I've been in marketing long enough to watch several waves of "required skills" come and go: social media management, growth hacking, data-driven decision making. Each one got added to job descriptions before most hiring managers understood what it meant in practice, and each one produced years of bad hires before the industry figured out how to evaluate it.
AI skills are in that same awkward phase right now, except the gap between what's listed and what's tested is wider than anything I've seen before.
The Checkbox Problem
Open five marketing job descriptions right now (senior content manager, SEO lead, digital marketing director) and most of them will include something like "experience with AI tools" or "proficiency in ChatGPT/Claude/Gemini" in the requirements.
Now ask the average hiring manager how they plan to evaluate that requirement in the interview.
Most of them don't have an answer. They'll ask "have you used AI tools in your work?" and accept "yes, I use ChatGPT regularly" as a complete response, which is like asking someone if they own a blender and taking that as proof they can cook.
The mistake is treating AI proficiency as a software skill, like knowing Excel or Figma. Those tools have defined feature sets and fairly clear mastery thresholds. With AI tools the ceiling is undefined, the floor is zero, and most people are operating somewhere in the lower third without knowing it.
What "Using AI" Actually Means
When a candidate says they use AI in their marketing work, they could mean any of these, and they are very different:
Level 1: They paste a brief into ChatGPT, accept the output with light edits, and call it done. The content is passable and it saves maybe 30 minutes. Most people who "use AI" are here.
Level 2: They have a personal prompt library and know which models handle which tasks better. They edit AI output heavily, cutting what's generic and keeping what adds value. They've probably been burned by Level 1 a few times.
Level 3: They've built multi-step workflows where AI handles specific functions (research aggregation, first drafts, structured data extraction) and a person applies judgment at defined checkpoints. Output at this level is qualitatively different from Level 1.
Level 4: They build tools: scripts, agents and automations that extend what one person can get done in a day. This is a small group. It takes some technical skill that most pure marketers don't have, though less than people assume.
A hiring manager asking "do you use AI?" will get yes from all four groups and have no way to tell them apart. That's a failure of the hiring process rather than the candidate pool.
The Specific Mistakes I See in Job Descriptions
Listing AI tool names instead of AI capabilities. "Experience with ChatGPT, Jasper, and Copy.ai" is not a meaningful requirement, because these tools change monthly. A marketer who understands AI as a system can pick up any specific tool in an afternoon. A marketer who only knows Jasper is barely more prepared than someone who knows nothing.
A better framing would be "Demonstrated ability to integrate AI-assisted workflows into content production, with measurable output on quality and throughput."
Conflating familiarity with fluency. "Familiarity with AI tools" and "ability to leverage AI to [specific outcome]" are different job requirements: the first means you've heard of it, the second means you've done it. Most job descriptions use the first and then wonder why the hired candidate isn't doing the second.
Not testing it in the interview. If AI skills are a genuine requirement for the role, give the candidate an AI task as part of the interview. Skip "tell me about how you've used AI"; sit them down (or send a take-home) and say "here's a brief, show me what you'd do with it." You'll learn more in 20 minutes than in an hour of behavioral questions.
I've heard the objection that this isn't fair to candidates who haven't used these tools as much. If the role requires it, though, finding out who can actually do it is the purpose of the test.
Treating AI as a bonus skill instead of a core multiplier. Some job descriptions still have AI buried at the bottom under "nice to have." For roles that involve any content production, SEO, paid media copywriting, or research at scale, that's backwards. On the right tasks, the productivity gap between a marketer who uses AI effectively and one who doesn't is large enough to change how a team is staffed. Treating it as a bonus is like making "ability to use Google" optional.
What the Strongest AI Users Have in Common
I've worked alongside a lot of marketers at this point. The ones who use AI well share a few traits that have nothing to do with which tools they know.
They have strong editorial judgment. The best AI users are the hardest on AI output. They can see where the model is coasting (generic transitions, vague claims, lists standing in for analysis) and they cut it. Being a better writer makes you a better AI user.
They think in systems. The most useful AI work comes from repeatable workflows rather than one-off prompts: a process for researching a keyword cluster, a template for turning an interview transcript into three content formats, a pipeline for updating old articles at scale. Marketers who work that way get compounding returns from AI, while marketers who go task by task get linear ones.
They can explain what they built. If someone can't articulate the specific workflow they set up, the prompting strategy they use for a recurring task, or the iteration it took to get output they were happy with, they probably haven't gone that deep. People who've built these workflows like talking about the details, and vague answers usually mean "I opened ChatGPT a few times."
They've hit the walls. Good AI practitioners have a mental catalog of failure modes. They know that AI hallucinates statistics confidently, that it defaults to "on the one hand, on the other hand" structure when you want a clear stance, that long prompts produce worse outputs than modular prompts in sequence. You only know these things from experience.
A Better Interview Framework
If you're a hiring manager and want to evaluate AI skills properly, this is what I'd do.
1. Pre-screen with a workflow question. "Walk me through the last time you used AI to accomplish a specific marketing task, step by step." Listen for specifics; vague answers tell you something.
2. Live task test. Give them a real brief: a blog topic in your industry, a target keyword, a rough audience description. Give them 20 minutes and access to any AI tool they want. Then look at what they produced and how they'd edit it before publishing.
3. Stress test the output. Take whatever they produced and ask them to critique it. "What would you change before this went live?" Great AI users will have an immediate, specific list. People who accepted the AI output at face value often can't answer this well.
4. Ask about scale. "If you needed to produce 50 pieces of content this quarter instead of 10, how would your process change?" A good answer involves workflow design and probably some automation. "I'd just use more AI" isn't one.
The Bigger Issue
Underneath the bad interview questions, most organizations haven't defined what good AI use looks like for their own marketing function, so they're evaluating a capability without a rubric.
The companies doing this well have marketing leaders who use AI themselves, who have opinions about where it helps and where it doesn't, and who can recognize quality output when they see it, and they hire differently.
The companies doing this poorly are copying AI requirements from other job descriptions, checking the box, and then being frustrated when the new hire doesn't transform their content operation.
If you're a marketer and feel like your AI skills aren't getting recognized, you might be working at Level 3 and interviewing with people who can only recognize Level 1. The fix is to be concrete about what you've built instead of general about what you've "used": show the workflow, the output, and the difference between what you can do and what your team was doing before.
I've spent 13+ years in SEO, I lead SEO at an agency, and I build AI agents. If you want to talk through how to evaluate AI skills on your team, or how to demonstrate yours, I'm reachable through this site.