There is a version of Google Ads automation that works brilliantly: Smart Bidding adjusts bids in real time based on signals no human could process at scale, and it genuinely improves CPA when given clean conversion data. But automation has a tendency to expand beyond its most useful territory. Device targeting and demographic targeting are two areas where platform defaults have gradually pulled control away from advertisers - often without anyone noticing.
Across the industry, PPC managers are asking for clearer guardrails around automation. They are not rejecting it outright - they are pushing back on how broadly it is applied. That distinction matters.
Why Device Targeting Still Matters in an Automated World
Smart Bidding does factor in device type as a signal. If your conversion rate on mobile is lower, the algorithm should theoretically reduce bids for mobile traffic. The problem is that this adjustment happens within the bounds of what the campaign allows, and those bounds are often set too wide by default.
For lead generation advertisers in particular, device mix has a direct impact on lead quality, not just lead volume. A B2B advertiser selling high-value professional services will typically see longer form completions, better qualified enquiries, and higher close rates from desktop traffic. Smart Bidding optimises towards conversion events - but if your conversion tracking fires on a basic form submit, it cannot distinguish between a mobile user who half-completed a form from a tablet on the sofa and a desktop user who spent twelve minutes reading the service page before converting.
This is not a criticism of the algorithm. It is a tracking and signal quality problem. Until offline conversion imports or CRM-matched data closes that quality loop, device-level targeting constraints are a practical way to steer spend towards the traffic that historically generates better outcomes.
Demographic Targeting: The Control That Performance Max Obscures
Standard search campaigns have long supported demographic bid adjustments - age, gender, parental status, household income. Advertisers can apply positive or negative bid modifiers, or exclude segments entirely. In practice, many accounts have never touched these settings, partly because search intent was assumed to be a stronger signal than demographics.
Performance Max complicates this considerably. Audience signals in PMax are recommendations, not restrictions. You can tell the campaign who you think your customers are, but you cannot stop it serving ads beyond that group. The campaign will test broadly and allocate budget based on conversion probability - which sounds reasonable until you consider that it may be converting the wrong demographic at volume, skewing your CPA data and distorting future bidding decisions.
For advertisers with genuine demographic constraints - age-restricted products, services with regulatory requirements around who can be marketed to, or simply businesses with a well-defined and validated customer profile - the inability to apply hard demographic exclusions in PMax is a real structural issue. Demographic bid adjustments in standard campaigns at least let you apply a negative modifier. In PMax, the levers are softer.
The Practical Case for Explicit Targeting Constraints
The argument for letting automation handle device and demographic allocation is straightforward: the algorithm has more data than you do and can react faster. That is true. The counter-argument is also straightforward: the algorithm optimises for the conversion events you have defined, and if those conversion events do not accurately represent business value, you will get efficient delivery of the wrong outcome.
Explicit targeting constraints serve a different purpose to bid optimisation. They are about business logic, not algorithmic efficiency. If your service is only available in the UK to adults over 25, there is no reason to serve impressions to under-25s and rely on the bidding system to self-correct. If your sales team consistently reports that mobile leads take longer to close and convert at a lower rate, reducing mobile exposure is a commercial decision, not a technical one.
Documenting these constraints clearly - and building them into campaign structure from the start - also makes performance analysis cleaner. When you can isolate device or demographic cohorts with confidence, you can feed more accurate data back into your bidding strategy and your creative planning.
How to Approach This in a Mixed Campaign Environment
Most accounts running in 2026 are running a mix of campaign types: Performance Max sitting alongside standard search, possibly with Demand Gen in the mix. Each campaign type has different levels of native control over device and demographic targeting. Managing this consistently across an account requires deliberate structure.
In standard search campaigns, review your demographic and device bid adjustments and treat them as active decisions rather than defaults. If you have not touched them, you are implicitly accepting the platform defaults - which is a choice, but not a considered one. Pull segmented performance data by device and by age bracket over a meaningful time window. If there is a clear pattern in CPA or conversion rate, act on it.
For Performance Max, the tools are more limited natively. Audience signals should reflect your actual best-customer profile, based on first-party data where possible - Customer Match lists, high-value converters from GA4, or CRM segments. You cannot enforce demographic exclusions, but you can provide strong positive signals that influence where the campaign focuses. If device skew in your PMax campaigns is causing problems, a structural approach - separating high-intent search traffic into standard campaigns where you can apply device controls - is often more effective than trying to manage it within PMax alone.
Closing the Loop With Better Conversion Data
None of this targeting work operates in isolation from measurement. If your conversion tracking is capturing form submits at the top of the funnel but not feeding in what happens downstream - whether leads became customers, which demographic or device cohort had the best close rate - then the algorithm will keep optimising towards the wrong signal regardless of how carefully you set your targeting parameters.
Offline conversion imports via the Google Ads API, or enhanced conversions connected to your CRM, allow you to pass actual business outcomes back to the platform. When you do that, Smart Bidding can start to distinguish between a lead that converted to a customer and one that did not. At that point, the tension between algorithmic control and manual targeting constraints starts to resolve itself - because the algorithm finally has the right data to work with.
Until that tracking infrastructure is in place, explicit device and demographic controls are not a workaround for a broken system. They are a sensible way to manage risk while you build towards better signal quality.