The ClassLens Blog

We bias-tested the AI grading ClassLens runs in Google Classroom on 975 PERSUADE 2.0 essays. No significant gap by race, gender or English-learner status.

See how ClassLens uses Gemini 3.8 Flash to prepare rubric-based draft grades in Google Classroom, with teacher review and a 30-day free trial.

Oklahoma is the one state we found with an enacted parental AI opt-out, and its wording may not even reach a teacher-facing tool. The EU’s high-risk rules apply 2 December 2027, not 2026. We shipped a per-student opt-out regardless.

We benchmarked Gemini 3.7 Flash against human-scored essays before it graded anything: QWK 0.639, versus 0.29 to 0.43 for the model it replaced.

Gemini rubric generation landed in Google Classroom August 5, and Gemini opens to students of all ages August 10. Neither one scores submissions.

The independent verifications a K-12 AI grading tool should hold: Google OAuth, CASA Tier 2, SOC 2, plus the six questions to ask any vendor.

Same-day test of Gemini 3.5 Flash-Lite for grading: human-rater agreement (QWK), grade reproducibility, and bias, versus the models ClassLens ran in July 2026.

ClassLens grades audio and video submissions in Google Classroom, not just essays. A teacher's guide to grading spoken and recorded work with AI.

AI can answer anything now, so the question is the instrument. How to write rubric questions that reveal what students understand, not what they memorized.

ClassLens pulls student work straight from Google Classroom and grades it with AI-assisted feedback. No file uploads, no pasting text. You approve every grade.

Gemini in Classroom drafts feedback one student at a time. It does not grade. What it does today, what it costs, and what fills the gap.

Two field trip days left me with 450 ungraded assignments. Every AI tool I tried only graded essays. So I built one that grades the rest of the pile too.

May 2026 study of Gemini 2.5 Flash-Lite: 200+ experiments, 8 datasets and no demographic bias detected. Historical results, not Gemini 3.8 validation.

Most teachers tried ChatGPT for grading once, got something generic, and quit. The model wasn't what failed. Everything wrapped around it was.

A criterion-by-criterion guide to marking IB English A Paper 1 for the 2026 and 2027 exams. The criteria carry forward unchanged; here is how to apply them.

A criterion-by-criterion guide to marking IB English A Paper 2 under the revised criteria, first assessed in 2026 and still the standard for the 2027 exams.

A teacher's guide to using AI to grade responsibly. Two pillars: student privacy and human connection. Plus the assignments you should never hand to a machine.

Knowledge Gap Reports show teachers where their whole class is struggling after every assignment. How AI grading creates instructional intelligence.

AI has the answers now. My washing machine says AI on it. So what do teachers teach? A 22-year veteran on why questions are the new curriculum.

A 22-year classroom veteran shares the strategies and tools that cut his grading time from 10+ hours a week to under 2.

A veteran teacher compares AI grading and manual grading on accuracy, feedback quality, and time. Where AI helps, and where it falls short.

Google Classroom automatic grading features include Forms quizzes, rubrics, grade import, and AI feedback. See what still needs teacher review.