The Pros and Cons of AI and PPC Automation in Modern PPC Technology

Why am I getting clicks, but not customers? 

That question has crossed everyone’s mind at least once. For paid media and specifically PPC (Pay Per Click), artificial intelligence and automation have the potential to be incredible but only with some direction from humans. With the evolution of ppc technology, advertisers now have tools that can both simplify management and amplify results. 

Artificial intelligence and automation have been in paid media for years first as suggestions and defaults  now they’re taking center stage. Think Responsive Search Ads, Dynamic Search Ads, automated bidding, Performance Max. On Meta, Advantage+ with regard to audience targeting, placements, and (more so than ever) creative assets. If you’ve run PPC over the past 5 years whether you opted into these technologies or they were turned on by default  you’re already using AI. Harnessing ai in ppc effectively is now less about the novelty of the tool and more about how humans guide its application. 

Sonnet treats AI and automation as power tools: awesome for speed and scale; deadly if applied without craftsmanship. Below is an outline of our view regarding each side and how we apply both with a human at the helm of the controls. 

What “AI + automation” really means in PPC  

Finding patterns at a large-scale: These systems are learning about your search queries, creative variations, and user behavior in order to predict what ad to display, to whom to display it, and how much to pay for that display. Modern ppc automation allows these predictions to occur at unprecedented scale and speed, unlocking opportunities previously unreachable with manual methods. 

Building dynamically: RSAs (Responsive Search Ads) are combining copy to build an ad. DSAs (Dynamic Search Ads) are mapping a website’s landing page to a specific user query. Performance Max ads are creating a blend of inventory and signals across all of Google.

Budgeting and bidding: Some strategies will automatically adjust bids in real-time (i.e. tCPA / tROAS). Others will allocate budget to what is “working” according to the platform (i.e. CBO / Advantage+). 

Lifting reporting: Automated connections to other platforms/dashboards can bring multi-channel data together to visualize trends and take away the time we would normally be spending on wrangling spreadsheets. By integrating ai in ppc into reporting pipelines, we can quickly spot actionable trends without losing sight of human strategy.

Yes they are useful but no matter how good the tool may be it is still people who determine the strategy. 

The upside 

Quick-to-respond to signals 

Automated test cycles reduce how much time is spent testing RSAs that evaluate headline/descriptions as well as smart-bidding micro-signal (device, timing, context) that humans are unable to identify. 

Long-term demand and new search terms/queries 

DSA and PMax provide visibility into demand that is typically unstaffed with manual advertising group creation. Advantage+, available on Meta, identifies adjacent audiences that may not be possible to target by a human. This is the power of advanced ppc technology in uncovering untapped opportunities. 

Increased operational efficiency 

Automated reports enable easier access to consolidated data across channels, enables the building of reusable views and allows for more time to analyze “Why” rather than gathering “What”. 

Adaptability in response to market fluctuations 

When there are changes in market conditions (e.g., seasonal, creative burnout or auctions), automated systems respond immediately. In comparison, humans do not. 

The downside  

Black-box drift 

Platforms are optimized toward their own proxy metrics; what is “best” in the platform doesn’t always align with “best” for the company; you can get a low CPA (Cost Per Acquisition) and still miss a qualified lead pipeline. 

A creative blandness 

Auto-generated or overly mixed content can strip away a brand’s uniqueness; emotional payoff of an ad will be replaced with “safe” messaging that could easily pass as any other company. 

Data dependence & bias 

Automations only know about the data from the previous time period. Therefore, if last month you overserved a particular segment of your audience, the automation is likely to double-down on that segment and possibly starve another high value cohort of opportunities to see your ads. 

Loss of control at the edge 

Pmax or Advantage+, while they offer additional reach, can make it difficult to understand exactly how your budget is being spent; placements, search terms, and the actual demographics of your audiences can be difficult to audit without tightly defined guardrails. 

Sonnet’s stance: automation with a conscience 

In our white label agency, we aggressively apply AI and automation wherever possible to help leverage human judgment where it matters. Sonnet has created its own “tool box” of tools based on real-world campaigns Sonnet has executed, rather than simply applying theoretical ideas or best practices gleaned from reading documentation. 

Crafting of Google Ads: We treat RSAs as modular stories. While Ad Strength does matter, we will never sacrifice meaningfulness for the sake of achieving a green bar. In addition to prioritizing relevant keywords to ensure relevance, we include unique selling points (USPs) into each campaign so both Google and the customer are happy. 

Scaling with a seat belt: We use tCPA and/or tROAS bidding models, however, we also track the rate of learning, the query mix and marginal CPC. With years of experience working in channel marketing, we know how to determine when to scale, switch strategies and set spend caps, rather than allowing the system to “learn.” This careful application of ppc automation ensures efficiency without sacrificing control. 

Performance Max with guardrails: Performance Max is great for coverage but can be very difficult to manage if you want to maintain control over your campaigns. To provide some level of control we organize our asset groups around the intent behind the group, we provide high quality first party signal input into each group and we regularly review search term insights. When visibility drops below our minimum threshold we restrict spending on Performance Max and move it to Search and Shopping breakout campaigns. 

Meta with creative control: We will utilize Advantage+ for audiences and placements whenever we believe it provides a tangible benefit to our advertising, however, we always default out of Advantage+ Creative. Ultimately, creative decisions related to brand messaging and pacing remain in the hands of humans. 

Time-saving reporting: As a result of implementing automated dashboards that pull cross-channel data and build-out the views we commonly use weekly, we have been able to save the time formerly spent creating those views and now invest that time in interpreting the data and developing new strategies, rather than generating additional screen shots. 

A simple framework: H.I.T. — Human-In-the-Traffic 

This is for evaluating which parts of your marketing and advertising strategy should be automated and which parts need to be driven by a strategist. 

H – First, write the Human hypothesis 

Write the human hypothesis: audience, job-to-be-done, promise, proof. The automation will test how to deliver the human hypothesis, but not what the human says. 

I – Input that you can rely on 

Input: Clean input: tight themes, purposeful assets, product-led value prop, negative list, conversion quality signals. If your input is unclean, then the automation will learn to operate in an unclean environment. Effective ai in ppc depends entirely on these clean inputs. 

T – Tightly coupled feedback loops 

Review weekly: Query your themes, Asset Rotation, Placement Quality, Lead Quality. Drive changes to the structure of the program — not the budget. 

Where we deliberately don’t automate 

Human creativity and storytelling in strategy and narrative development: While we will use data to inform our creative iterations, the ideas for stories come from people. 

Test offers: Algorithms will never create a better test offer than a human can, only an automated system will be able to optimize a poor one at a much quicker rate. 

Manual measurement architecture design by default 

We keep the core framework in human hands to avoid black-box gaps: 

• UTM governance and naming conventions 

• Offline conversions mapped back into the ad platforms 

• Consent and privacy settings configured intentionally, not blindly 

• Spend caps and budget guardrails on assets/placements, so automation doesn’t quietly overspend or drift off-brief 

Quick operator checklists 

Before Automation Is Activated. 

A single sentence that defines your campaign (Who is this campaign targeting? Why should I do this campaign now? What will be the results of this campaign?) 

• The search structure or asset grouping(s) linked to each tier of intent. 

• Queries/placements that MUST appear AND NOT appear in your search results. 

• Your primary and secondary conversions are both tied to a specific business objective. 

• Your baseline dashboard is in place so you can determine if the “learning” improves your campaign’s performance. 

During the Campaign 

• Check the query mix (Are we getting traffic from branded searches just to make our numbers look better?) 

• Track CPA/ROAS margins, as well as average CPA/ROAS. 

• Check the contribution of each asset to the overall campaign performance (Which assets are really driving conversions?). 

• Check the quality of leads (Is the quality of the leads increasing with volume?), not just volume. 

• Document all changes. Allow systems sufficient time to learn before intervening. 

When To Pivot 

• You may see creative fatigue (your ads are getting tired) before the algorithm tells you they are fatigued. 

• Your query themes are deviating from your intent. 

• Your platform metrics improve but your downstream performance declines. 

• CPA/Cost Per Incremental Conversion increases on a weekly basis. 

A small, real scenario  

An online retail company relied heavily on PMax to grow their user base and were seeing average metrics improving as well as increasing revenue. However, they began to notice that new customers were slowing down and blended Cost per Acquisition (CPA) was also rising. The company decided to break PMax out based on customer intent, rebuild the feeds to better match those intents, and added an automated Search Layer for users searching high value terms. The company chose to keep the automated bidding in place, tighten the negatives and refresh the creative using human written ad copy angles. After making these changes, the company saw their Return On Ad Spend (ROAS) became much more consistent, they saw a healthy increase in new customers and the ads started sounding like the actual brand not templates. This scenario demonstrates how ppc automation, guided by human strategy, can maximize both efficiency and results. 

The Point 

Automation is supposed to allow us to regain some of the time that we have lost as digital marketers and use it to focus on things that only humans can do, such as sharpening our offers, writing more true ads, making more smart decisions about how we will allocate our budgets across different channels and aligning our marketing efforts with real business objectives.  

At Sonnet, we want to use the machines to help us be more efficient and to help us get there faster , without allowing the machines to dictate who we are or where we are going.