AI Overviews have quietly broken how we measure marketing.
Buyers are researching, deciding, and moving on without ever clicking. If you’re still measuring success through last-click attribution, you’re not just reading the wrong data. You’re making decisions based on a story that has a major chapter missing.


Outbound • Inbound


What’s breaking and why it matters right now
What’s happening, in plain terms?
Someone searches for a solution your company provides. Google’s AI Overview reads your content, pulls out the relevant parts, and answers the question directly on the search results page. The person gets what they need. They don’t click through to your site. Your analytics register: zero visits, zero sessions, zero conversions.
But that person just had a meaningful encounter with your brand’s thinking. They might be primed to remember your name. They might search for you directly next week. To your current analytics stack, though, that interaction simply didn’t happen.
Now multiply that by thousands of searches a day.
83% of searches end without a single click when AI Overviews or instant answers appear.
Users get what they need and move on.
Here’s where it gets expensive. When GA4 organic traffic flattens or drops, the natural instinct is to assume demand is softening. You cut organic content production budgets, pull back on thought leadership, and shift investment to paid media. But demand hasn’t dropped, it’s just moved somewhere the old tools can’t see.
The paid search problem is just as bad and possiblyworse. Branded search – people typing your company name into Google – is one of the strongest signals your marketing is working. It shows someone has heard of you, thought about you, and wants to find you specifically. But AI answers are now intercepting some of those journeys before the branded search ever happens. So, your branded click-through rate looks weaker. Your paid ROAS looks worse. The instinct is to cut branded spend. Again, that’s the wrong direction.
The third rupture is in how we qualify pipeline. B2B buyers now do significant research inside ChatGPT, Perplexity, Claude or Gemini before they visit a vendor’s site. When they eventually fill in a contact form, they are not at the start of their buying process; they’re near the end. But the MQL model treats the form as the beginning. The pipeline looks like it came from a single touchpoint – when really it came from six weeks of AI-assisted research that left no footprint in your CRM.
The result: MQL costs look inflated. Conversion rates look low. Marketing looks inefficient. And the CFO starts questioning budgets that are working well, but not showing up in the data.
Five things to change before next quarter
You don’t need to overhaul everything overnight. You do need to stop optimising against a broken signal and start building a new baseline. Here’s where to start today.
01
Treat branded search volume as your primary demand signal
Branded search – people actively looking for you by brand name – is the clearest downstream signal that your wider marketing activity is working. When your brand shows up credibly inside AI answers, branded search volume goes up. It might take a few weeks, but it moves. Start tracking it weekly, with a trend view, not just a monthly total. If branded search is rising, demand is growing. If it’s flat while paid budgets increase, something isn’t connecting.
02
Start auditing your share of voice inside AI answers
You need to know how often your brand appears when buyers ask the kinds of questions you want to be associated with. This isn’t complicated to start. Pick the twenty or thirty prompts or questions your ideal customers ask when they’re evaluating solutions in your space. Run them through ChatGPT, Perplexity, and Google’s AI Overview. Note where you appear, where you don’t, and who’s showing up instead. Do this weekly. Log it in a simple spreadsheet. You will learn more from this than from most of the reports in your analytics platform.
03
Look at your server logs – your analytics engineer will know what to do
AI systems like GPT Bot and Perplexity Bot crawl your website to pull content into their answers. Those crawlers leave a trace in your server logs. Looking at those logs tells you which content is being used, by which systems, and how frequently. This is not an SEO exercise. It’s intelligence about where your thinking is influencing AI answers before any human has clicked anything. If you don’t have someone looking at this, ask your analytics or dev team to start.
04
Build a pipeline model that doesn’t start at the form
The most important change you can make to how you report on marketing is also the simplest: ask better questions in discovery. Train your sales team to capture how the buyer first heard about you – not just what they last clicked. Make it a required field in your CRM. Then use that data to build a model of qualified pipeline conversations that accounts for brand exposure, not just last-touch channel. It won’t be perfect on day one. But it will be far more honest than what you’re using now.
05
Add a branded search column to every channel report
This is structural, not just a metric swap. Every channel – paid, organic, content, events, PR – should now be asked the same question: did this move branded search? A campaign that drives traffic but leaves branded search flat has bought attention without building demand. A campaign that barely moves direct traffic but lifts branded search by 15% might be doing exactly the right job. If your current reporting doesn’t show this, it’s giving you an incomplete picture.
None of this requires a new platform or a six-month implementation. It requires a decision about what you’re willing to hold youeself accountable to.
The agency problem no one wants to say out loud
Here’s the uncomfortable and necessary thing to say. If your agency is optimising your campaigns to last-click conversions, they are – without meaning to – making it harder for you to see your brand working. Not because they’re doing a bad job. Because the system they’re optimising to is set up to ignore everything that happens before the click.
Think about what that means in practice. Brand impression share? Not tracked. Content that gets cited in AI answers and shapes buying intent weeks before a form is filled? Not tracked. Organic articles that build category authority and get read by AI crawlers training future answer engines? Not tracked.
Everything upstream of the click becomes invisible. And if it’s invisible, it gets cut. Which means the very activities that build long-term demand get defunded in favour of activities that generate cheap, late stage clicks that look good in a dashboard but don’t represent growing demand.
This is why the ‘brand budget versus demand budget’ argument has always been a false choice. It’s a split that was invented for organisational convenience – not because buyers behave that way. A buyer doesn’t experience ‘brand activity’ in week one and ‘demand activity’ in week six. They experience a series of impressions, ideas, and encounters that gradually shift their thinking. Brand and demand are one motion. Measuring them separately makes it impossible to understand how that motion actually works.
In a zero-click world, that separation doesn’t just produce bad data. It produces decisions that actively work against each other. The brand team invests in content that builds AI answer authority. The demand team pulls back on branded spend because ROAS looks weak. The net result is the demand your brand team created never gets captured. Both teams look less effective. Both budgets shrink in a downward spiral of marketing’s influence in the boardroom.
The fix is to measure them together. Shared KPIs that span brand and demand. Branded search lift as a metric every team is accountable to. Share of voice in AI answers reported alongside pipeline velocity. Pipeline modelling that includes brand exposure, not just last touch.
Some agency partners will embrace this. The good ones will already be thinking along these lines. Others will push back – because unified measurement makes it harder to attribute success to a single channel and harder to justify siloed retainers. That conversation is worth having. The answer tells you a lot about whether your agency is set up to help you succeed in the next two years or the last two.
Reader, you’ve got roughly one planning cycle to get this right before it becomes a board-level conversation. The CFO is not going to keep accepting ‘strong top-of-funnel activity’ as a substitute for pipeline evidence. The tools and the approach exist to do this properly. The question is whether the measurement culture inside your team – and your agency relationships – is ready to change.
Last-click attribution is not going to get more accurate. The click is not coming back. The only real question is whether your analytics stack is going to tell you the truth about the demand you’re building – or a flattering story about the clicks you happened to capture.