Here's the question that quietly kills most small business interest in a new marketing channel before it even gets tested: how would we actually know if it worked?

It's a fair question, and with conversational AI advertising, it's a genuinely harder one to answer than it is with Google or Meta. There's no familiar click-to-conversion trail yet. Attribution windows, multi-touch tracking, and incrementality modelling for this channel are all still being built out by the platforms themselves.

None of that means measurement is impossible. It means the enterprise version of measurement isn't ready, and the SMB version doesn't need to wait for it.

What's actually available right now

Conversational AI platforms are moving fast on this. OpenAI has recently turned on configurable attribution windows for ChatGPT ads, covering both click-through and view-through activity, and is working with third-party attribution vendors to track its role across a longer customer journey. Incrementality testing and marketing mix modelling are described as further down the roadmap, not live yet.

Worth being honest about the broader context here too. Measurement dissatisfaction isn't unique to AI. The IAB's 2026 State of Data report found that among buy-side marketers actively using AI-powered measurement tools, somewhere between 60% and 75% say those tools still fall short on rigor, timeliness, trust, or efficiency. That's not a stat specific to conversational AI ads, it's a snapshot of how marketers feel about advanced measurement generally, AI-powered or not. If your existing channels aren't measured to your satisfaction either, "conversational AI isn't measurable yet" starts to look less like a reason to wait and more like business as usual.

Why waiting for perfect measurement is the wrong instinct

Every channel goes through an early phase where the tracking lags the opportunity. Paid social went through it. Programmatic went through it. The businesses that got comfortable early usually weren't the ones with the best measurement, they were the ones willing to define "good enough" for themselves and start collecting real data while everyone else waited for permission.

For a business with real budget constraints, that patience has a cost most people don't count: competitors testing now are building institutional knowledge about what works in this channel while you're still waiting for the dashboard to look right.

What "good enough" measurement actually looks like for an SMB

You don't need a multi-touch attribution vendor. You need to answer one question honestly before you spend a dollar: what result, if you saw it, would make you call this test a success?

That answer is different for different businesses, and that's fine. A business selling a $150 product wants to see orders this week. A business with a six-month sales cycle should care more about whether a lead mentions ChatGPT unprompted in their first conversation with sales. Neither is more legitimate than the other. What matters is picking one before you start, not after.

Once you've picked it, here's how to actually track it without any specialist tooling:

Add one question to your enquiry or checkout flow. "How did you hear about us?" with an AI/chatbot option, or simply "Did you use an AI tool like ChatGPT before contacting us?" This costs nothing to build and gives you a direct signal no pixel can.

Use a dedicated landing page or offer code for anything you run in this channel. If you can't track a click all the way through yet, you can still isolate the channel by giving it something unique to point to.

Put conversion tracking in place now, even if you're not running paid AI campaigns yet. Clean, working conversion tracking on your site is the foundation this entire channel gets built on later. Getting it right before you need it means you're not scrambling when the platform-side tools mature.

Keep one simple source of truth. It doesn't need to be a centralised intelligence platform. A shared spreadsheet or a CRM field that every lead gets tagged against is enough, as long as everyone on the team actually uses it.

Where Blend sits on this

We tell clients the same thing about this channel that we tell them about most new channels: don't let the absence of perfect measurement talk you out of learning something real. Define your own bar for proof, put the cheapest possible version of tracking in place, and treat the first few months as genuinely a test, not a bet you need to win immediately.

The platforms will keep building out their side of this. Attribution windows, third-party integrations, incrementality models, all of it is coming, on its own timeline. Your ability to know whether this channel is working for your business doesn't have to wait on any of it.

If you're not sure what "good enough" measurement should look like for your business in this channel, that's worth a conversation before you spend on the test itself.