Eight hundred million people use ChatGPT every week. That is roughly 10% of the global adult population (OpenAI/Harvard, July 2025). Among U.S. knowledge workers, 43% use AI tools at work, and 28% of employed adults specifically use ChatGPT in their jobs (Pew Research, via OpenAI).
When a candidate tells you "I use ChatGPT" in an interview, they are telling you roughly the same thing as "I use email." It confirms that they have adopted a tool that hundreds of millions of other people have also adopted. It tells you nothing about how they use it, whether they use it well, or whether they have the judgment to use it safely in a professional context.
And yet recruiters continue to treat this statement as a meaningful signal of AI competence. It is not.
Consider the equivalent in any other domain. "I use Excel" does not tell you whether someone can build a financial model, write a macro, or merely open a spreadsheet and sort a column. "I drive a car" does not tell you whether someone is a safe driver, a skilled one, or a risk on the road. The statement describes adoption. It says nothing about capability. With AI, this distinction is more consequential than with any previous tool, because the risks of poor AI use are invisible until they cause damage.
What "I use ChatGPT" actually means
OpenAI's own research, conducted with Harvard and published in July 2025, analyzed 1.5 million ChatGPT conversations and found a clear picture of what people actually do with the tool. Three-quarters of all conversations fall into three categories: practical guidance, seeking information, and writing. For work-related use specifically, 56% of messages involve task execution, and nearly three-quarters of those are writing tasks.
The typical ChatGPT user, in other words, is asking questions and getting help with drafts. They are using the tool the way most people use it: as a faster way to get a first draft, an answer to a question, or a summary of something they do not want to read.
This is useful. It is not the same as AI readiness.
55% of professionals use AI at least weekly (Section AI, 2026). But less than 3% have progressed beyond basic prompting into what the report calls "integrated" or "advanced" use, the kind of use that involves systematic verification, contextual judgment about when AI is appropriate, and critical evaluation of output quality. The gap between using ChatGPT and using AI well is enormous, and the phrase "I use ChatGPT" sits firmly on the wrong side of it.
The five levels of AI use that actually matter
When a candidate says "I use ChatGPT," they could be anywhere on a spectrum that ranges from negligible to genuinely impressive. The challenge is that the statement itself does not differentiate between these levels.
Level 1: Basic prompting. The candidate asks ChatGPT questions and uses the output directly. They draft emails, summarize documents, and generate ideas. This is where the vast majority of users sit. It requires no critical evaluation, no understanding of limitations, and no judgment about when AI is appropriate. It is a productivity tool, not a professional competency.
Level 2: Iterative use. The candidate refines prompts, gives feedback, and improves outputs through multiple rounds. This is a step above basic prompting and indicates some understanding that first outputs are not final products. But it still does not require the candidate to evaluate accuracy, identify hallucinations, or exercise judgment about sensitive content.
Level 3: Critical verification. The candidate treats AI output as a starting point that requires checking. They verify claims, cross-reference data, and recognize when AI-generated content sounds authoritative but may be wrong. This is the threshold where AI use becomes professionally reliable, and it is the level that 38% of business executives fail to reach, having made decisions based on hallucinated AI output (Deloitte, 2024).
Level 4: Contextual judgment. The candidate adjusts their AI use based on stakes, context, and audience. They use AI differently when preparing an internal summary than when producing a client-facing analysis. They recognize that some tasks should not involve AI at all, because the data is sensitive, the stakes are too high, or the domain requires human judgment that AI cannot provide. 57% of enterprise employees have entered confidential information into public AI tools (TELUS Digital, 2025), which means the majority of professionals have not reached this level of contextual awareness.
Level 5: Strategic integration. The candidate understands where AI adds value and where it introduces risk across their professional domain. They can design workflows that incorporate AI productively, establish verification protocols, and help colleagues use AI more effectively. They also understand the regulatory context. With the EU AI Act's Article 4 literacy requirements already in force and enforcement beginning August 2026, strategic AI integration includes knowing what compliance looks like. This level is rare, and it is the level that genuinely differentiates a candidate as AI-ready.
"I use ChatGPT" maps to levels 1 or 2. Levels 3 through 5 are where professional value and professional risk diverge. And the only way to know which level a candidate occupies is to put them in a situation that tests their behavior, not their self-report.
The practical implication for recruiters: if you are interviewing a candidate and they tell you they use ChatGPT, your next question should not be "how do you use it?", because that invites a rehearsed narrative. Your next question should be a scenario. "A colleague sends you an AI-generated market analysis for a client presentation. You have thirty minutes before the meeting. Walk me through what you do." The answer reveals whether they are at level 1 or level 3. That distinction matters more than any claim about tools used or frequency of use.
See the gap for yourself
Take the free Aptivum Snapshot (10 questions, 8 minutes) and find out where you actually stand on AI readiness.
Why the interview answer is unreliable
The problem with "I use ChatGPT" is not just that it is vague. It is that the social incentives in a hiring context guarantee it will be both vague and inflated.
Candidates know that AI skills are valued. They know that claiming AI experience is expected. And they know that interviewers rarely probe beyond the surface claim. In an environment where 79% of tech workers admit to pretending to know more about AI than they actually do (Pluralsight, 2025), it would be naive to take interview self-reports at face value.
Research from Aalto University (2026) found that AI use itself inflates self-assessment: the more people use AI, the more they overestimate their own abilities. This means that the candidates who sound most confident about their AI skills may be the least calibrated. Confidence in AI proficiency is not a signal. It may be an anti-signal.
Even well-intentioned candidates struggle to describe their own AI competence accurately, because most AI use is habitual and unreflective. A person who uses ChatGPT daily to draft emails may genuinely believe they are proficient with AI. They have used it thousands of times. They have gotten useful results. What they have not done, and what they may not realize they have not done, is critically evaluate those results, consider what the tool might have gotten wrong, or think about whether certain information should have been kept out of the prompt.
What to ask instead
If "I use ChatGPT" is not a useful signal, what is? The answer is not a better interview question, though better questions help. The answer is a fundamentally different evaluation method.
Interview questions can probe for level 3 and above. Instead of "do you use AI tools?", ask candidates to walk you through a specific decision they made about AI output. Did they catch an error? Did they choose not to use AI for a specific task, and why? Did they encounter a situation where AI output seemed convincing but turned out to be wrong? These questions reveal behavior. "I use ChatGPT" reveals adoption.
But interviews still rely on self-report, which, as we have established, is systematically unreliable for this dimension. The more robust approach is scenario-based assessment that puts the candidate in a realistic situation: here is an AI-generated document with a subtle factual error, a plausible but fabricated citation, or a recommendation that sounds reasonable but ignores an important contextual factor. What does the candidate do? The behavior reveals their actual level, and no amount of interview preparation can fake a verification instinct that does not exist.
As we explored in our analysis of how AI has decoupled appearance from competence, the traditional hiring process rewards presentation at every stage. "I use ChatGPT" is presentation. What you need to know is what lies underneath.
Find out what lies underneath. Take the free Aptivum Snapshot: eight minutes, five dimensions. See where you actually stand on AI readiness.
