A new generation of tools is taking aim at one of the most stubborn barriers to a clean record: not eligibility, but the act of assembling, filing, and following up on the paperwork. Nonprofits, law schools, and a handful of legal-tech startups are now deploying AI to screen people for expungement eligibility, draft petitions, and translate dense court rules into plain English. Recent reporting suggests these tools are reaching people that traditional legal aid clinics have never fully served. The result is a quiet shift in who actually gets their records cleared.
What these tools actually do
The most common entry point is a screening chat. A visitor enters the state where the case was filed, the type of offense, the year of conviction, and whether any new cases have piled up since. The tool returns a plain-language verdict on whether the case is potentially eligible, which statute or rule applies, and what the next step should be. The screening logic is usually a rules engine trained on state expungement statutes, not a free-form chatbot.
Once a case looks eligible, some tools go further. They can pre-fill the court petition, generate a proposed order for a judge to sign, and draft the cover letter a person needs to send to the prosecuting agency. A few partnerships with self-help centers now push the completed file directly into the clerk's electronic filing portal, which cuts out the step that used to require a lawyer or a long trip to the courthouse.
Where the tools are running
Michigan, Utah, and California have been the most active testing grounds, in part because their automatic or near-automatic expungement laws generate the most potential applicants and the most data on what works. Reentry clinics in those states use AI-generated drafts as a first pass, then have a supervising attorney or law student confirm the final filing.
Smaller states without big legal aid budgets have used the tools to scale capacity that did not exist. A rural courthouse that once fielded a few dozen expungement questions a month now routes most of them to a chatbot linked to a real human callback service. The shift matters because most people who qualify for relief never apply, and the reason is procedural friction, not unwillingness.
The real limits of an AI expungement draft
An AI-generated petition can be wrong. Statutory eligibility rules change, prosecutorial consent rules vary by county, and the underlying court records sometimes show details that the screening tool never saw. A draft that claims a case is eligible when it is not can lead to a denied petition, a wasted filing fee, or, worse, an order that the court later vacates after the person has told their employer the record was cleared.
The tools also struggle with cases that span multiple jurisdictions. A person arrested in one county and convicted after a transfer in another, or with both state and federal records tied to the same incident, will often get a partial answer from a tool built around one state's rules. A human attorney is still the right move whenever the situation is more complex than a single, clearly eligible case in one jurisdiction.
What to verify before relying on a tool
Anyone using an AI-driven expungement service should confirm a few basics before signing anything. The tool should clearly identify which state's law it applies. It should name the specific statute it believes makes the case eligible, not just a generic verdict. It should explain what records it has and has not reviewed, and it should let a human attorney or legal aid reviewer look at the file before it is filed.
It is also worth checking who runs the tool. A nonprofit legal aid organization backed by a state bar is a different proposition from a private startup that charges a fee to file the petition. Both can be useful, but the protections, the conflict-of-interest rules, and the appeal path differ significantly between the two.
What this changes for applicants with records
The larger effect is a slow redistribution of who reaches the system. Petition-based expungement favors people with steady internet access, time to gather certified records, and a basic familiarity with court procedures. AI tools, even imperfect ones, lower the cost of taking the first step. They do not replace a lawyer for a hard case, but for the long tail of clearly eligible misdemeanor and low-level felony convictions, they have become the most practical entry point.
The next bottleneck is no longer intake. It is verification: confirming that the case the AI drafted against is actually the case the court has on file, and that the resulting order reaches every background check vendor and the FBI's federal records. That part still requires a person, a checklist, and patience.