During a recent University of Florida College of Journalism and Communications PR Advisory Council call, I heard a term that immediately made me think about how we’re working as marketers and communicators, and how our audiences are making decisions.
The term was “cognitive surrender.”
It’s an uncomfortable phrase, especially when you start wondering how often you’ve done it yourself.
Here’s how it goes. You ask AI a question, it gives you a clear, organized answer, then there’s some reasoning, maybe a few sources, and a recommendation that sounds pretty reasonable.
“That sounds good,” you think. And you move on.
But did you check whether the sources actually supported the recommendation? Did you question the assumptions? Could you explain why the answer was right?
Or did you accept it because it sounded right enough to keep things moving?
Researchers Steven Shaw and Gideon Nave explore this in their work on cognitive surrender at Wharton, and looking at people’s uncritical acceptance of AI’s answers in place of our own reasoning. They distinguish it from cognitive offloading, where we simply delegate a task while retaining control over how we use the result.
I’m all on board for cognitive offloading, at least if it’s done correctly, ethically, and strategically. But surrendering cognitive, or critical, thought is a different story.
Having AI summarize a report can be useful. You can examine what it included, check what it missed, and decide what the findings mean for your organization.
The concern comes when you accept that summary, and the conclusions attached to it, without doing that work.
The moment worth examining is when “AI suggested it” quietly becomes “I believe it.”
For marketing and communications professionals, that creates a challenge on both sides of our work. An audience member might ask AI which company to trust, which product to buy, or what to believe about an issue. If they accept the answer without looking further, it can shape their understanding of our brand before we ever speak to them.
Meanwhile, we might ask AI to interpret audience research, recommend messaging, or help develop a campaign. If we accept its conclusions without scrutiny, we’re making decisions about people we may not fully understand.
At that point, we have a campaign built on unchecked assumptions reaching an audience that may also be accepting unchecked answers.
That deserves our attention.
Using AI responsibly requires us to recognize when assistance becomes deference, and what happens to our audiences, our brands, and our counsel when we stop asking questions.
It’s easy to understand the appeal.
Someone wants to compare products, choose a service provider, or understand a complicated issue. They could spend an afternoon sorting through websites, reviews, and conflicting explanations. It’s not an easy task to be sure.
Or they could ask AI to pull it together in seconds.
And I get it…who doesn’t want a little of that time back?
The answer arrives organized around their question, with a clear explanation, and perhaps a recommendation. It can feel like the research has already been done, even when there are still important things to verify.
That confidence matters. In Shaw and Nave’s reasoning experiments, access to AI increased participants’ confidence even when the system supplied incorrect answers. The experiments weren’t about purchasing decisions, but they raise a relevant question for us: what happens when someone feels more certain about our brand than the available evidence warrants?
We’re also seeing signs that people don’t always go beyond the answer. Pew Research Center’s analysis of Google browsing behavior found that, in March 2025, users clicked a traditional search result in 8% of visits with an AI summary, compared with 15% without one. That doesn’t prove cognitive surrender, but it does show why we can’t assume people will click through to examine the information themselves.
As I wrote about in Your Next Customer May Send an AI Agent First, someone may arrive with a shortlist already assembled and questions about your organization already framed.
Now take that a step further. What if they never check the recommendation? What if an outdated detail, an unsupported comparison, or a missing piece of context becomes something they believe about you?
Your first impression may be an answer someone else assembled, that is completely wrong.
That person might contact your team to learn more. They might also eliminate you from consideration without ever visiting your website. During a crisis or evolving issue, they could accept a summary as the explanation and repeat it to someone else.
We can’t build every campaign around a website visit that may never happen. We have to consider what people believe when they encounter us, and where that understanding came from.
That means we need to expand what we’re looking at when we build a marketing or communications plan.
We already ask who we’re trying to reach, what they need to know, and what we want them to do. Now we also need to ask: what answers might they encounter before they hear from us?
Start with the questions people actually bring into a decision. What makes this company a good fit? Does its product solve my problem? What do customers say about their experience?
Then explore those questions through AI tools and examine the responses. Follow the sources, and look for missing context, outdated details, unsupported claims, and contradictions. One response won’t tell you what every customer sees, but it can reveal gaps worth investigating.
Some of those gaps may be ours.
If our website makes a promise that we haven’t supported publicly, or our sales materials contradict an outside listing, we’re giving people, and the tools helping them, an incomplete picture.
Giving Customers AND AI Proof
That’s why the exercise in Your Brand Has a Pitch. Does It Have Proof? belongs in the planning process. Pick the reasons someone should choose you, then check whether the evidence is available to support them.
Our goal is to make our claims easy to understand, and our evidence easy to check.
This also requires coordination across the PESO Model® ecosystem. Owned content can explain what we do and document results. Earned coverage can provide independent validation. Shared conversations can reveal experiences and questions we need to address. Paid media can bring useful information to more people.
As I discussed in You STILL Can’t Buy Your Way Into AI Relevance, those disciplines need to work together to build credibility.
There’s a practical foundation here, too. Google’s guidance for its AI search features emphasizes established SEO practices and helpful, reliable content. It doesn’t promise inclusion, but it reinforces the value of making important information accessible and keeping business details current.
And we still need direct audience engagement. Ask customers what brought them to you, what they already understood, and what they needed to confirm. Evaluate whether people understand your offer and take meaningful action, alongside tracking visits and clicks.
The goal is to give people a credible basis for a decision, wherever they encounter us.
We’re asking our organizations to provide credible evidence and considering whether audiences will question the answers they receive.
So, the logical next question should be: are we applying that same scrutiny to our own work?
It’s easy to ask AI for audience insights, messaging recommendations, or a campaign strategy and get something that looks ready for a client presentation. The structure is solid, the language is polished, and the rationale sounds convincing.
But what supports it? Actual audience research? Information we supplied? Or assumptions that nobody has checked?
In a survey of 319 knowledge workers, Microsoft researchers found that higher confidence in generative AI was associated with less reported critical thinking. That’s self-reported evidence, but it gives us a reason to examine how confidence affects our review process.
Then there’s the idea of sycophancy: the tendency for AI to agree with or flatter us instead of meaningfully challenging our ideas. In April 2025, OpenAI rolled back a GPT-4o update it described as overly flattering and agreeable.
If you’ve used AI tools, you’ve likely encountered this. The tool enthusiastically endorses your idea, and then it builds a convincing argument around your premise. That can feel like a second opinion, even when your premise still needs testing.
Let’s be clear—agreement from the tool is not validation of the strategy.
And changing a few sentences doesn’t resolve that problem. Editing the answer isn’t the same as evaluating the thinking.
Our review has to reach beyond whether the work sounds like us. We need to know whether it reflects the people we’re trying to reach, is based on credible evidence, and makes strategic sense.
And understanding those people takes more than asking AI to describe them.
Someone using an AI assistant to compare healthcare providers may be worried about a diagnosis. A customer researching software may be responsible for a decision that affects their entire team. A parent planning a vacation may need to know whether their child will feel comfortable and included.
They can delegate the research, but they still have to live with the outcome.
That’s why preserving the human element is both a professional responsibility and a strategic necessity.
We need to understand what people are trying to accomplish (yes, that’s the intent element we talk about so much), what concerns them, and what would help them feel informed enough to act.
That understanding comes from listening, talking to customers, asking frontline employees what questions they hear, and paying attention to the experiences people share, especially when they challenge the story we want to tell.
Then bring that understanding into the content. Use real experiences, specific explanations, and language people can understand. Make important limitations clear. Give someone a way to reach a person when a summary doesn’t answer their question.
As I discussed in Your Brand Has a Pitch. Does It Have Proof?, documented customer experiences can give substance to the promises we make.
Our content may reach someone directly or contribute to an AI-generated answer. Either way, it needs to address something that matters to the person receiving it.
The route to the audience may change, but our responsibility to the person doesn’t.
If someone is willing to accept an answer without looking further, that’s a reason to be more careful with their trust.
So, how do we make that responsibility part of the way we work?
It all starts before the prompt, but here’s what you can do.
Write down what you think the problem is, what you know about the audience, and what you still need to learn. Give yourself a perspective to test.
Then use AI to explore alternatives, identify gaps, and challenge your assumptions. Asking it to critique its own answer can help, but you still need independent evidence and input from people who understand the situation. These AI tools are looking for experts, but you need to verify that’s what they find.
Then, before moving forward with AI-generated content or recommendations, ask these five questions:
A human review needs to be more than a human clicking approve.
If those questions expose a gap, resolve it before the work goes out. That might mean finding better evidence, talking to someone, revising the recommendation, or deciding it isn’t ready.
That judgment is part of the work we’re paid to do.
If AI saves us an hour, what do we do with it?
We could produce more content, or we could spend some of that time checking the evidence, talking to a customer, or questioning whether the recommendation solves the right problem.
That choice matters.
A campaign assembled faster still needs to reflect a sound strategy. A convincing recommendation still needs support, and a message that reaches more people still needs to mean something to them.
If we save time creating the work, some of that time should go into questioning it.
That’s especially important when people may encounter our organizations through answers we didn’t write and accept conclusions we never had the opportunity to explain. We need to understand that behavior, provide credible information, and preserve the human connection throughout the experience.
We also need to recognize the same temptation in ourselves. “That sounds good” can be the beginning of a useful conversation with AI. It can also be the moment we stop doing the work that makes our counsel valuable.
Our expertise shows up in the questions we ask, the assumptions we challenge, and the decisions we can explain and defend. AI gives us another way to explore those decisions, but we’re still responsible for what we recommend and how it affects people.
So, the next time AI delivers an answer that looks ready to use, pause before moving on, and ask: Are we making better decisions, or getting faster at accepting answers?
