Medicaid Work Requirements and AI: Is Your Agency the Answer Residents Get?

A resident opens a letter about Medicaid changes. It uses unfamiliar words and a deadline. They don't call, and they don't search for “community engagement requirements.” They open an AI assistant and type: “I got a letter about Medicaid changes. Does it mean I am losing my insurance?”

Whatever comes back is, for that resident, the agency's answer. The question for every state Medicaid and human services agency is simple and uncomfortable: is that answer yours, and is it right?

What is changing, and when

Under federal law (H.R. 1), states must implement Medicaid community engagement requirements for many expansion adults no later than January 1, 2027: at least 80 hours a month of work, training, school or community service, or an income test, with eligibility checked every six months instead of once a year. Certain groups are exempt. CMS issued an interim final rule in June 2026 (Healthcare Dive summary), and states are now sending outreach notices (Georgetown CCF). Details and start dates vary by state, so always rely on your agency's own guidance.

The blind spot: the AI layer

Agencies have done what responsible agencies do: mailed notices, updated web pages, run digital campaigns, briefed community partners and put leadership in front of the media. Leadership then asks a fair question and can't answer it: is any of this reaching residents, and what are people being told when they ask for themselves?

Teams can see web traffic, call volume and notice response rates. What they can't see is what ChatGPT, Gemini, Google AI Overviews and AI Mode tell residents. Wrong answers are heard about secondhand, from a call center agent, a caseworker or a community partner. There is no baseline, no ranking and no way to see who is shaping the conversation.

What research and state experience show

• Awareness is low. CMS qualitative research from April 2026, summarized by Georgetown CCF, found that enrollees are largely unaware of the work requirements and may struggle to tell whether they are affected, with confusion highest among caregivers and people with disabilities. It is qualitative research, not a nationally representative survey.
• People don't know the terms. “Community engagement,” “medically frail” and “redetermination” don't tell anyone what to do. People ask in their own words, and AI is where they ask.
• Scams and look-alike mail confuse people. An outreach notice can be mistaken for a termination notice, and residents ask AI, “is this letter real?”
• Guidance is still moving. AI can repeat outdated or other-state rules long after your own page is updated.

Four questions every answer has to settle

Effective outreach, and effective AI answers, need to answer the same four questions clearly:

• Does this affect me?
• What do I need to do?
• By when?
• Who can help me?

If your plain-language page answers these and a third party's page answers them better, AI will cite the third party.

What residents actually ask AI

• “I got a letter about Medicaid changes. Does it mean I am losing my insurance?”
• “My hours change every week. How do I show that I meet the work requirements?”
• “Do I need to renew more often now? How do I find my deadline?”
• “I care for a family member with a disability. Do I have to meet the work requirements?”
• “I got a notice that my coverage may end. Where do I find review or appeal instructions?”
• “Is this text about Medicaid real? How do I verify it without clicking the link?”
• “What counts as 80 hours?”

Each of these has a different consequence if the answer is missing, outdated or wrong, and each is a prompt your agency can measure.

What an AI visibility baseline shows you

Adobe Brand Visibility turns the AI layer into something you can see and manage. In the walkthrough below, which uses a mock agency and sample data, it moves a team from anecdote to evidence in one session:

• Visibility overview: an AI visibility score, mentions, citations, cited pages and the split by AI platform, so you can see whether one assistant carries nearly all your visibility and whether mentions dipped after a notice wave.
• Topics and prompts: the Medicaid topics where the agency is cited, mentioned or absent, and the high-volume topics where it has no presence, so the content plan is prioritized instead of guessed.
• What AI actually says: the full AI answer to each prompt, which sources it cites, and how it compares with current guidance. A wrong answer becomes a trackable item with a volume, a risk and an owner.
• Who is speaking for you: cited sources, source opportunities and comparisons with neighboring-state agencies, a health plan, an advocacy site and an aggregator.

From evidence to action

Finding the gap is only half of it. In the demo the team picks the top-priority prompt (for example, six-month renewals), reviews the recommended content change for the page AI should be citing, edits the wording to match state policy, and has program and legal reviewers approve it as usual. They preview exactly what will change, deploy it, and can undo it with one click. Then they return to the prompt and track citations and accuracy.

Two fixes matter most for AI. First, plain-language, agent-ready versions of key pages. Second, making sure AI agents can actually see the content: pages that load their text with JavaScript can be invisible to AI agents, so the content has to be made readable to them.

Connect it to the outcomes you already report

AI visibility is a leading indicator. It becomes powerful when it sits beside program KPIs your leadership already tracks:

• Raise awareness: AI visibility score and citations on “what is changing” and “does this affect me” prompts, alongside awareness among affected adults and visits to work-requirement pages.
• Prevent avoidable coverage loss: factual accuracy on deadlines, documentation, exemptions and notices, number of wrong answers corrected and time to fix, alongside procedural disenrollment and notices resolved on time.
• Maintain coverage: citations and accuracy on renewal and reporting prompts, including non-English prompts, alongside on-time renewal rates and churn.
• Increase new enrollment: presence on “how to apply” prompts and accuracy of applicant rules, alongside application starts and completions.
• Reduce misinformation and staff burden: count and severity of inaccurate answers, alongside repeat contacts and call handle time.

What to do in the next 30 days

• Write down your ten highest-stakes resident prompts, in residents' words, not program terms.
• Run them across the AI assistants residents use and record who is cited and whether the answer is right.
• Publish or fix plain-language pages that answer the four questions, and make sure AI agents can read them.
• Set a baseline and a cadence: measure monthly, review quarterly, and tie the results to your program outcomes.

Prolifico launches Adobe Brand Visibility for public agencies in about six weeks and supports it with monthly recommendations and quarterly reviews. If you would like to see where your agency stands, ask for an AI visibility scan of your domain and a baseline on the questions residents ask about your priority programs.

Health and Human Services program · Medicaid 2027 readiness · Request a proposal