The Detection Paragraph Parent AI Letter Templates Leave Out

Five things, and a letter that covers 4 of them still leaves the family with a question. Name the tool the school uses. Say that its output describes text rather than whether a student cheated, and that a reading is never a finding on its own. Name the role that reviews a flag before a family hears about it. Then the appeal route, with its deadline.

Search for a parent AI letter template and the pages that come back explain permitted AI use. Of the 7 results checked for this query in September 2026, the one carrying an email template never mentions detection, and the guidance page beside it says nothing about flagged work.

A use policy is a statement of values, and a detection paragraph is an admission that the school checks. The checking is already happening. The Center for Democracy and Technology surveyed 460 US middle and high school teachers in late 2023, and 68 percent said they use an AI content detection tool regularly, up 30 percentage points in a year ⁵. A tool that 2 teachers in 3 reach for is not a detail a letter can hold back.

The clause that belongs in the policy document itself, rather than in the letter home, is worked out in a school AI policy template with a detector clause.

Here is the model text, written in the third person so it drops into a letter, a handbook or a policy page without rewriting. In 6 sentences it names the tool and states the limit, then assigns the review and gives the appeal route. Swap the bracketed items for the district's own names, offices and deadlines, and keep the order they are in.

Model text, for a letter home.

[District name] checks written work with [tool name], software that estimates whether, and how much of, a document reads as machine-written. The number it returns describes text rather than a person, and it is not proof that a student used AI. No grade is changed and no integrity report is filed on that number alone.

A flagged assignment goes to [the teacher and the assistant principal], who read the drafts and the version history alongside the student's earlier writing, and who speak with the student first. Families are contacted within [n school days] of that conversation, whichever way it goes. A family that disagrees may ask [the principal's office] to see the record, and may appeal in writing to [the integrity committee] within [m school days].

What the model text leaves out is an accuracy figure. Vendor numbers move. A letter sent in September gets quoted back in April, and a percentage on district letterhead becomes the thing the meeting argues about. The limit is stated in words instead, and the words hold.

Three of the brackets will take up the drafting meeting. Who is named on the review line decides whether one adult can act alone. The number of school days decides whether a family hears about a flag from the school or from the student at dinner. The appeal window is the third, and it is the one that gets set too short. Ten school days is long enough for a family to find the drafts, read the policy and write something calmer than the first draft of the email they wrote on the night the letter arrived.

A parent who gets a detection letter will search the vendor's name, and the vendor's own false positive figures are a click away. The letter states the limit in words. The figures below are for the administrator who will be asked about them, and for the policy page, where a date can sit beside each one.

In a June 2023 post, Turnitin put its sentence-level false positive rate at around 4 percent, roughly 1 highlighted sentence in every 25, and its document-level rate below 1 percent, for documents it reads as 20 percent or more AI writing ¹. Both numbers are Turnitin's own.

Vanderbilt University announced on 16 August 2023 that it was switching Turnitin's AI detector off, and published the arithmetic. At the 1 percent false positive rate Turnitin claimed at launch, around 750 of the 75,000 papers Vanderbilt submitted in 2022 could have been wrongly labeled ². The sum treats the rate as applying to every paper, which is Vanderbilt's simplification, and it is the kind a board member will run anyway.

The same one at district scale is worked through in the false positive arithmetic behind a detector's accuracy rate.

The error is not spread evenly across writers. A 2023 Stanford study published in Patterns ran 7 detectors of that year, Turnitin's not among them, over 91 TOEFL essays written by people. The average false positive rate was 61.22 percent, and the authors attribute it to detectors reading a narrower range of expression as a machine signal ³. Whether a district's own tool behaves the same way on the work of its multilingual students is a separate question, unmeasured here, so a letter that raises the Stanford finding names it as a reason to check rather than a result.

A separate survey looked at the flag from the student's side. Common Sense Media found 1 teen in 10 saying a teacher had flagged their schoolwork as AI-written when it was not, and Black teens reported it at about twice the rate of their White and Latino peers ⁴. It asked about teachers' flags, not about any detector.

Borrowed wording beats invented wording on district letterhead. Human's page on how to read a result states the limit of its own readings in two sentences. This is an estimate from our detector. A flag should be reason to look closer and not verified proof of cheating. The second of those drops into a letter home with nothing changed.

The teacher who set the work, and nobody else until that teacher has spoken with the student. Name a role and a sequence, never a department. The teacher reads the drafts and the student's earlier writing, then asks the student about them. Only after that does anything reach an administrator or a family. Turnitin's own guidance tells instructors to use a highlighted result to start a conversation instead of drawing a conclusion ¹.

A role outlives a staffing change and a name does not. A letter naming the assistant principal for academics still means something in March, when the post has a different occupant.

The second reviewer matters more than most drafts admit. A single teacher acting on a percentage is the shape of the cases that get reversed, and a second reader costs minutes. Where a district cannot staff that, the rule gets shorter: the teacher reviews and the principal signs. A committee that never meets is worse on district letterhead than a plain two-person rule.

Write down what gets recorded, and when. A conversation that clears the work leaves no record at all. A conversation that does not clear it creates a record the family can read.

That record is where the next conversation starts: what a teacher does when a parent disputes an AI flag, one week after this letter went home.

Three things, and a letter that lists them answers the questions that would otherwise become calls. A parent, or the student once they are 18, can ask to see the record. In the United States, FERPA, as the regulation stands in 2025, requires a school that takes federal education funds to comply with a request for access within a reasonable period and no later than 45 days ⁶.

Amendment is the second ask, and where the school refuses, 34 CFR 99.21 gives a right to a hearing ⁷.

The third one is the one almost no letter mentions. Where the hearing changes nothing, the regulation still lets a parent or eligible student place a written statement in the file. That statement says why they disagree, and it travels with that part of the record whenever that part is disclosed ⁷. A family that knows this in advance behaves differently from one that finds it out through a lawyer.

Whether the detector output is itself part of the education record depends on what the school keeps and where. That is worth settling with counsel before the letter goes out.

The federal shape of the question, and what a district owes when a paper sits on a vendor's server, is set out in how FERPA applies to student papers run through an AI detector.

Deadlines are the district's to set. What the paragraph does is put the appeal route in a family's hands on the day the news arrives.