Fixing the AI industry’s PR problem

Gammatek ISPL

Sep 3

7 min read

Author : Gammatek ISPL, Published: 3rd Sep 2026 If you sell, deploy, or support any product with "AI" in its feature list, the AI industry's trust problem is already your problem, whether or not you had anything to do with causing it. Recent survey data shows a majority of Americans are now more concerned than excited about AI, and a majority say they distrust the industry's most prominent leaders individually, not just the technology in the abstract. That gap between rising usage and falling trust doesn't stay contained to the frontier labs making headlines. It shows up in your support tickets, your sales calls, and the hesitation a customer has right before they ask "wait, where does this data actually go?"

The trust gap, in numbers

The data on this shifted meaningfully over the past year, and it shifted in a direction most industry roadmaps didn't plan for. A December 2025 YouGov survey found that only 5% of Americans trust AI "a lot" when it comes to making recommendations or providing information, with 41% expressing outright distrust and no industry, including finance and healthcare, earning a net-positive trust score. The 2026 Edelman Trust Barometer found trust in AI has fallen to roughly a third of the population, and for the first time since Pew began tracking the question, a majority of adults under 30, the demographic using AI tools most heavily, now say they're more concerned than excited about it.

It gets more specific than general unease. A CNBC Generation Lab poll asked respondents whether they trusted nine of the AI industry's most prominent leaders to act responsibly, and found a majority distrusted every single one of them, with even the best-performing executive still distrusted by 65% of respondents. Meanwhile, Pew's June 2026 research found 63% of Americans think AI is advancing too fast, and only 16% expect its 20-year societal impact to be net-positive.

Here's the part that should reframe how any AI-adjacent business reads this data: usage is climbing at the same time. This isn't a story about people avoiding AI. It's a story about people using a technology they don't believe in, which is a specific and unstable condition, not a temporary adoption curve that resolves itself.

Why this became a PR problem instead of just a technology problem

The trust gap didn't open because the technology stopped improving. It opened because the way AI got deployed outran the explanation of how it works, who's accountable when it fails, and what happens to a person's data along the way. That's a communication and design failure as much as a technical one, which is exactly why a recent Financial Times analysis framed the AI industry's challenge explicitly as a PR problem rather than purely an engineering one, and proposed a handful of concrete fixes rather than resigning to the "doomer" position that public trust simply can't be recovered.

The uncomfortable part for anyone reselling, integrating, or supporting AI-enabled products: your customers don't distinguish between "the model" and "the product you sold them." If Gemini, ChatGPT, or Copilot behaves in a way that erodes trust, that erosion attaches to every business built on top of it, including yours, even when you had no role in training the model or designing its safety behavior.

There's also an income and generational split worth planning around, not just an average. The 2026 Edelman Trust Barometer found the gap in trust between high- and low-income groups has more than doubled globally since 2012, with the US now showing the largest income-based trust gap of any country surveyed, roughly 29 points. Separately, low- and middle-income respondents were far more likely than higher earners to say generative AI will leave them behind rather than benefit them. If your customer base or workforce skews toward either end of that income spread, a single company-wide AI messaging strategy will land very differently depending on who's reading it, and treating "the public" as one audience with one trust level is itself part of how the industry got here.

Five ways to close the trust gap, reframed for vendors and IT providers

The ideas circulating at the industry level, transparency, accountability, better communication about limitations, aren't just principles for OpenAI or Google's comms teams. They translate into specific, concrete moves for any business that sells or supports AI-enabled products to end users or employees. Here's how each one maps down to that level.

Industry-level fix	What it looks like from a vendor or IT provider's side
Be transparent about what the system actually does	Document, in plain language, which features process data on-device versus in the cloud, and make that documentation something a non-technical customer can actually find and read
Acknowledge limitations instead of overselling	Train sales and support staff to say "here's what this doesn't do yet" as clearly as "here's what this does," instead of letting marketing claims outrun the product
Give people a way to opt out or intervene	Make sure every AI feature you deploy for a client has a documented off-switch and an escalation path to a human, and that both are easy to find
Hold someone accountable when it fails	Name a specific internal owner for AI-related incidents, the same way you'd name an incident owner for a security breach, rather than treating "the AI did it" as an explanation
Rebuild trust through consistent, small proof points	Track and report AI-feature outcomes (accuracy, escalation rate, time saved) the same way you'd report uptime, so trust is earned with data instead of asserted with marketing copy

None of these require you to have built the underlying model. They require you to treat the deployment of AI features the way you'd treat any other product change that touches customer data and customer trust: documented, owned, and measured.

An implementation checklist for rolling out AI features without triggering the trust gap

If your organization is about to roll out AI-enabled devices, software, or support tools, whether that's an AI-heavy phone like the Pixel 11, an AI-assisted help desk, or an internal copilot, the trust research above points to a specific set of things worth doing before launch rather than after a complaint.

Write the data-flow explanation before you write the marketing copy. If you can't explain in two sentences where a feature's data goes, you're not ready to explain it to a customer either.
Separate "AI-assisted" from "AI-autonomous" in every customer-facing description. The trust data shows people are far more comfortable with AI that assists a human decision than AI that acts on its own, so blurring that distinction in your messaging actively works against you.
Build the human escalation path before launch, not after the first complaint. 68% of Americans say they wouldn't let AI act without specific approval, so a visible, fast path back to a person isn't a nice-to-have, it's the feature that makes the rest of the rollout tolerable.
Assign incident ownership in writing. Decide now who's accountable when an AI feature produces a wrong or harmful output, the same way you already have an owner for a security incident or a data breach.
Report outcomes on a cadence, even the unflattering ones. A short, regular internal or client-facing summary of what an AI feature actually did, including its error rate, does more to rebuild trust over six months than any single reassurance statement.

This is the same discipline we've recommended to clients navigating AI feature rollouts on managed devices: the businesses that treat AI transparency as a standing operational practice, not a one-time announcement, are the ones whose support volume doesn't spike every time a headline about an AI failure runs. It's the same logic we walked through when covering the privacy tradeoffs in Google's Pixel 11 AI features: the fix isn't avoiding AI, it's being specific about what it does and giving people real control over it.

Frequently asked questions

Is the AI industry's trust problem actually getting worse, or is this just a vocal minority? The data points to a genuine and widening gap. Pew's 2026 research found a majority of adults under 30, the group with the highest AI usage, are now more concerned than excited about it, and no industry sector surveyed by YouGov earned a net-positive trust score for AI. This isn't a fringe position, it's closer to a majority sentiment sitting alongside record adoption.

If we didn't build the AI model, why does this affect our business? Because customers experience AI through the product it's embedded in, not through the underlying lab. If a customer's trust in AI generally is low, that skepticism attaches to your product the moment it touches an AI feature, regardless of who trained the model.

What's the single highest-leverage fix from this list? The human escalation path. It's the one item on the list that directly addresses the finding that 68% of people won't let AI act without specific approval, and it's usually the fastest to implement since it doesn't require changing the underlying AI feature at all.

Does being transparent about AI limitations hurt sales? The research suggests the opposite risk is larger: overselling AI capability and then having a customer discover the gap themselves does more lasting damage to trust than disclosing limitations upfront. Trust, once lost with a specific customer, is expensive to rebuild.

How does this connect to compliance requirements like the EU AI Act? Directly. Several of the AI Act's human-oversight requirements, a named accountable person, a documented override path, clear disclosure before deployment, are the same practices that close the trust gap described here. Regulatory compliance and trust-building point at the same set of operational fixes.

Does the income and generational trust gap actually change how we should communicate about AI, or is that overthinking it? It's worth planning around rather than ignoring. A single AI rollout announcement written for a higher-income, higher-adoption audience can read as tone-deaf to a lower-income or older audience who are statistically more likely to expect AI to work against their interests rather than for them. Segmenting the message, even briefly, tends to cost little and avoids the most common version of this mistake: technically accurate communication that still manages to erode trust because it assumed a shared starting point that wasn't there.

Not sure whether your AI-enabled products or internal tools would hold up if a customer or auditor asked "where does this data go and who's accountable?"

Our AI governance and compliance review documents exactly that, mapping data flows, ownership, and escalation paths for every AI feature you're currently running or about to deploy, so you can answer that question with a document instead of a shrug.

Book an AI governance and compliance review: https://www.gammateksolutions.com/post/new-google-ai-model-said-to-narrow-gap-on-coding-ability