Agencies must address their AI discoverability gaps

Google’s AI-generated Overviews now appear on 43% of U.S. searches, up from 15% a year earlier, according to research by market intelligence firm Similarweb. When those summaries appear, the Pew Research Center found, users are about half as likely to click through to a website.

In the private sector, this dynamic is a marketing problem. But for federal agencies, it presents a public trust and mission delivery challenge.

Here’s why: About a third of U.S. adults have turned to AI tools for health information in the past year, according to a March 2026 KFF poll. They might get a response rooted in accurate, current, and authoritative information from a federal health office. But they may also get one from an outdated page, a wrong answer on Reddit, or deliberately misleading content.

In short, being the source an AI tool chooses to cite is becoming vitally important. Agencies must adapt to this new reality so that unreliable sources don’t fill the gap.

The discoverability challenge for agencies

Practitioners now distinguish three related disciplines when it comes to search visibility. Search engine optimization (SEO) focuses on ranking a website on a search engine results page. Generative engine optimization (GEO) focuses on getting large language models (LLMs) like ChatGPT, Gemini, and Claude to cite an agency in their answers. Answer engine optimization (AEO) focuses on becoming the definitive source a search engine’s AI selects to answer a question.

A benefits eligibility page, public health advisory, or safety recall notice is the kind of stable, authoritative content that search engine AI generally favors and selects. But this content needs to be written and structured so the AI can find and trust it. If an agency's own content relies on stale PDFs, complicated navigation, or writing that AI systems can’t easily parse into a direct answer, the AI will likely look elsewhere and cite something else—like outdated official guidance or advocacy content presented as neutral information.

5 moves to improve AI discoverability

Solving the discoverability challenge requires an integrated approach, not one communications initiative. A public-facing web page written for AEO, a social presence built around named experts, and a media relations strategy earning third-party coverage all help an AI tool decide what to surface and cite. On social, that increasingly means being the answer inside the platform itself, not only feeding the open web. In our work with federal agencies, we’ve seen that those who manage website, social media, and earned media strategies together tend to make faster progress.

These five moves allow agency leaders to begin identifying and addressing their AI discoverability gaps:

1. Audit your most important public-facing content for AI-readiness. Ask an AI tool questions your agency should be the authority on and see if it’s cited. If it is, your discoverability foundation is likely working. If it’s not, you can start by rebuilding those pages in plain language, organizing answers around the questions the public actually asks, and making sure each answer features clear definitions, current figures, or attributed expert quotes. A peer-reviewed study led by Princeton researchers in 2024 found that adding quotations and statistics to web content made it more visible in AI-generated answers. A 2026 working paper examining more than 21,000 AI citations found that pages with statistics and clear definitions played a bigger role in shaping AI answers.

2. Expand the sources you’re listening to. Traditional and social media monitoring may no longer be enough for agencies to stay attuned to public conversations. What people say on Reddit and in creator communities can shape what AI tools say about a topic, and those spaces can become places where misinformation about your programs spreads before your team ever sees it. By mapping the communities and voices that own each issue in your agency’s portfolio and building a listening capability for them, you can proactively spot emerging questions and false claims and respond with accurate information before they take hold.

3. Make authentic, credentialed voices central to your strategy. Put real subject-matter experts and spokespeople in front of the camera within your agency's existing review process. Short, plainly spoken videos from named experts can feel more human than highly produced institutional content, and that matters for trust. Just 17% of Americans say they trust the federal government to do what is right “just about always” or “most of the time,” according to Pew Research Center, while scientists remain among the most trusted groups in the country.

4. Update your crisis communications for an AI-driven environment. Deepfakes, synthetic content, and false or misleading information create new categories of risk for agencies. These falsehoods also tend to break and travel on social first, in a Reddit thread, a short video, or a private Group, sometimes in spaces that standard listening tools can't see until the claim is already spreading. When it surfaces, the correction often has to happen on-platform and fast, which makes real-time social monitoring and pre-cleared responses as important as the holding statement itself. Credibility can be especially fragile in fields where agencies operate, like public health, benefits, safety, and emergency response. Crisis playbooks should include real-time monitoring, predictive analytics, a deepfake-response protocol, pressure-testing, and pre-cleared holding statements, reviewed continuously rather than filed away after the last exercise.

5. Measuring progress also has to change. Being cited is the goal, but the old scorecard of followers and impressions won't show whether you're getting there. On social especially, the signals that matter now are non-follower reach, saves and shares, and whether your agency is the answer the platform surfaces when someone asks.

Agencies can’t afford AEO inaction

AI tools favor current, clearly sourced content. If agencies allow pages to go stale, AI may opt to cite whatever content it deems fresher, regardless of whether it’s right. It will be harder to compete for public trust once the public has formed a habit of getting that information from somewhere else. And it will be harder to respond credibly in an actual crisis if the deepfake protocol, pre-cleared statements, and listening infrastructure don't exist when the first fast-moving falsehood appears.

At the same time, adapting to AI-driven discovery doesn’t have to be overwhelming for agencies. We’ve helped federal clients adopt structured, evidence-based approaches to help fill these gaps, including rebuilding public-facing content, expanding listening capabilities, and running deepfake and misinformation tabletop exercises. The agencies that start taking these steps now are building the muscle before they need it under pressure. Those that don’t will have to build that capacity live, in public, in the moment it matters most.