The conversation in paid search right now is about AI. AI Max, agent search, automatic bidding. The skills are really impressive and the excitement is worth it. But there’s a precondition that rarely makes it into the conversation: none of it works well if your measurement infrastructure is broken. Feed the smart bidding algorithm imperfect data and make bad decisions with confidence. AI works as designed. The problem is what you feed it.
The scale of the infrastructure gap is bigger than most people think. A recent Google study found that 84% of CMOs make budget decisions based on ROI data, but only 28% of agency teams running those campaigns believe they can accurately measure ROI. Money flows based on data that people generate that they don’t trust. And the AI features that sit on top of that data evolve into a version of reality that doesn’t exist.
One in five conversions are not counted
The most common source of measurement loss is also one of the least discussed. Safari and Firefox do not allow the Google Tag to run as a third-party script. For any user who arrives at the client’s website from any browser, the tag cannot collect conversion data. The visit happened, the purchase happened, but the rating system didn’t see it. On average, that causes more than 20% data loss before another challenge is involved.
That 20% shortfall flows directly into how AI campaigns behave. Intelligent bidding algorithms make choices about where to spend your budget making those choices based on imperfect information. Keywords are considered important. The audience gets the lowdown. Attribution models show a picture that is always darker than the truth. The algorithm does exactly what it is asked to do – it is not given the full picture.
Why does this keep happening
Measurement infrastructure is rarely prioritized. Fixing it requires access to website tags, development resources, permission frameworks, and a willingness to spend time fixing the invisible before developing the visible. The path of least resistance, for most groups and most agencies, is to go ahead with the campaign and treat measurement as good enough.
The problem is that good enough is hard to sustain. Signal loss builds up slowly over the years, combined with all browser updates and privacy regulations. And as AI factors make more decisions in campaigns, the quality of the data they feed them is more important, not less. A manual campaign by an experienced clinician can compensate for gaps in judgment. An AI-powered campaign can’t.
What a fix looks like
Google has been putting a set of tools under the umbrella name of Data Power – the measurement infrastructure that must be in place before layering on AI-enabled campaign features. Features are not new: Permissions Mode, Advanced Conversions, and Google Tag Gateway have been available for some time. What has changed is the recognition that they need to be treated as a foundation, not an optional add-on.
In this case, Google Tag Gateway directly addresses the problem of signal loss. Think of it as a tag VPN: instead of Google Tag loading from googletagmanager.com (which Safari and Firefox identify as third-party and block), it loads from your first-party domain. The browser recognizes the first company’s script, transmits the data, and the measurement gap closes. Implementation is fast compared to your development service, but someone needs to make it a priority.
Advertisers who used Google Tag Gateway saw, on average, a 14% increase in tracked conversions from conversion alone. That is not an incremental improvement. Those are your campaigns that finally show what they’ve been doing.
Why is it more important than performance
There is a commercial effect of getting this wrong that outweighs the effectiveness of the campaign. Last year, 61% of clients who left their agencies cited dissatisfaction with ROI. A significant part of that dissatisfaction probably had nothing to do with campaign performance. The gap was moderate, and no one raised the alarm. If you don’t know that the data is incomplete, the inefficiency appears to be a strategic problem rather than an infrastructure one.
Managing Data Power as a routine audit is worth seeking out for your agency. Running through your accounts to identify where the signal is leaking, where the approval parameters are incomplete, where Advanced Conversion was not used, is the type of activity that creates written evidence that your investments are being carefully managed.
AI in paid search is only as good as the data it works on. Get the basics right before chasing features, and you’ll get more out of both.
