In the Age of AI, Writing Has an Authenticity Problem

A growing number of professionals are worried about being mistaken for a machine.

“AI slop” is an imprecise cultural term for content that feels generic, careless, or mass-produced. The discomfort it describes often has less to do with a sentence’s polish than with the sense that no one brought any judgment to the work. That uncertainty matters more in reports, recommendations, and other writing that can shape what happens next.

In interviews we conducted as part of our ongoing UX and marketing research, 73% of knowledge workers said they regularly encounter writing at work that they suspect was generated by AI. Most said that suspicion changes how they view the sender.

For Superhuman, the question isn’t simply whether people use AI to write, but how that use affects the perception of their communication. “We think carefully about how people perceive AI use, both from the perspective of people receiving AI-assisted writing and those doing the writing themselves,” says Heather Breslow, Senior Director of UX and Marketing Research at Superhuman. “When people use our products, we want to ensure that their own voice comes through in the moments that matter most to them.”

AI can produce writing that resembles familiar professional styles. Readers need better evidence that a person’s judgment, knowledge, and care stand behind the words.

AI suspicion shapes first impressions

Writing carries social signals. A message can suggest effort and competence before the reader finishes the first paragraph. The rise of AI-generated writing makes perceived authorship also part of the message. Readers may now assess both what a piece of writing says and whether its sender stands behind it.

That assessment can affect how a message lands. In a study of 1,100 working professionals, participants rated a supervisor’s message with low AI assistance as professional 95% of the time. When they were told the supervisor had relied heavily on AI, that figure fell to 69%–73%. Perceived sincerity dropped from 83% to 40%–52%.

Participants were told how much AI assistance a supervisor used, so the study measured how disclosure shaped perceptions of the message and the sender, rather than whether readers could identify AI use from the prose alone. They may question whether that person wrote the message, cared about the recipient, or had the confidence and capability to stand behind it.

Professional writing has weighted value. A performance review communicates a manager’s judgment. A proposal signals what a team believes. A recommendation puts someone’s reputation behind an idea. Readers will judge the message and the person who sent it.

Confidence needs more than a score

AI detection can surface a reason to look more closely, but the tools don't establish authorship.

Research on AI detectors finds that their performance varies by tool, text type, and the role AI played in the writing process. One study of several detectors recommended using their results as part of a broader review rather than relying on them alone.

So what happens when human-written work gets flagged as potentially AI? The concern is really around misattribution.

When a pattern in someone’s prose begins to function as evidence of inauthenticity, a writer can end up defending how they wrote before anyone engages with what they wrote. A false flag carries real consequences when the work is an executive recommendation or an important internal decision. It can cast doubt on the writer’s expertise, effort, and professional integrity.

That risk is especially acute because professional writing often shares the qualities people associate with AI: clarity, structure, neutral language, clean transitions, and a polished tone.

In our research, 72% of professionals said an AI-detection tool would be valuable. Only 26% reported using one. That gap suggests that recognizing the value of detection is different from knowing how to use its findings.

Detection works best as the starting point for a broader review. Confidence comes from examining the claims, citations, originality, revision history, and reasoning behind the work.

Writers want to own the final draft

When professionals worry that their writing sounds AI-generated, their most common response is to revise it manually.

That behavior says something important about the problem they’re trying to solve. Manual revision gives writers control over phrasing, emphasis, and context. These choices determine whether a message represents them.

The stakes rise when a message carries judgment or care. One participant in our study described trying AI for performance reviews and rejecting the result because it didn't sound like their voice. The writing felt “mechanical” in a process where tone and personal accountability matter.

Another participant explained the standard they used for work presented to leadership: everything gets filtered through their personal review before being shared. That reveals a clear expectation of human responsibility.

Voice comes through in the writer’s point of view, the details they know firsthand, the reasoning behind their conclusion, and the context they choose for a particular audience. Those choices give writing its weight. They help readers understand where an idea came from and what informed it.

Welcome to the authenticity economy

AI has made content abundant. A polished first draft, a tidy summary, and a competent-sounding message now take less time to produce.

This is the beginning of an authenticity economy: a period when abundant, competent-looking content makes the evidence behind a piece of work more valuable. Readers expect supported claims, sound sources, original work, and judgment applied before someone puts their name behind the result.

The broader digital world is already building systems for that problem. The Coalition for Content Provenance and Authenticity has developed Content Credentials, an open technical standard that records the origin and edit history of digital media. It gives people information about how an asset was made and how it changed, rather than asking them to rely solely on visual impressions.

Professional writing raises a parallel need. When the final page cannot fully show how the work was made, confidence depends on the evidence around it: the sources, the reasoning, the revision process, and the person prepared to take responsibility for the final judgment.

That is the standard worth building toward. AI can help people produce better work. People remain responsible for the work they share.