Grassroots Advocacy and the Authenticity Gap:
Trust and Influence in the Age of AI

By Bryan Dawson

“The Liar’s Dividend”

A compelling story can change our minds. A photograph or video can move us to action. A familiar voice can persuade us to trust. Behind those reactions is an assumption: there is a real person, experience, or event behind what we are seeing and hearing. Artificial Intelligence is making that assumption harder to defend.

A convincing account of something that never happened can be generated in seconds. Images can depict events that never occurred. A recording can put words in someone’s mouth. A deepfake video can distort perceptions and influence someone to change their vote. Fabrication has always existed, but the cost, speed, and accessibility of producing it have changed dramatically.

The National Institute of Standards and Technology’s examination of synthetic content lays out the resulting challenges of establishing where content came from and identifying artificially generated material. Creating something believable can require far less effort than establishing whether it deserves to be believed.

This affects more than our ability to spot misinformation and disinformation. It changes how influence works. If we cannot confidently identify who is speaking, whether their experience is real, or whose interests a message serves, we have less reason to trust it.

Even awareness of the problem creates another vulnerability. Researchers at the Brennan Center for Justice and Georgetown’s Center for Security and Emerging Technology describe the “liar’s dividend” like this: The existence of convincing fakes gives dishonest actors an opening to dismiss genuine evidence as fabricated. Falsehoods can gain credibility while truthful accounts become easier to deny.

AI’s Impact on Grassroots Advocacy

For grassroots advocacy, that erosion of trust threatens the relationship on which the entire enterprise depends: a real person communicating a real concern to someone empowered to act.

A constituent’s message matters because someone in the district cares about the issue. What happens when a decision maker can no longer be confident that the message reflects that person’s beliefs, experience, or even their decision to participate? An early experiment offers a useful starting point, but its findings require care.

A response is not the same as influence

In a Cornell study published in 2023, researchers described an experiment conducted in 2020 involving more than 32,000 emails sent to approximately 7,000 state legislators. Human-written messages received responses at a rate of 17.3%, compared with 15.4% for messages generated by GPT-3.

The narrow difference might suggest that AI-generated messages could substitute effectively for human-written appeals. But the study measured responses, not persuasion or policy influence and a response is not the same as influence. An acknowledgment may be automated. A staff member may send a standard reply without finding the message credible or persuasive. Those response rates do not establish that the messages changed legislative priorities, altered a vote, or earned anyone’s trust.

The timing matters, too. This experiment took place in 2020, before the public release of ChatGPT in November 2022 and the subsequent expansion of generative AI into advocacy platforms. Importantly, the experiment did not test how offices would respond after years of exposure to AI-assisted campaigns or widespread concern that apparently personal messages might be manufactured. My concern is what happens as decision makers learn to distrust the correspondence itself. Today, many advocacy technology providers explicitly offer AI-powered message creation for grassroots campaigns.

The Growth of Distrust

My concern is what happens as decision makers learn to distrust the correspondence itself.

If an office repeatedly encounters invented experiences, constituents who know little about messages sent in their names, or thousands of seemingly personal appeals generated without meaningful human involvement, it has reason to become skeptical. As that skepticism grows, these letters risk losing influence because the office no longer trusts the messenger.

The damage can spread beyond the campaigns responsible. A sincere letter from a nurse, business owner, or parent may arrive in the same inbox and face the same suspicion. Once credibility deteriorates across a channel, genuine participants can pay the price.

This concern is already appearing in professional guidance. The Public Affairs Council warns that legislative staff may question or discount AI-generated content. It also describes the credibility problem that arises when staff follow up with a constituent who has no knowledge of a message sent in their name.

For our industry, this creates a difficult conundrum: AI helps us become better at producing messages while making those messages less influential.

For our industry, this creates a difficult conundrum: AI helps us become better at producing messages while making those messages less influential. That does not mean AI has no legitimate place in advocacy. AI can help someone understand a policy issue, translate a concern, organize their thoughts, explain an experience clearly, or develop campaign descriptions.

We have long accepted assistance through sample letters, talking points, and advice from organizations we trust. Shared wording does not make a position insincere. Ten thousand people can legitimately endorse the same sentence. Ten thousand differently worded messages can still be fabricated.

Authenticity Matters

Authenticity depends on the connection between the person and the message. That brings us to an ethical question: When is a constituent justified in trusting AI to speak on their behalf? Agreement with a broad goal does not automatically authorize every policy position a system might attach to it.

Authenticity depends on the connection between the person and the message.

The constituent needs control over that distinction. Platforms should explain the request clearly, display the actual message, allow changes, and obtain approval before sending. Personal details must come from the person, not from a model’s effort to make the appeal more persuasive. This aligns with the Public Affairs Council’s recommendations on consent, review, and approval. This is also where direct conversation deserves renewed attention.

The Case for Direct Calling
AI cannot, at least not yet, replace a live human conversation between a constituent and a lawmaker’s office. A live call gives legislative staff an opportunity to engage directly with a constituent, ask follow-up questions, and explore how an issue affects their life. That exchange offers more opportunities to assess understanding and credibility than a standalone email. Even a personally delivered voicemail gives someone an opportunity to express what they want their representative to understand.

While the case for calling rests on the real participation it enables, email remains an important part of that relationship. A constituent’s email can draw attention to an overlooked issue, explain a concern, and put a specific request before a decision maker. It can also provide an effective follow-up to a call, reinforcing the conversation, supplying supporting information, and creating a written record that staff can reference or share. Its value depends on whether it credibly represents the person sending it.

A call and a thoughtful follow-up email can work together to give an issue both personal immediacy and lasting substance. That is the approach behind the company I founded, The Virtual Citizen. AI-generated letters can be produced at scale with little constituent involvement, but getting real people to initiate calls from their own phones and personally explain their concerns can be hard and requires a different level of participation. The Virtual Citizen makes that participation easier, giving decision makers an opportunity to hear directly from the people affected.

The Switchboard is NOT Enough
Making that participation easier requires more than publishing the congressional switchboard number and asking supporters to call. That leaves the constituent to navigate the connection and prepare what to say. It also leaves the sponsoring organization without visibility into whether someone acted. Providing a phone number alone supplies no call verification, campaign attribution, or contact data for follow-up.

The Virtual Citizen brings those steps together. Advocates dial #250 and speak a campaign keyword, or scan a direct-dial QR code. Before being connected to the appropriate office, they hear a short campaign message explaining the issue and the requested action. That preparation helps callers approach the conversation with greater confidence while leaving them free to express their concerns in their own words.

Organizations can place that invitation virtually anywhere: radio, television, billboards, point-of-sale displays, print publications, direct mail, digital advertising, social media, and live events. Different spoken keywords can be used per channel for foolproof attribution to gauge performance of advertising spend. After the call, a text message can provide a link to send a follow-up letter or email, reinforcing the request in writing.

The Virtual Citizen captures call activity, including call duration, mobile data, and call origins mapped by Designated Market Area (DMA), regardless of the phone’s area code. Knowing where people call from offers insights to help organizations understand which outreach channels generate participation, where engagement is occurring, and whom they can re-engage as an issue develops.

The platform captures call activity, including call duration, mobile data, and call origins mapped by Designated Market Area (DMA), regardless of the phone’s area code. Together with campaign attribution, those insights help organizations understand which outreach channels generate participation, where engagement is occurring, and whom they can re-engage as an issue develops. Call activity does not by itself prove persuasion, but it gives organizations a more useful measure of participation than simply distributing a number and hoping people dial.

For many people, caring is not the obstacle. Knowing whom to contact, getting connected, and feeling prepared to speak can be. Removing those barriers, then supporting informed follow-up, helps turn concern into participation and participation into an ongoing relationship.

Conclusion

Calling should be part of an accessible mix of communication options. Making it easier for people to participate personally can strengthen the connection between constituents and decision makers. Across all channels, the standard should be the same: A real person understands, authorizes, and stands behind the message delivered in their name.

The long-term risk is clear.

When decision makers stop believing that constituent messages represent authentic participation, those messages lose their power to persuade. Generating more of them will not restore the trust that made them valuable.

A citizen’s voice has value because there is a person behind it. Preserving that connection is our responsibility.