Pro Tips for Assessment Generation

These tips make a noticeable difference in the quality of what TeachersFlow generates. They're based on how the AI actually uses your input — so understanding the logic behind each one helps.

Upload quality training documents

Upload assessments you were genuinely happy with — not everything you've ever written. More importantly, include variety: different student ability levels, different topics, different assessment styles. The AI learns your voice better from a diverse sample than from ten similar documents.

Be specific in your context field

The context field is where most of the personalization happens at generation time. Don't just write "grammar quiz" — write something like "oral grammar quiz for B1 students who struggle with past tense, 10 minutes, no multiple choice." The more specific, the less generic the output.

Rate your generated assessments

After generating, rate the result using the 1–5 star system. This feeds back into your training profile. A few weeks of consistent rating makes a noticeable difference — the AI starts to understand not just your style but your quality bar.

Select a student when generating

When you choose a specific student, you can also check Include student's last assessment for inspiration. This tells the AI to consider what that student has already been assessed on and adjust accordingly. Useful for building on previous feedback rather than repeating it.

Export and re-train

Go to Groups, filter by assessment type or group, and export the assessments you were happy with as DOCX files. Then upload those DOCX files back into Training Docs. This creates a feedback loop — your best outputs become training data for the next round.

Include a previous assessment for inspiration

When generating, you can select a specific past assessment already saved to a student's profile and ask the AI to use it as a structural reference. Works best when you pick a prior assessment with a similar format to what you're generating now.

Include teaching goals

Check Include teaching goal when generating to anchor the assessment to your active curriculum objectives for that group. The difference is especially noticeable for feedback-style assessments where the goal provides the "why" behind the suggestions.

Use photos for context

You can attach a photo during generation — a photo of a student's written work, a whiteboard, a textbook page, anything that gives the AI something concrete to work with. It significantly reduces generic output.

Use speech-to-text for context input

If you find yourself typing the same kinds of context descriptions repeatedly, try speaking them instead. Speech-to-text is faster and often more natural — you tend to give more detail when talking than when typing.