Gains:
- Being able to distinguish where artificial intelligence saves time in preschool preparation work (plans, materials, communication) and where decisions such as development evaluation and diagnosis are left to humans, depending on the task risk level.
- Anonymization of data and implementation of KVKK/child privacy discipline without giving the child's name, identity, photograph and private information to artificial intelligence
- Understanding that artificial intelligence outputs may contain hallucinations and being able to check developmental appropriateness and accuracy one last time as a teacher
A preschool teacher's day looks like "playing games" when viewed from the outside; but in reality it is a knowledge-intensive work consisting of planning, observation, material preparation, parent communication and evaluation. Planning a daily routine, observing 20 children one by one and taking notes, preparing boards and worksheets, finding songs and stories, writing a parent newsletter every month—all of these take time. Artificial intelligence (AI), that is, computer programs that can produce text, images and ideas, saves the teacher time in these repetitive preparation tasks, like an assistant. But let's draw a clear line from the start: AI is a draft generator; It is the teacher who knows the child, evaluates his development and makes the decision. Throughout this module, think of the AI like an intern: fast and diligent, but inexperienced; It does not enter the classroom without you checking everything it produces.
Where is AI useful in preschool and where is it not?
The tasks that AI can really accelerate are obvious: generating activity ideas appropriate to the age group, suggesting different games that explain a concept (for example, "big-little"), turning observation notes into an organized text, writing a draft of a parent newsletter, simplifying the lyrics of a song, restructuring a story according to a certain value, drafting a worksheet idea and visual material. What they have in common is this: all produce outlines, none make decisions about the child.
The area where AI should not enter is also clear. AI cannot decide whether a child has a developmental delay, whether inclusion guidance is required, or what lies behind a behavior. These are the jobs of the teacher who observes the child for weeks, and, when necessary, the guidance counselor and the specialist. Developmental assessment, diagnosis and guidance belong to humans. An observation sentence produced by the AI may be well written, but it has not seen the child; Only you can confirm its accuracy.
Attention: Just because the AI writes fluently and confidently does not mean that it writes correctly. AI sometimes produces non-existent information as if it were real; this is called a "hallucination" (fabrication). Don't assume that an age norm, a developmental milestone, or a source is correct just because an AI tells you to; Verify from reliable source.
Segregation by task risk level
Separating the teacher's work through an eye of "risk" clarifies where to use AI comfortably and where to use it cautiously. A classification like this works:
- Low risk (use comfortably): Generating event ideas, song/rhyme draft, billboard title, newsletter draft, material idea. Even if it is wrong, you can correct it and the child will not be harmed.
- Medium risk (use but verify closely): Converting observation notes into a report, text to be sent to parents, activity requiring developmental appropriateness. Here the AI gives the draft, you check each sentence according to the child and reality.
- High risk (human decision, AI support only): Developmental assessment, inclusion referral, concern about a child to be conveyed to the parent. At best, AI corrects the draft language here; The responsibility for the content is entirely yours.
Child data privacy: the most sensitive issue
The data you work with in preschool education is one of the most sensitive data in the world: information about young children. In Türkiye, personal data is protected by KVKK (Personal Data Protection Law); Child data carries special sensitivity. The basic rule is simple: The child's name, surname, T.R. Never write your ID number, photo, health information, family status into an AI tool. Because many AI tools process what you write on their servers and sometimes store it to improve their model; You cannot retrieve this data that is beyond your control.
The solution is to anonymize the data (making it unidentifiable). Refer to the child as "a 5-year-old girl", not "Elif". Instead of "Ali's grip on scissors is poor", say "A child in my group's fine motor skills are still behind his peers." This is enough to allow the AI to do its job while also protecting the child's privacy. The same rule applies to photos: do not upload children's photos to image-generating AI tools.
Tip: Give yourself this test: "If this text was screenshotted and shared in the parent group, would it be clear which child it was meant for?" If so, you haven't anonymized it enough yet.
Developmental appropriateness: not every idea fits all ages
Preschool is a wide range between the ages of 3-6 and there are huge differences between these ages. An appropriate activity for a 3-year-old child may be boring for a 6-year-old; A task suitable for a 6-year-old may seem impossible to a 3-year-old. Developmental appropriateness means that an activity corresponds to the child's age, developmental level and interest. If you give the age information, AI will adjust it accordingly, but you have the final say: even if AI recommends a small-part activity, you do not apply it in the 3-year-old group due to the risk of swallowing. AI does not know the reality of security and development as much as you do.
three mini cases
Case 1 — Time savings, right limit. A kindergarten teacher spent an average of 3 hours each week creating 5 activity plans. He started using AI as a draft generator: he gave the age group and the theme of the week, asked for 8 activity ideas, selected 5 of those and adapted them to his class. Prep time went down from 3 hours to about 1 hour; He devoted the 2 hours he earned to one-on-one observation with the children. AI provided the idea, the choice and adaptation was up to the teacher.
Case 2 — Privacy failure and return. A teacher wrote a child's name, date of birth and family information such as "mother does not work, father is abroad" into the AI to print a progress report. The director of the institution noticed this and warned: this was a leak of the child's and family's private data. The teacher deleted the text and learned the new method: "A 5-year-old child, I have these observations in the social-emotional domain..." he took the anonymous draft and added the names and special situations manually on his computer.
Case 3 — Catching the hallucination. A teacher asked AI, "How many words does a 4-year-old child use?" he asked; The AI gave an exact number and made up a "source" name. When the teacher searched for this source, he saw that there was no such publication. He confirmed the age norm from National Education and reliable development sources. Lesson: The number and source the AI gives is just a claim until verified.
Four copyable templates
1) Role and boundary setting opening (at the beginning of each conversation):
Your role: assistant to the preschool teacher. I am a kindergarten teacher working with 4-5 year olds. Give me sketches and ideas; Do not make an evaluation/diagnosis about the child. I will never give you the child's name, identity or private information. If you are not sure about the information, say "I am not sure, verify"; fitting.
2) Requesting anonymous event ideas:
Give 6 activity ideas for 5-year-olds with an "autumn" theme, lasting 20-25 minutes, supporting fine motor skills. For each idea, write the purpose, simple materials needed, steps (3-4 items) and a safety note. Be warned if you recommend small swallowable pieces.
3) Developmental fitness check:
Could you adapt the following activity to 3-year-old groups: [paste the activity]. Write down what I should simplify and what security risks I should review according to the 3-year-old development level. Write in short sentences.
4) Privacy self-control:
Is there any information in the text below that identifies any child (name, age+circumstance combination, specific family/health information)? If so, mark it and suggest how I can anonymize it. Text: [paste text]
Weak prompt / Strong prompt
Weak prompt:
Give me activity.
Age, duration, theme, purpose are not clear; AI is general and produces ideas that do not fit the class.
Powerful prompt:
I work with 4 year olds. Give me 5 activity ideas on the theme of "Emotions" that last 15-20 minutes, support social-emotional development, and can be done in the classroom with few materials. Write down the goals and steps for each; Note developmental appropriateness for 4-year-olds.
Quest
Role of AI
Decision/responsibility
Risk
Generating event ideas
draft, idea
Teacher selects/adapts
low
Edit observation note
collects the text
teacher confirms
medium
Report language to parents
draft writer
Teacher approves content
medium
Developmental assessment
Language support only
All teacher/expert
high
Common mistakes
- Writing the child's real name/data into the AI. The most common and dangerous mistake; always anonymize.
- Using AI output without verifying it. Age norm, developmental stage, source—all require confirmation.
- Leaving developmental fitness to AI. Only you know the truth of age and safety.
- Thinking that AI is the decision maker. AI produces blueprints; The decision about the child is the teacher's.
- Using verbatim a text that your class has never met. Adapt each outline to your group.
Tip: Start every AI conversation with an opening sentence that identifies the role and boundary (see template 1). This makes the AI operate more accurately and more safely.
In summary
AI is a powerful assistant in the repetitive preparation work of the preschool teacher: it quickly produces drafts of activities, materials, songs, stories and parent communications. But he cannot know, see or evaluate the child. The two most critical rules: never give child data to AI, anonymize it; and developmental appropriateness, you, as the teacher, have the final say on every decision. The AI is an intern, you are in charge of the class.
Application task
Choose a real task (an event plan or a newsletter) that you will prepare this week. First start an AI conversation with the “Opening role and boundary setting” template, then describe your job anonymously and get an outline. Modify the incoming draft in at least three points according to your class. Write down the changes you made and a "what did I change and why" note in 5 items. Also run your text through the "Privacy self-check" template.
checklist
- [ ] Have I started the conversation with an opening that indicates roles and boundaries?
- [ ] Did I leave any child name/identity/private information in the text (anonymous)?
- [ ] Have I verified the age norm/source information provided by AI?
- [ ] Have I personally checked developmental suitability and safety?
- [ ] Did I, as the teacher, make the final decision and adaptation?