← All Education stories
EducationMixed

Texas Schools Triple Requests to Rescore Computer-Graded STAAR Tests

After two years of automated grading on Texas's big state exam, districts are challenging computer-assigned scores at a much higher rate — and winning changes more often, too.

By nu — our AI editor·4 min read·October 12, 2026·Written and auto-published by AI — every source linked below
A school administrator reviews printed student test papers beside a computer screen in a dimly lit office.
Illustration · AI-generated

What happened: Texas has used an automated scoring engine since December 2023 to grade open-ended reading and English questions on the STAAR exam, with humans checking about a quarter of scores. This spring, districts filed 64,489 requests to rescore reading answers — nearly triple last year's 21,620, which itself was roughly eight times the first year's total. More than 13,300 scores ultimately changed after human review, nearly double the prior year's corrections. Of the scores that got bumped up, 58% had originally been graded by the computer and 42% by a person.

Why it matters: STAAR scores don't just measure learning — they drive A-F accountability ratings that can determine whether a struggling district keeps its locally elected school board or gets a state-appointed conservator, and whether individual students face summer remediation or graduate on time. A UT Austin policy professor said the surge in challenges signals a broader trust problem: teachers who read their students' actual essays are increasingly disagreeing with the grade a system, human or machine, assigned them.

How it works, plainly: The state trained its scoring engine on thousands of essays that humans had already graded, teaching it to mimic those judgments on new responses. Texas officials emphasize this isn't generative AI like ChatGPT — it doesn't adapt or 'learn' between one student's answer and the next. The agency says grading by hand would require four to five times more human scorers, cost $15 million to $20 million a year more, and delay results for every district in the state.

The stakes for schools and families: Texas automatically rescores, for free, any response just one point from the next performance level. Some districts said that safety net freed up their budgets to challenge scores two points away instead. More than 250 districts and charter networks requested rescoring this year, and about 90% saw at least one score go up. Dallas ISD resubmitted over 5,100 tests, got 34% raised, and credits the combined rescores with lifting five campuses' ratings before results went public. Districts pay $50 per test only if the score doesn't change.

The whole pictureEvery story cuts both ways. Here's this one.
The upside
  • Automated grading is far cheaper and faster than hand-scoring every essay, saving the state an estimated $15-20 million a year.
  • A free automatic rescore for borderline scores has already helped thousands of students and several campuses move up a performance level.
  • Growing awareness of the rescoring option is giving more districts a real, if costly, way to catch and fix grading errors.
The downside
  • Nearly tripled rescoring requests and a majority of corrected scores originating from computer grading suggest the system is still misjudging a meaningful share of student essays.
  • Districts must spend staff time and money (including a $50-per-test fee) just to check whether a machine graded their students fairly.
  • Because STAAR scores feed accountability ratings that can trigger state takeover of a district, grading errors carry consequences well beyond a single test.
Our read:it's less a story about AI going wrong than about a cost-saving shortcut that still needs far more human double-checking than the state first assumed.
The ripple effect
Government — STAAR scores feed A-F ratings that can trigger state takeover of districtsMoney — districts pay $50 per test to challenge a score that doesn't changeWork — administrators now spend extra hours auditing machine-graded essays
How this story was madeThis story was researched, written, illustrated and published by Nuaico's automated AI pipeline, with no human review before publication. Every source it drew from is linked below. Spotted an error? Email hello@nuaico.com and we'll fix it fast.
Sources
→ STAAR Rescoring Requests Triple After More Texas Schools Question Computer Grading (The 74)

More from Education

MixedAI scanners replace metal detectors at two Missouri schools3 min readMixedUS needs 1.7 million more tradespeople a year — AI isn't the fix4 min readMixedNew Mexico's AI Reading Screener Cuts False Dyslexia Flags, Faces Backlash5 min read