.svg)
The honest answer depends entirely on disclosure. Here's what the data says about how customers react when AI identifies itself, and when it doesn't.
TL;DR
Ask whether customers can tell when they're talking to AI, and the honest answer is: it depends almost entirely on whether the business tells them. That's a more useful and more actionable answer than a simple yes or no, because it points directly at what a business actually controls.
There's a version of this question that treats AI detection as some ongoing arms race, customers getting sharper, AI getting better at sounding human, an endless cat-and-mouse game. That framing misses the more practical point. Whether a customer knows they're talking to AI is, in the vast majority of cases, simply a function of whether the business chose to say so. This isn't a detection problem. It's a disclosure decision.
The data here is genuinely reassuring for businesses willing to be upfront. According to SalesAPE's 2026 workplace AI survey of over 250 US professionals, 75.7% of people find it positive or neutral when an AI identifies itself as a bot immediately.
“That's a strong majority who simply aren't bothered by AI involvement, provided they know about it from the start. Disclosure isn't a liability to manage carefully. For most people, it's a non-issue.”
The picture changes considerably when disclosure doesn't happen. The same survey found that 24.3% react negatively specifically when they realize, after the fact, that they'd been talking to an AI without being told upfront. That's a meaningfully different number from the reaction to upfront disclosure, and the difference isn't about the AI itself. It's entirely about the gap between what the customer assumed was happening and what was actually happening.
This connects to a broader pattern of public attention on this exact issue. EY's research found that 73% of respondents fear being unable to tell real from AI-generated content at all, a widespread, general concern about deception rather than about AI specifically. Read together with SalesAPE's own findings, the picture is consistent: people aren't primarily worried about AI itself. They're worried about not being told the truth about what they're interacting with.
This is worth being precise about, since it's changed recently, carries real penalties, and the picture varies significantly by jurisdiction.
As of 2 August 2026, Article 50 of the EU AI Act legally requires that AI systems designed to interact directly with people be built so those people are informed they're interacting with an AI system, unless that would already be obvious to a reasonably well-informed, observant, and circumspect person.
This applies specifically within the EU AI Act's jurisdiction, generally based on exposure to people in the EU rather than where a company is based. Enforcement carries real weight: according to the European Commission, penalties can reach up to €15 million or 3% of a company's global annual turnover, whichever is higher, with proportionality taken into account for small and medium-sized businesses.
One important caveat worth flagging directly: the official text of Article 50 currently carries a notice that it has been amended by something called the Digital Omnibus on AI, and the published text hasn't yet been updated to reflect those changes. That means the specific provisions described here could shift, and anyone relying on this for compliance purposes should check the current, amended version directly rather than treating this as the final word.
In the US, the picture is more of a patchwork. California's AI Transparency Act (SB 942), confirmed directly from the bill text, applies to "covered providers," generative AI developers with over one million monthly California users, and carries a $5,000-per-violation, per-day penalty, with each day of continued violation counted separately.
That threshold puts most small and mid-sized businesses using a third-party AI tool outside its direct scope, since it targets large AI platform providers rather than the businesses deploying them. A companion law, AB 2013, similarly targets developers releasing or substantially modifying a generative AI system, again closer to an AI vendor's obligation than a typical deploying business's, though this detail comes from a secondary compliance summary rather than the bill text itself. Other states, including Utah, have moved toward disclosure requirements that read as more directly relevant to smaller businesses using chatbots, though that detail is similarly unverified against primary statute text here.
Rather than assume a single global or national standard, it's worth checking what actually applies to:
What the data above already makes clear, though, is that disclosure isn't just increasingly required and increasingly enforced, it's also simply what most customers respond well to regardless of what the law happens to require where you operate.
The practical takeaway here is refreshingly simple, and it doesn't require picking a side in a debate about whether AI should be used at all. Disclosing clearly and immediately, "you're chatting with an AI assistant" stated plainly at the start of an interaction, converts a meaningful risk into a non-issue for the overwhelming majority of people.
Trying to make an AI interaction indistinguishable from a human one, whether intentionally or through neglect, takes a manageable disclosure decision and turns it into a trust problem that surfaces later, at a worse time, in a worse way.
If you're weighing how to handle AI disclosure in your own customer interactions, it's worth building it in from the start rather than treating it as an afterthought. SalesAPE offers a free demo if you'd like to see how straightforward, upfront disclosure works in practice, no pressure either way.
It depends heavily on timing. According to SalesAPE's 2026 workplace AI survey, 75.7% react positively or neutrally when an AI identifies itself immediately. The negative reaction, found in 24.3% of respondents, is specifically tied to finding out after the fact, not to AI involvement itself.
Disclosing upfront is clearly the better approach based on the data. Immediate disclosure is a non-issue for most people, while discovering AI involvement after the fact produces a meaningfully worse reaction, even when the interaction itself went well.
This reflects a broader public concern rather than something specific to any one business. EY's research found that 73% of respondents fear being unable to tell real from AI-generated content at all, suggesting the underlying worry is about deception generally, not AI as a category.
Stating plainly and immediately that the customer is interacting with an AI assistant, right at the start of the conversation, rather than leaving it ambiguous or only revealing it if asked. This straightforward approach aligns with what the data shows most people are actually comfortable with.
In the EU, yes, in certain circumstances, under Article 50 of the EU AI Act, in force since 2 August 2026, with penalties reaching up to €15 million or 3% of global turnover. That said, the provision is currently under amendment via the EU's Digital Omnibus on AI, and the published text hasn't yet been updated to reflect the changes, so it's worth checking the current version directly. In the US, requirements are more of a patchwork: California's SB 942 carries a $5,000-per-day penalty but only applies to large AI developers with over one million monthly California users, while other states have moved toward more direct chatbot disclosure requirements. It's worth checking what applies specifically to your business and where your customers are located.