How the fifty-state e-filing table was built

The question, stated narrowly on purpose

May a party representing themselves electronically file a divorce, and what does e-filing add to or subtract from the cost?

That is deliberately narrower than “does this state have e-filing.” Nearly every state does. The question that matters to a self-represented filer is whether the system is open to them, and that turns out to be a different question with a different answer in more states than we expected.

Source hierarchy, in the order we accept it

1. The court’s own rule, administrative order, or standing order

   Preferred, always. A rule is the thing that governs, and it is the only source that settles the question rather than describing it. Where a state’s answer is in a rule, the row cites the rule by number.

2. The court system’s own published help page or portal documentation

   Accepted where no rule addresses the point, because in several states the practical answer lives in the portal’s own registration policy rather than in any rule. Cited as what it is — the court’s guidance, not the court’s rule.

3. The published fee schedule

   For amounts only. A fee stated anywhere other than the schedule that sets it is treated as unconfirmed.

4. Nothing else

   No vendor marketing, no law-firm blog, no legal-information aggregator, no wiki, and no other compilation. This last exclusion is deliberate and it is the reason the table took as long as it did: had we been willing to copy another compilation, it would have been a day’s work and it would have inherited that compilation’s errors without knowing which ones.

What “confirmed” means, and what we do instead when it is not

A cell is confirmed when a named source in the hierarchy above states the answer and we recorded the date we read it. Everything else says not confirmed, in those words, in the cell.

No state is currently marked not confirmed on the e-filing question. Three were on 30 September — Louisiana, Missouri and Virginia — and all three closed on 1 October. None of them closed by finding a better copy of a rule we had already read, which is the part worth recording.

Louisiana was not a missing rule but a misread one. La. C.C.P. art. 253(B)(1) is permissive and names nobody, and the mandate added at art. 253(B)(2) from 1 January 2026 is aimed only at filings “filed by an attorney” — so eligibility is a parish question by design. We had been looking for a statewide answer that the statute deliberately does not provide. The Jefferson Parish Clerk of Court then answers it for the 24th Judicial District Court in one sentence, and says “self-represented litigants” in those words.

Missouri was a dating problem and we solved it by reading the rule instead of the FAQ. Court Operating Rule 27 carries its own dates — adopted 1 September 2011, revised effective 1 October 2019 — and makes self-represented registration conditional on the automation system providing for it. Four circuits, including the largest, state in dated local rules that the condition is not met for a civil case.

Virginia we had wrong, and not in the not-confirmed direction. We said the rule appeared to contradict itself and that we would not resolve a contradiction in a court rule by picking the reading we preferred. There was no contradiction. Rule 1:17(c)(5) sits under the heading “System Operational Standards” and its own clause limits “remote electronic access to documents” — it lists who may view the electronic file, not who may file into it. We had quoted a fragment of the sentence and reasoned about the fragment. That is a different failure from the ones this page was built to prevent, and it deserves its own line in the method: quote the whole sentence, including the subsection heading it sits under, before concluding that a rule is incoherent. A drafting body is far more likely to be coherent than our excerpt is to be complete. It is on the corrections page.

Tennessee was the fourth for one day, and should not have been. We marked it not confirmed because we could not read Tenn. Sup. Ct. R. 46 — the judiciary’s site serves an automated-traffic challenge instead of the rule. But R. 46 approves e-filing court by court, so the rule we could not read is not the rule that answers the question. The local rules are, and four of them are readable: Davidson County Circuit Court lists “Pro se and indigent litigants” among authorized e-filers, Davidson Chancery does not, Shelby offers it free to self-represented filers, and Hamilton says paper filing continues. Tennessee now reads depends on your county. The mistake was applying the not-confirmed test to the wrong source, which is worth recording: an unreadable source only makes a cell unconfirmable if it is the source that governs.

Connecticut used to be the entry in this paragraph. It came off the not-confirmed list on 29 September when the standards document finally downloaded, and on 30 September it became the best-sourced row in the table — the Judicial Branch publishes a case-type table that answers the question by docket code rather than leaving it to be inferred. The gap we had named for weeks turned out to be a document we could not fetch, not a rule that did not exist.

Those two were the same mistake twice, and that is the part worth keeping. In each case the reasoning was sound and the thing being reasoned about was a stand-in for the source rather than the source: half a sentence standing in for the sentence, an unreadable statewide rule standing in for the local rules that actually govern. Neither error would have survived one more look at the primary material, and in both cases we had already decided what the answer was before taking that look.

So the rule we work to now is narrower than check your sources, which nobody disagrees with and which did not prevent either of these. Before publishing a conclusion drawn from a source we could not fully read, we have to name what would tell the possibilities apart — the sentence that would settle it, the document that would settle it — and go and get that, rather than picking the explanation that fits what we already have. A confident mechanism built on partial evidence reads exactly like a confident mechanism built on complete evidence. That is what makes this failure worth a rule of its own: nothing about it feels like guessing at the time.

The findings the method produced

All of these are in the table with their sources, and all were re-checked against the live page on 30 September 2026 before being repeated here. The count changed that day: what had been four states is sixteen — and eighteen once the last three unconfirmed cells were read on 1 October — and the eleven-row correction is on the corrections page with the date and the sources.

Sixteen states do not permit it, and most of them never say so

That is three of the four quoted. The pattern worth noticing is that none of these is an accident of software. Each is a deliberate policy written down somewhere a self-represented person would never look.

The cost runs the opposite way to expectation

E-filing is generally assumed to be the cheap option. For a self-represented filer it is sometimes the expensive one.

Colorado: the portal charges $18 a filing plus $18 to e-serve (from 1 October 2026; $12 before), while paper filed over the counter carries no portal fee at all. Arkansas: $20 a filing and $100 to register. A registration fee is a barrier of a different kind from a per-filing fee, because it is paid before anything has been accomplished.

What we do when we are wrong, which we have been

Two of the four no-e-filing states were wrong on this site for weeks before we caught it. That is written up on our corrections page with the date, what changed, and how long the error was live.

The corrections page is not a gesture. It is the reason to trust the rest, and it is the answer to the only question a librarian actually needs answered: not whether this is right, but what happens when it is not. Every correction carries a severity, the date, and how long the error stood. Some of them are corrections to earlier corrections.

Re-checking, and how stale this gets

E-filing rules move. Mandatory-filing orders get amended, portals change registration policy, fee schedules get adjusted by a legislature.

Every row carries the date it was read, which is the honest way to publish something that goes out of date: not a promise of currency, but a statement of when it was true. An automated check reads the whole site daily and reports pages whose figures no longer agree with our own source-of-truth records, and the statutory material is swept on a schedule — quarterly for the most dangerous categories, monthly for court fees. Those checks report to a person. They do not edit the site. We would rather a correction took a day longer than have a machine quietly rewrite a rule.

The same answer, written twice, then compared

The e-filing answer for each state exists in two places here, written at different times: the row in the fifty-state table, and the how you file block on that state’s own divorce-kit page. They were produced from the same source hierarchy but not from each other, which makes comparing them a real test rather than a formality. Agreement means two separate readings of a rule landed in the same place. A mismatch would mean at least one of them is wrong, and we would not know which without going back to the rule.

On 30 September 2026 all fifty pairs were compared: zero mismatches. The next day that result was shown to be almost worthless, and it is worth saying why. The check compared the table’s summary line against the kit page’s summary line. On eleven states both said the same thing and both were wrong — and on six of those eleven, the correct answer was already sitting on the kit page, a few lines below the wrong headline, in the block’s own Account field and in the rule quoted in its Authority note. New Mexico’s page said the reader might e-file at the top and quoted the rule prohibiting it at the bottom. A check built to compare two summaries of the same research can catch a transcription slip. It cannot catch a summary that contradicts its own source, because both copies of the summary agree. The check now compares each badge against the rule quoted underneath it, which is the comparison that catches six of the eleven. The other five carried the right answer nowhere on the page, so no comparison could have found them; those were failures of research rather than of summary, and they need a different remedy. That is the whole result, and it is worth stating plainly because it is a claim we can point at rather than one you have to take on trust — both halves are published, so the comparison can be repeated by anyone who wants to do it.

What it does and does not prove. It proves the site is internally consistent: it does not say twice about a state what it could not say once. It does not prove either reading is right. Two passes over the same misread rule would agree with each other perfectly. The date on each row, and the citation beside it, remain what you should check if the answer matters to a filing.

A second automated check began running nightly on 30 September 2026. It covers the kit pages rather than the statutory material: each kit’s form list and cost table are copied from that page’s own verified sections, so the check re-derives them and reports any page where the two no longer agree. It found twenty-nine such pages on the day it was written, which is why it exists. Like the others, it reports to a person and does not edit anything.

Who made this, and what we get out of it

The Cusp is a free guide to divorce in all fifty states. It is run by a real estate agent, not a lawyer. That is stated on every page because it is the most important thing to know about the source.

The e-filing table is free, requires no account, and has nothing to buy on it. The site earns money from labeled paid placements in a professional directory, and the rate card, the caps and the arithmetic are published. None of it touches this page.

Using it

Link it, quote it with attribution, or take the underlying data. If a table of this shape would be more useful to you as a file than as a page, write and say what format and we will send it — there is no form and no sign-up.