AI Is Explaining Your Agency. Is It Getting the Full Picture?

Imagine your agency releases a new preparedness toolkit, publishes a major research report or announces a change in how it carries out its mission. Your team checks the facts, clears the language and coordinates publication. The announcement goes out. The website is updated. Partners receive the materials.

Now imagine someone asking an AI tool what resources are available, what the report actually found or what your agency is doing about the issue. The answer they receive may not be the explanation your team approved. Google, for example, describes how its AI search features can conduct multiple related searches across topics and sources to develop a response. Your agency’s content could be part of that response without being the entire explanation.

That raises a question that belongs in more federal communications discussions:

Is AI helping people understand your agency’s work or leaving them with an incomplete picture?

This is not only a question for employees managing applications or public-facing programs. It also belongs to the people developing resources, publishing research, responding to media inquiries, conducting outreach and explaining the agency’s mission.

Your Website May Be the Source Without Being the Destination

Pew Research Center’s analysis of browsing data from 900 U.S. adults found that users clicked traditional search results on 8% of visits when a Google AI summary appeared, compared with 15% when one did not. Websites ending in .gov accounted for 6% of sources linked in AI summaries, compared with 2% in traditional results. These findings reflect March 2025 browsing activity, not a study specifically of federal-agency searches.

The study does not establish how people encounter information about any particular agency. But it gives us a reason to question an assumption: that people will visit the official website before forming an understanding of the agency’s work.

I would therefore expand the question we ask about public communications. Not only, “Did we publish and promote the information?” But also, “What might someone understand if they encounter that information through an AI-generated answer?”

For a toolkit, that means whether people can discover and use it. For research, it means whether findings retain their context. For public affairs, it means whether people understand what the agency has done, what it has not done and what happens next.

This Is Bigger Than Getting an Agency’s Name Into an Answer

Consider three hypothetical situations.

• A local emergency manager asks what federal resources could help with preparedness planning. The response offers generic advice but never identifies a relevant, publicly available agency toolkit.
• A researcher asks about a federal study. The answer repeats a finding but leaves out the study period or a limitation that materially affects its interpretation.
• A journalist asks about a recent agency announcement. The response describes a planned action as though implementation is complete.

These are different problems. The first concerns discovery. The second concerns context. The third concerns factual accuracy.

Google itself warns that its AI responses may contain mistakes. An official citation should therefore be a starting point for verification, not a reason to assume that everything around it is correct.

For agencies reviewing these answers, the useful question is not simply whether their name or website appears.

What would someone believe — or do — after reading the answer?

Would they find the resource they need? Understand the limits of the evidence? Know which office is responsible? Distinguish an announcement from a completed result? That is a more meaningful standard than visibility alone.

An Announcement Is Not a Lasting Explanation

A press release can explain what happened on a particular date. It should not have to carry the entire burden of explaining an initiative indefinitely.

When planning a significant announcement, I would identify the public page that will provide the ongoing explanation: what the initiative does, whom it serves, its current status and where people can find supporting resources. Then connect the announcement, related resources and subsequent updates to that explanation.

The same principle applies to research and educational materials. Keep the full report, toolkit or training materials, but give people a useful introduction. Explain the intended audience, the problem the resource addresses, its publication or revision date and any important limitations. Do not make a document title do the work of an explanation.

Nor should historical announcements be quietly rewritten to appear current. Preserve their context and make the relationship to updated information clear through the agency’s established publishing and records processes. This is consistent with Digital.gov’s emphasis on maintaining content throughout its lifecycle, reviewing legacy material and establishing ownership, not treating publication as the end of the work.

The objective is not to publish more. It is to make the information you already publish work together.

Make the Information Useful Before Chasing an AI Fix

Before investing in a specialized “AI visibility” solution, I would check some fundamentals.

Can someone understand the resource without knowing the agency’s internal terminology? Are important findings explained alongside their qualifications? Do related pages agree? Are current instructions clearly distinguishable from historical information?

For priority resources, provide readable web content and descriptive links to the complete materials. Important information should not exist only inside a graphic. Google’s guidance specifically recommends making important content available in text and helping people discover it through internal links.

This is not a recommendation to rewrite everything for machines.

Google’s guidance says established search practices remain relevant to its generative AI features. It does not require special AI text files or a particular writing style, and following its guidance does not guarantee inclusion. Those statements apply to Google’s systems, not every AI platform.

My recommendation is to start with changes that are useful regardless of which technology someone uses: clearer explanations, stronger connections between related materials, accurate dates and identifiable sources.

Those improvements do not give an agency control over external answers. They do give people better information to discover and verify.

Public Affairs Should Focus on Accuracy, Not Favorability

There is an important distinction here.

An AI answer is not wrong simply because it is critical of an agency. Nor is it complete simply because it repeats the agency’s preferred message.

The objective should be an accurate, evidence-based understanding of the agency’s responsibilities, actions, resources and results, including uncertainties, limitations and legitimate criticism.

That means making the evidence behind a claim available, not just publishing the claim. It means distinguishing an intended outcome from a demonstrated result. And it means providing a clear, dated explanation when circumstances change.

I would bring this work into existing communications and content-review processes rather than assign it entirely to the web team. Public affairs staff can identify recurring questions. Subject matter experts can verify the facts. Digital teams can address content and access issues. Resource owners can confirm whether materials are current.

Digital.gov’s content-governance guidance similarly emphasizes ownership, subject matter expert verification and clear approval responsibilities. The aim is not to make AI repeat agency talking points. It is to make authoritative information available for comparison and scrutiny.

Start With One Topic and Five Real Questions

A manageable first step is to select one high-priority topic and examine how AI-enabled search represents it. Choose questions that reflect different public needs: What is the agency responsible for? What resources are available? What changed in a recent announcement? What does the supporting research show? Where can someone verify the information or get help?

Do not give the tool the agency’s preferred answer in the question. Test the kind of question someone unfamiliar with the organization might actually ask. Use only agency-permitted tools and public information. Record the question, tool, date, response and cited sources. Have an appropriate subject matter expert review consequential errors, missing qualifications and incorrect or outdated resource links.

Treat the results as a spot check, not an agency-wide accuracy score. Google notes that its AI search features can use different models and techniques, producing different responses and links. Then distinguish between problems your team can address and those it cannot directly fix.

Conflicting agency pages need content-owner review. An incorrect external answer may warrant platform feedback. Missing resources may justify reviewing how those resources are described and linked.

Document the action and check again. Do not assume that changing a webpage proves an external answer has been corrected.

Add Public Understanding to the Communications Plan

Website visits, downloads and media coverage can remain part of the assessment. I would add another question: whether people can find useful resources and develop an accurate understanding of the agency’s work through the information channels they encounter.

That does not require monitoring every answer on every platform. Start where misunderstanding would have meaningful consequences.

The next time your team reviews a communications plan, place one question alongside “Where will this be published?”

“What should someone understand if they encounter only an AI-generated explanation of it?”

The goal is not to make AI speak for your agency. It is to help people find what the agency provides, understand what it has (and has not) done, and verify the information on which they act.

Raitchele Arnell is the CMO of ArtForm Business Solutions, a women-owned digital agency supporting government contractors, enterprise organizations, and public-sector initiatives. She specializes in AI adoption, market intelligence, strategic communications, and modernization strategies for highly regulated and mission-driven industries. Her work spans cybersecurity, healthcare modernization, emergency communications, critical infrastructure, and federal technology initiatives. Known for helping organizations rethink how marketing supports mission outcomes, Raitchele focuses on the intersection of AI, operational efficiency, audience intelligence, and trust-driven communications.

Photo Credit: Matheus Bertelli, Pexels