AI is arguably the most disruptive and polarizing technological innovations we’ve seen in recent years. People tend to either be firmly in the AI optimism camp or see AI as a meaningful risk to the way we work and live. But of course, the truth is more nuanced than either extreme. AI is not some big bad villain here to take every job and flatten the employment market. It is also not a silver bullet that can solve every and any problem with a well-worded prompt. AI is a tool, and like every tool, it reflects the effort, judgment, context, and intent that people bring to it.

Two focus groups, one made up of self-proclaimed skeptics and the other of AI enthusiasts, met to discuss their experiences with AI. The members came from Women in Tech SEO and represented a mix of seasoned (10+ years working in digital) and newer (1-3 years and 3-5 years) practitioners, business leaders, and marketing minds. We had representation from both EMEA and the Americas in the cohort.

They discussed where they see value, what gives them pause, and what they believe needs to change. The most useful takeaway was not that the groups were radically different. It was that skeptics had things they were genuinely excited about, while enthusiasts had things they were clearly skeptical of. That tension matters because it reflects where most practitioners actually are: curious, cautious, hopeful, and frustrated all at once. If you feel skeptical about AI and the role it will play in work, school, and life, you are not alone. If you feel energized by what it unlocks while still worrying about misuse, trust, and the impact on human skill, you are not alone either.

Who Participated

The participants represented a broad mix of digital marketers, SEOs, PR, paid media, content marketers, consultants, and in-house practitioners, with experience ranging from roughly 3–5 years to over a decade. They were globally based and brought perspectives shaped by agency work, enterprise environments, independent consulting, and cross-functional marketing roles. This was not a hypothetical or academic conversation. It was a practitioner's conversation grounded in daily use, professional accountability, and the very real pressure of figuring out what AI should do, what it can do, and what it absolutely should not be trusted to do on its own.

Top Three Takeaways

1. AI skeptics and AI enthusiasts are not opposites. Both groups are already using AI in meaningful ways. The real difference is not adoption. It is the level of trust they need before they are willing to hand AI more responsibility.

2. Trust is the make-or-break issue. Across both conversations, participants consistently came back to hallucinations, overconfident outputs, weak attribution, unclear sourcing, and market hype that frames AI as a replacement instead of a support system.

3. The healthiest relationship with AI is human-led. The practitioners who are getting the most value from AI are using it to accelerate research, analysis, drafts, and repetitive work while keeping human judgment, creativity, and accountability firmly in place.

Theme 1: Both groups are already using AI

The first myth these conversations punctured was the idea that “skeptics” are sitting AI out. They are not. Multiple skeptics described using AI every day, testing different models, and even building tools or workflows with it. One skeptic said she uses AI daily and has used it to build tools and automate processes, but only with strong human review because her years of SEO experience help her spot where the model is likely to make mistakes. Another said she tests tools constantly to see whether they can actually do what people claim they can do. The skepticism was not about whether AI should be touched. It was about whether it should be trusted without scrutiny.

The enthusiasts expressed the same level of participation, but with a different emotional posture. They described using AI to accelerate research that once took days or weeks, to build workflows, to create prototypes, to generate creative concepts, and to support accessibility needs.

One participant said AI opened up opportunities for people with dyslexia, difficulty typing, or trouble organizing thoughts. Another described using AI to create a social media assistant for her company. A third talked about using multiple tools together (one for transcripts, another for synthesis, another for presentations) because no single tool does everything equally well. In both groups, AI was already inside the workflow. The difference was not whether people were engaging with AI tools. It was how much confidence they had in what the technology was giving back.

Theme 2: Trust is the central issue

If there was one word that defined both conversations, it was trust. For skeptics, trust breaks when AI hallucinates, sounds certain when it is wrong, or gets used by people who do not have the knowledge to catch errors. One participant said her biggest pause with AI was its “propensity to hallucinate” and the fact that people are not thinking critically about outputs. Another worried about “people trusting it indiscriminately” because some models are so affirming that they make users feel brilliant even when the result is flawed. A third said, flatly, “I wish that if it didn’t know, it would actually tell you it doesn’t know instead of just saying rubbish.” Those comments are not just complaints about accuracy. They are direct statements about credibility. When a system is wrong with confidence, it teaches the wrong lesson twice: once through the bad output and again through the false certainty.

Enthusiasts agreed with the trust problem, but they tended to respond by building frameworks rather than backing away. One participant said that what gave her trust was her own knowledge first and learning how to use the tool responsibly. Another explained that she includes error-handling blocks in prompts specifically to reduce hallucinations and bias. Another said she had to learn to brief AI the same way she would brief a human colleague: be detailed, specify what not to do, ask it not to hallucinate, and ask it to be honest if it cannot find the answer. In other words, enthusiasts do not trust AI because it is inherently trustworthy. They trust the workflows they have built around it. That is an important distinction. The trust is procedural, not unconditional.

Theme 3: Both groups reject replacement narratives

One of the strongest points of alignment was the rejection of the idea that AI should replace human expertise. Skeptics were the most direct about this. One participant said she worries about what happens to her brain if she becomes too reliant on AI, especially for basic cognitive tasks like summarizing and assimilating information. Another described frustration with people using AI to produce client commentary that merely restates the dashboard without adding any insight. Others pushed back on the industry messaging that AI removes the need for an SEO, strategist, or specialist. The objection was not anti-technology. It was anti-replacement. These participants were defending judgment, craft, and professional accountability.

Enthusiasts reached a very similar conclusion through a more optimistic frame. They talked about AI as a “human-enabling” capability, particularly when it helps people work around barriers, move faster through tedious tasks, or prototype ideas that would otherwise get deprioritized. But even the most enthusiastic voices emphasized that AI is useful when it extends human potential, not when it substitutes for it. One participant described AI as powerful because it removed layers of approvals and let her prototype ideas herself. Another said it leveled the playing field for career changers because everyone was, in some sense, learning together. That is a very different vision from replacement. It is a vision of human extension. Across both groups, the healthiest framing was the same: AI is an assistive partner, not a decision-maker.

Theme 4: Hallucinations and overconfidence are still the biggest trust breakers

Although trust showed up everywhere, hallucinations deserve their own theme because of how often and how forcefully they surfaced. Skeptics described seeing AI generate commentary where half the data was incorrect because the model made it up. They described tools producing plausible but false keyword landscapes, content audits, and recommendations that could mislead less experienced practitioners. One participant summed up the danger by saying that the real issue is not AI itself, but “inappropriate use of AI to solve a problem,” especially when people treat the output as gospel. Another worried that the longer this goes on, the more “duff information” will flood the ecosystem.

Enthusiasts did not dispute any of that. Instead, they normalized the need for active mitigation. They talked about asking for honesty when the model cannot find an answer, testing the same task across tools because results vary dramatically, and grounding AI with better sources and better instructions. One participant said her first experiences were disappointing because she expected the AI to produce what she wanted from very short prompts, and she only began to trust it after learning how much detail she needed to provide. Another said that trust came through “play,” experimentation, and repeated failure. The common lesson was clear: hallucinations are still a live issue, and mature users are compensating for that reality rather than pretending it has been solved.

Theme 5: Measurement and attribution remain unresolved

If trust is the emotional barrier, measurement is the operational barrier. Both groups raised the same practical question: how do we know any of this is working? Skeptics repeatedly said they do not have a reliable way to report on the impact of AI-driven visibility, content, or optimization efforts. One participant said her biggest pause with AI is that she cannot clearly report on success. Another described GEO and AI-era measurement as the “wild west.” There was a strong sense that practitioners are being asked to optimize for new environments without a mature, agreed-upon measurement system.

Enthusiasts echoed the same concern. They talked about the chaos around naming, inconsistent data across tools, and uncertainty about whether any platform can truly say an AI tactic is working. One participant described GEO and AEO as inseparable from the need to track share of voice, citations, visibility, and downstream impact, but also admitted that the data layer is still forming. Another said that attribution is the central difference between traditional SEO and the AI search era: the underlying fundamentals of good information architecture and content quality still matter, but the way visibility gets credited and measured is not yet settled. This is a major blocker to adoption because teams can tolerate a learning curve more easily than they can tolerate a measurement vacuum.

Theme 6: Creativity and brand differentiation are a real point of tension

Skeptics were especially vocal about the flattening effect of AI on voice, messaging, and differentiation. One participant said, “AI makes us all sound the same.” Another said everything starts to sound generic, even when she gives the model a clear style guide and continues correcting it. In paid media, the concern was even more pointed: if the job is to stand out from competitors, but AI consistently pushes toward similar patterns, words, and structures, then the tool risks eroding one of the most valuable parts of brand identity: distinctiveness. There was also concern that more junior practitioners who rely heavily on AI may not develop the instincts needed to write strong ads, content, or meaningful commentary in the first place.

Enthusiasts saw the same issue, but they often interpreted it as a setup problem rather than proof that AI cannot support creative work. They talked about the value of brand kits, knowledge bases, and strong inputs. One participant said she was impressed by how closely a workflow could follow an established brand kit when it was properly configured. Another said the biggest unlock in custom tooling was the realization that the model did not need to already “know” everything; it needed access to the right knowledge base. For this group, generic output was often evidence of weak context, weak prompting, or unrealistic expectations. That is not a trivial difference. It shows that both sides are staring at the same creative risk, but one side sees an inherent limitation while the other sees an opportunity for better orchestration.

Theme 7: Education is lagging far behind adoption

Another theme with remarkable agreement was the need for better education. Enthusiasts described many AI learning resources as too theoretical, too long, and too far removed from practice. One participant said she had tried courses that were “not actionable” and full of talk with very little practical value. Another said she missed short, easy explanations and summaries. A third said the most valuable learning came from practitioners who were actively using the tools, even if the teaching was a little chaotic, because real use cases mattered more than polished theory.

Skeptics described a parallel problem from the user experience side. They said tool environments often feel like they are built by coders for coders, with too much jargon, too much engineering-first framing, and too many unexplained terms. One participant said terms like “workflow,” “MCP,” and “connectors” left her feeling lost before she even began. Another described the broader AI space as deliberately phrased in an impenetrable way that makes experienced practitioners doubt themselves. The message from both groups was clear: people do not just need more access to AI. They need clearer explanations, better examples, and a more human on-ramp.

What a healthy relationship with AI looks like

The healthiest relationship with AI is not one-way enthusiasm and it is not total resistance. It is a disciplined partnership. It means using AI where it is strongest (research acceleration, summarization, first drafts, repetitive analysis, ideation, and workflow support), while preserving human ownership of strategy, judgment, and final output.

It means treating validation as part of the process, not as an optional extra. It means resisting the temptation to outsource thinking just because a tool can produce something quickly. It also means being honest about where AI is already genuinely useful, especially for accessibility, efficiency, and reducing low-value friction in the workday.

Just as importantly, a healthy relationship with AI requires boundaries. People need to know what tasks should stay human-led, what outputs must be reviewed, and what level of confidence or sourcing is required before AI-generated work can move forward. The practitioners in these conversations were not asking for perfection. They were asking for transparency, accountability, and tools that make good behavior easier. That is a much more grounded and productive ambition than either fear or hype.

Closing Thoughts

The most useful insight from these conversations is that “skeptic” and “enthusiast” are not fixed identities. They are lenses people use to make sense of a fast-moving technology that is both exciting and imperfect.

Skeptics are not anti-innovation. They are asking for more honesty, more restraint, and more respect for human expertise.

Enthusiasts are not uncritically optimistic. They are asking for better tooling, better education, and better ways to turn possibility into practical value.

Both groups, in their own way, are asking for the same thing: AI that helps people do better work without asking them to surrender the judgment, creativity, and critical thinking that make the work matter in the first place.