If you manage Google Ads accounts, being “good” with AI is expected, but employers probably never checked (or didn’t know how to check) if you are. Being able to show your judgment is a career must if you can, and a risk if you can’t.
Fresh research from CodeSignal surveyed 1,000 U.S. employees and 1,000 U.S. people managers about AI readiness, and two numbers sit side by side in the results. 72% of workers surveyed say their AI skills have never been formally evaluated, yet 69% of managers call AI ability a major or moderate factor in hiring. Indicating that employers are hiring for a skill they mostly don’t measure.
I sent CodeSignal 15 questions about how solid those numbers are and what practitioners should do about them. The written answers were more candid than I expected, including on the obvious conflict of interest of an assessment vendor concluding that nobody assesses anything. Here’s what holds up, what needs a footnote, and what it means for people who run paid search for a living.
Industry observers note that start with what this survey is and isn’t. CodeSignal commissioned two online surveys through Pollfish in September 2026, and both samples were entirely U.S.-based. The firm defined knowledge workers by the work they do rather than by job title, meaning people whose jobs run mainly on computers and digital tools. It also didn’t screen out people who skip AI. 12% mentioned they never use it at work.
In a recent development, codeSignal reports that roles where AI use is required or expected “jumped 78%.” I asked whether that meant percentage points. It doesn’t. It’s a relative increase, self-reported by employees about their own roles, measured across each person’s time in their current job. When respondents started those jobs, 9.4% remarked AI use was required or expected of them. Today, 16.7% say so.
According to recent reports, that’s nearly double. It’s also fewer than one in five.
Paid search people are probably well past that curve already, since automated bidding has been Google’s default recommendation for years. So, the useful question for our field is whether anyone can tell who is using AI well.
See also: From T-Shaped To M-Shaped: The PPC Career Evolution Nobody Is Talking About.
It has been revealed that 62% of employees told CodeSignal they received employer training on how to use AI at work. Only 35% said employer training is where they built the AI skills they actually use.
Industry observers note that I asked whether most of that training simply wasn’t about AI, or whether workers weren’t counting it. CodeSignal noted neither. The training was specifically about using AI at work, and the firm reads the gap as a sign that it wasn’t substantial or specific enough to apply on the job.
The follow-up number is worse. Only about one in four trained workers said they can apply most or all of their employer’s AI training. The other 74% apply some, very little, or none of it. Run that against the full sample and roughly 46% of all workers got AI training they can’t fully use. When CodeSignal asked what gets in the way, employees named lack of time, training that doesn’t fit their role, and lack of access to relevant training.
So, where are people learning? The question allowed multiple answers, which is why the figures don’t add up to 100. Free resources people found on their own led at 46%, followed by trial and error on the job at 39%, and employer training at 35%. Colleagues came in at 22%, and courses people paid for themselves at 12%.
Tigran Sloyan, CodeSignal’s co-founder and CEO, noted, “Companies are buying AI courses or building their own, but if most workers are going elsewhere to learn, you have to wonder if that training is actually helping.”.
Paid search has worked this way for a long time. Plenty of PPC managers learned bidding by watching their own accounts and reading practitioners on LinkedIn and X. What has changed is that the thing they now learn by trial and error is the ad platform’s own AI, and each error is paid for out of a client’s budget.
According to recent reports, 72% of workers say their AI skills have never been formally evaluated, but 66% say the same about their skills in general. That’s a six-point gap.
It has been revealed that I asked CodeSignal whether the real story is that companies don’t assess much of anything, and AI just makes that visible. The answer began with “Yes, exactly.” Companies have long made people decisions on proxies such as resumes, job titles, tenure, and manager impressions. CodeSignal’s argument is that those were always weak signals, and they get costly when roles shift in months. More than six in ten managers in the survey said they make people decisions based on assumptions rather than reliable data.
This bothers me more in paid media than almost anywhere else. We measure campaigns to the decimal point. We argue about attribution windows, incrementality tests, and conversion lag. Then we judge the people running those campaigns with a once-a-year “meets expectations.” For a discipline built on measurement, that’s a strange double standard, and AI is making it harder to defend.
Recent findings suggest that codeSignal says the hiring side already indicates the strain. When AI fluency is a criterion nobody has measured, people get judged on what’s easy to see, such as who is most visible or sounds most confident, instead of on the quality of their work.
It has been revealed that anyone who has sat through an agency pitch knows how that plays out. The person who says the automation will figure it out can sound more fluent than the person who spent the afternoon reading search terms and adding negatives. Only one of them is protecting the budget.
It has been revealed that “A resume is a story someone tells about their skills,” Sloyan remarked. “The work is the proof. Right now, most companies are making their biggest decisions about people based on the story.”.
See also: PPC Automation Layering: How Smart Advertisers Combine Automation With Strategy.
Sources indicate that I asked CodeSignal to separate “AI fluency” from AI literacy and prompt-writing, because those terms get used as if they were the same thing. The organization drew the lines this way. Literacy means understanding what AI is. Prompt-writing is a tactic. Fluency means knowing what to delegate to AI, checking and correcting its output, and recognizing when not to use it at all.
That is the best short job description for a modern PPC manager I’ve read this year.
Map it onto a Google Ads account, and each part becomes something you can point to. Delegation is deciding which campaigns get Smart Bidding, which conversion actions you feed it, and what budget it gets to learn on. Checking is reading the search terms, placements, and automatically created assets that the system produces, then correcting what’s wrong. Knowing when not to use AI is the judgment call that rarely demonstrates up in a dashboard. That might mean holding back a campaign with too little conversion data, or keeping a human hand on ad copy in a regulated category.
Industry observers note that none of that is visible in a certificate. All of it is visible in the account.
See also: The Conversion Setup Errors That Break Smart Bidding.
I also asked CodeSignal to critique its own product category. What can a timed, simulated assessment not tell you about whether someone uses AI well on the job? The company answered that an assessment is only as useful as what it was designed and validated to measure. To understand how someone performs over time, it stated, you need to look at their actual work.
Sources indicate that that answer applies just as well to the credential most PPC people already hold. Google’s AI-Powered Performance Ads Certification on Skillshop is free, and third-party study guides describe it as a timed, 50-question exam. It’s worth taking, because it teaches how Google wants its AI campaign types used. But it’s written by the platform whose recommendations a good PPC manager sometimes overrides. Passing it shows you know Google’s playbook. It can’t show the afternoon you decided not to follow it, or what happened to cost per acquisition when you didn’t.
Industry observers note that then there’s the conflict of interest. CodeSignal sells assessments, and its survey concludes that skills aren’t being assessed. I asked how readers should weigh that, and what finding cut against its business.
In a recent development, the organization remarked its interest in skills is the reason it commissioned the study. Then it volunteered a number that does cut the other way. 82% of employees said their manager understands their skills very or somewhat well. CodeSignal called that knowledge real and valuable. Its caveat is that the knowledge is informal, varies from manager to manager, doesn’t follow people when they change teams, and can’t be compared across a company.
Sources indicate that agency people will recognize that last part. When an account manager moves to another client team, everything their old boss knew about their judgment usually stays behind.
Industry observers note that I give CodeSignal credit for offering up the 82%. I’d still want to see the full questionnaire before treating any of these figures as settled, and the organization’s launch announcement for its Agentic Skills Intelligence product doesn’t describe the survey method at all. The details above came from its answers to my questions.
CodeSignal’s advice to individual workers was blunt. Saying you use AI for everything doesn’t prove much. Fluency is being able to explain exactly how you use it and, just as important, when you don’t. For a PPC manager, that turns into a short routine.
If you lead a PPC team or run an agency, the manufacturer’s advice for a 200-person business with no assessment budget translates well. Start with managers, since most employees think their managers already understand their skills. Then give those managers something more concrete than “meets expectations” by writing down what good AI use looks like for each role on the team. A specialist’s standard might be catching irrelevant search terms within a week. A strategist’s might be deciding when a client’s conversion tracking is clean enough to move to value-based bidding.
Industry observers note that next, look at real account work before training and again after it. A sample of a few accounts per person is enough to see whether anything changed.
According to recent reports, find your self-taught experts, too. With 46% of workers learning from free resources on their own, someone on your team has probably already worked out what the automation gets wrong. Have them show everyone else.
According to recent reports, finally, build training around the work people do every week and give them time to practice. Lack of time was the obstacle employees named first. Track whether people use what they learned, not whether they finished the course. A Skillshop badge is a completion metric, and you’d never let a client judge a campaign on impressions alone.
Sloyan’s version of this is worth quoting. “Training only works when it’s built around the job someone actually does.”.
Recent findings suggest that my view is that PPC managers have an advantage most knowledge workers in this survey don’t. Our work already leaves a record. Every bid strategy change, every negative keyword, and every recommendation we dismissed is logged somewhere, with performance data attached.
According to recent reports, most of us treat that record as housekeeping. The managers who turn it into evidence of judgment are the ones who will get promoted while their employers are still figuring out how to measure AI skills at all.
Everyone else will be telling a story.
Featured Image: ImageFlow/Shutterstock.
