80% of consumers now get answers from zero-click results in at least 40% of their searches. That's not a platform preference shift. That's a behavior shift. When Google Overviews summarize product comparisons or answer FAQs directly on the SERP, users don't click through.
For e-commerce and SaaS, this cuts organic traffic by 15-25%. For cannabis operators marketing compliance details, customer FAQs, or product info, the impact runs deeper because AI Overviews tend to prioritize answers over clicks.
The problem: your analytics see a 20% drop in organic traffic, but you don't know if that's from algorithm changes, SERP competition, or zero-click behavior. You retarget, shift budgets, or panic. The real issue (users are getting answers elsewhere) gets buried in noise.
Analytics dashboards now show conflicting data from multiple channels, with hallucination artifacts and incomplete...
AI measurement tools are rolling back like it's 2015.
In June 2026, a mid-market B2B marketing team found that their AI-powered analytics dashboard had been confidently reporting 40% higher conversion rates than their actual CRM showed. The AI wasn't hallucinating on purpose.
It was predicting patterns based on click behavior, ad spend, and seasonal trends, then filling in the conversion gaps with "likely" scenarios. Those scenarios were wrong 87% of the time.
This is becoming systemic. AI models are prediction engines, not fact-checkers. When you ask an LLM-powered analytics layer "How many conversions did this campaign drive?" it doesn't know the answer. So it predicts. Confidently. Your dashboard shows the prediction.
The FTC is watching this closely. Draft guidance on AI transparency in marketing now includes explicit warnings about unsubstantiated measurement claims. Using an AI tool to report metrics you can't verify is starting to look like the kind of thing enforcement actions get built on.
Channel opacity is now the default.
You can't see into:
- ChatGPT Ads Marketplace (no query data, no demographic breakdowns, aggregated metrics only)
- TikTok Shop integrations (conversion data is sparse and delayed by days)
- WhatsApp Business APIs (no user journey visibility, no attribution trail)
- Reddit ads (limited placement data, opaque inventory allocation)
Meanwhile, iOS privacy changes, cookie deprecation, and first-party data fragmentation mean you're losing visibility into the channels you used to own.
Result: you have metrics from 7 channels and verifiable attribution from maybe 2. You're guessing at the other 5, and your AI tool is filling the gaps with plausible-sounding numbers that nobody can verify.
Why this matters
Your budget decisions are based on incomplete information, and you know it.
CMOs I've spoken with over the last quarter are in a specific bind. They want to invest in AI marketing tools because competitors are and VPs expect it. But they can't show the board clean ROI because the data won't support it. So they're spending conservatively on AI while watching competitors move faster.
That gap (strategic priority versus actual spend) is rational. It's the market saying: "This could work, but we can't prove it yet."
Marketing measurement is fragmenting into channel-specific mini-stacks.
Instead of one unified attribution model, smart teams are now building separate measurement approaches for each channel:
- Search: last-click attribution (because SERP behavior fundamentally changed)
- Social: direct response or brand lift studies (because pixel data is now unreliable)
- Email: UTM tracking plus actual revenue (because email has its own trustworthy conversion data)
- Direct traffic: CRM matching and cohort analysis (because it's the most honest signal)
This is more transparent than forcing everything into a single model. But it also means there's no true "multi-touch attribution" anymore. Just honest channel silos with clear guard rails.
Teams are now building separate measurement frameworks for each channel instead of relying on a single unified attribution...
For cannabis brands, the stakes are higher.
Marketing compliance is already complex. If you're using an AI tool to report customer acquisition cost, and that tool is hallucinating data, you could end up with budget allocation that triggers inventory shortfalls (overselling products you didn't have demand for) or compliance violations (spending on prohibited channels because the AI misreported channel mix).
Real measurement in cannabis now requires manual reconciliation: CRM data versus ad platform metrics versus POS data. Anything less is guessing, and guessing in a regulated vertical has real costs.
What to do about it
Stop treating attribution as solved. Treat it as an open question.
Build separate measurement frameworks for channels where you can actually verify data:
- Email: trackable, high-fidelity data
- Paid search: click data is real, conversion data is lagged
- Organic: traffic is real, attribution to keyword is harder
- Social: impressions are real, conversions are guesstimated
For channels where you can't verify conversions (zero-click, AI Overviews, brand influence), measure differently. Use brand lift, share of voice, engagement, reach. Don't force zero-click traffic into an attribution model that assumes clicks exist.
Audit your AI measurement tools weekly.
If you're using an AI analytics platform or AI-powered BI tool to report metrics:
- Compare its reported conversions against your CRM or source-of-truth database
- Run this check every week, not annually
- If discrepancies exceed 5%, stop using the tool until it's fixed
- Document every mismatch and report it to the vendor
The tools aren't malicious. They're just prediction engines, and prediction engines can be wrong with total confidence.
Shift budget toward measurable channels and honest frameworks.
If you can't verify ROI on a channel, don't scale it aggressively. This is the opposite of "move fast and break things."
- Scale email and owned channels (you own the data)
- Run smaller tests on social and emerging channels (measure lift, not attribution)
- Reduce paid search spend if zero-click is stealing your clicks (shift to brand and upper-funnel campaigns)
- Use CRM data as your source of truth, not platform-reported metrics
For cannabis specifically: manual reconciliation is not a bug, it's a feature.
If you're reconciling POS data against ad spend and customer data, you're doing measurement correctly. It's slower, but it's honest. Build that process into your monthly reporting, and you'll know which channels actually drive revenue. Not which channels look profitable in a hallucinating analytics layer.
FAQ
A: Yes. Bain reports 15-25% organic traffic drops for competitive verticals. For cannabis, where search queries are high-intent but also compliance-heavy, AI Overviews are answering most questions before users click. If your site used to capture FAQ traffic, you're probably seeing real decreases now.
A: No, not yet. Compare its reported metrics against your CRM or source of truth weekly. If you see discrepancies above 5%, the tool is hallucinating. This is common and the vendors know it, but they're working through it.
A: Not if you can verify it. But most multi-touch models are now fed by AI layers you can't see into. Build attribution from clean data only: direct conversions, trackable channels, verified transactions. Be honest about what you don't know.
A: No. Better data infrastructure, standardized measurement APIs, and more transparent AI tooling will get built. But in 2026, the honest answer is: most marketing measurement is less reliable than it was in 2020.
A: "Attribution is fragmenting. We're shifting to channel-specific measurement. We're reducing spend on unverifiable channels and focusing budget on campaigns we can actually track. We're building separate frameworks for AI-driven and traditional channels."
A: Yes. That's why CMO spend on AI is lagging strategic interest significantly. Everyone knows AI marketing tools could work. Nobody can prove it yet. ---