Describe the assessment. Get the test back ready.

Subject, topic, number of questions and level of the class. The AI writes the stems, spreads the difficulty and builds the Google Form with the answer key already configured — ready for you to review and run.

Instructed for assessment, not for chat

The model underneath is general purpose — what changes is the instruction. FormsFlow reaches the model with the shape of the test, the spread across cognitive levels, the rules for plausible distractors and the answer-key format Google Forms accepts. That layer, not a bespoke training run, is what separates the result from a loose request to a chatbot.

Why asking a chatbot for questions is not enough

Asking for ten questions on a topic is easy and any assistant will do it. The hard part is arriving at an assessment that measures the right level, covers the content in the right proportion, and comes out ready to run without manual work. The generic route breaks in the three steps below.

Shallow or ambiguous questions

Without an instruction about form, the model drifts to definition questions — the ones a student gets right by recognising a word rather than understanding a concept. And to distractors so obviously wrong that they hand the answer to anyone who skimmed.

Difficulty with no control

A useful test mixes recall, application and analysis in a deliberate proportion. Without that instruction everything lands at one level: either a memorisation list, or ten hard questions that fail the whole class while diagnosing nothing.

The manual work comes back at the end

Questions as text still have to become a form: create each question, type each option, switch on quiz mode, mark the key and assign points. It is the longest and most error-prone step, and it is exactly the one the chatbot does not do.

How the assessment gets built

Describe the goal

Say the subject, the topic, how many questions you want and for which grade or level. By text, by voice, or by attaching the material the class studied. The more specific the brief, the less reviewing afterwards — stating the grade level changes the result more than anything else.

The AI spreads the difficulty

Questions are distributed across cognitive levels and formats instead of landing all the same. Each objective question comes with its correct option identified — which is what lets Google Forms grade on its own afterwards.

Adjust by talking

“Replace question 4”, “make 7 harder”, “add two written questions”. Adjustments happen inside the same conversation, with no charge per message; very long conversations use extra credits.

Export to Google Forms

The assessment lands as a form in your own Drive, with quiz mode on and the answer key filled in. From there it is an ordinary Google Form: you edit it, share it and collect responses in the usual spreadsheet.

For every kind of assessment

The format changes with what you need to measure. Describing the intent — diagnose, drill or grade — changes the result more than picking a question type.

Tests and exams

Summative assessment with multiple choice, written answers and true/false in a proportion you set. Objective questions are graded by Google Forms; written ones are left for you to read.

Mock exams

Long lists in entrance-exam style, with fresh questions on the content rather than items pulled from an existing bank. Useful once the class has already seen the mocks in circulation.

Quick diagnostics

Five to ten questions to check what the class absorbed before moving on. Because grading is automatic, the feedback arrives the same day — the only way a diagnostic can change the next lesson.

Retakes and make-up tests

Another version of the same content, with rewritten stems and reordered options. Purely mechanical work by hand, seconds when automated.

Assessment from the class material

Attach the handout, chapter or summary the class received and ask for a test on that. The assessment then covers what was actually distributed, not what the model knows about the topic in general.

Frequently asked questions

What people ask before their first assessment.

Does the AI spread questions across cognitive levels?

Yes, when you ask. The default mixes levels; if you specify “half application, two analysis”, or name a taxonomy such as Bloom's directly, the spread follows the brief. The model underneath is general purpose: what drives the result is the instruction, not a bespoke training run.

Can I choose the question types?

Multiple choice, written answer, true/false and matching, in any combination. Worth keeping the practical consequence in mind: only objective questions are graded automatically by Google Forms. A written question can go in the test, but its mark is still yours to give.

Are the questions original?

They are generated from the topic and the material you provide, not retrieved from a question bank. That means they are fresh, and also that they need review: a language model writes a wrong stem as fluently as a right one.

What do I have to review before running it?

The answer key, always — the shortest read and the only non-negotiable one. After that, checkable facts: dates, formulas, proper names and figures. The gain is reviewing instead of drafting, not skipping the review.

Does it work from a PDF or a recorded lesson?

Yes. Creating from PDF, DOCX and TXT is on every plan, including Free; creating from a YouTube link is on the Advanced plan.

Does the form stay in my account?

It does. The Google Form is created in your own Drive, under your name, and responses land in the linked spreadsheet like any other form of yours. If you stop using FormsFlow, the assessments stay exactly where they are.

Does it work for higher education and corporate training?

It does. Generation works over any textual content — from primary school to postgraduate study and internal company training. What changes the result is stating the expected level of the respondent, not the field.

Do I have to pay to try it?

No. The free plan gives 10 credits a month with no credit card: 1 credit starts the conversation and 1 credit creates the form. Adjusting inside the same conversation costs nothing per message; very long conversations use extra credits.