When a parent wonders how to open a child support case, when a teacher needs mandated-reporter training, or when a Medicaid member hears that the rules are about to change, the first stop is increasingly an AI assistant. The answer they get may or may not come from your agency.
Every state has a Health and Human Services agency, usually among the largest in state government by funding and by people served. Each typically runs 10 to 20 programs, including Medicaid, SNAP, TANF, child support, child welfare and health insurance exchanges, with significant overlap in who is eligible for what. Some states administer everything centrally; others supervise programs that counties administer, as in North Carolina and parts of California and Colorado.
That scale and complexity make HHS agencies unusually exposed to how AI search describes them. Here are six challenges we see, and what to do about each.
1. Negative-sentiment bias in AI answers
People who had a smooth experience rarely post about it. People who waited too long, or whose benefits were suddenly canceled, often do, on Reddit and elsewhere. AI assistants draw on what is public, so complaints can carry more weight in answers than the agency's own pages. That is a service problem and also a reputational one: agency leaders are accountable for how the public perceives their tenure.
What to do: measure how the agency and each program are described, find which sources shape those answers, and publish clear, accurate, current official content that AI systems can cite.
2. Revenue leakage on optional programs
Medicaid and SNAP are mandatory programs; child support is optional for many families. If a parent can't easily find or trust the agency's channel, they may hire a private attorney or handle it informally. One large state program handles more than 700,000 cases and over $1 billion a year. Every case that never reaches the agency is lost collections for families and the state.
What to do: make the agency the obvious first answer to questions such as “how do I get child support without a lawyer?”, with trust-building content and easy digital entry points.
3. Mandated-reporter visibility gaps
Teachers and health and mental health workers are the source of most child welfare reports. One state agency had to redo its mandated-reporter training because people couldn't find the resources through a simple search. It was a pure content-visibility failure with real consequences for children.
What to do: monitor the prompts mandated reporters actually use, and make training, definitions and reporting steps easy to find, current and accessible.
4. The Medicaid work-requirements cliff
Under federal law (H.R. 1), states must implement community engagement requirements for many Medicaid expansion adults no later than January 1, 2027: at least 80 hours a month of work, training, school or community service, with eligibility redetermined every six months instead of annually. Exemptions apply to groups such as pregnant women, caregivers of a dependent child and some people with disabilities.
Arizona’s Medicaid agency launched KeepMyAHCCCS.com on September 8, 2026 for the nearly 417,000 members who could be affected, a sign of how much awareness work remains. With only months of runway left, search is where people will land first, and AI is increasingly the front end of search.
What to do: publish plain-language explainers, exemption guidance and reporting instructions in the languages residents use, and measure whether AI answers get them right. See our Medicaid 2027 readiness page.
5. Constant federal policy churn
Beyond Medicaid, rural health funding and other still-undefined federal changes keep agencies in a permanent “we have to get the word out” cycle. There is no reliable way to guarantee that accurate information surfaces in AI search each time something changes.
What to do: treat visibility as an ongoing program, with a prompt library that is refreshed quarterly and monthly recommendations tied to policy and program changes.
6. The downstream cost of unanswered questions
When people can't find accurate answers themselves, call centers fill up and frontline caseworkers field the same basic questions repeatedly. That drives cost and slows service for everyone.
There is a flip side. If an assistant helps an applicant gather documents and pre-fill paperwork before their first caseworker meeting, the same technology becomes an efficiency win. The difference is whether the assistant is pointing to your accurate, official information.
What to do: fix the sources that generate repeat questions and publish checklists and forms that AI can hand to residents directly.
What an HHS visibility program looks like
Adobe Brand Visibility shows how agencies and programs appear in AI-generated answers: where they are described accurately and cited, where information is missing or outdated, and which other sources are winning. Prolifico turns that into an operating program.
- Listen: build a prompt library of up to 1,000 resident questions organized by program, audience, county and journey stage.
- Measure: establish a baseline and benchmark against peer agencies.
- Fix: close gaps with accurate, accessible content (including Section 508 and WCAG), technical improvements that make pages readable to AI agents, and stronger authority for official sources.
- Prove: report monthly and review quarterly, so leaders can see what improved and what policy changes need attention.
The launch takes about six weeks and can start with one program, one agency or a coordinated statewide effort. We recommend starting where the pressure is highest: for many states, that is Medicaid readiness for January 2027.
Learn more on our Health and Human Services page, see program-by-program examples, or request a proposal.