Updated on
August 12, 2026
AI and EdTech Tools for Teachers: An Evidence-Informed Guide
Choose AI and EdTech tools using evidence, accessibility, safeguarding, data protection, workload and total cost.
AI and EdTech tools can support planning, practice, feedback and administration. Schools should evaluate learning value, accuracy, accessibility, safeguarding, data protection, workload and cost before adoption.
The Department for Education says evidence about generative AI's effects on learner development and educational outcomes remains limited. It currently sees more immediate benefits and fewer risks in teacher-facing use than in pupil-facing use. Structural Learning therefore recommends beginning with a defined need, professional review and a small monitored trial rather than a league table of fashionable products.
Jurisdiction note: the DfE and Ofqual sources cited in this guide apply to England. Readers elsewhere in the UK should check guidance from their national government, regulator and awarding bodies.
Key takeaways
- Start with the educational problem: identify what should improve before choosing a platform.
- Test the actual task: generic claims about AI or EdTech do not prove that one product will help your learners.
- Keep professional review: staff remain responsible for checking accuracy, curriculum fit, bias and appropriateness.
- Protect learners and data: check safeguarding, age restrictions, accessibility, privacy and assessment rules before use.
- Measure total cost: include licences, training, setup, support, data work and the time required to check outputs.
What are AI and EdTech tools?
Educational technology, or EdTech, includes digital services used for teaching, learning, assessment, communication and school administration. Generative AI is one part of that larger category. It can create text, images, audio, code or other content in response to instructions. A quiz platform may use no generative AI; a planning assistant may rely on it heavily; a learning platform may combine several approaches.
This distinction matters because the risks and evidence differ by function. A teacher using AI to draft five retrieval questions is not the same use case as a pupil holding an unsupervised conversation with a public chatbot. A school should therefore approve defined uses, users and data rather than adopting or banning “AI” as one undifferentiated thing.
Start with a teaching or operational need
The Education Endowment Foundation's 2019 digital-technology guidance advises schools to consider how technology will improve teaching and learning before introducing it. Write the need in observable terms: “Year 8 pupils need more opportunities to practise balancing equations with timely corrective feedback” is testable; “we need an AI platform” is not.
Describe the current process, the intended users and the outcome that matters. If the problem is slow feedback, establish the present turnaround time and the quality of feedback learners receive. If the problem is workload, record which task consumes time and whether that time is actually reduced after checking, correcting and distributing the output. A tool that makes drafting faster but creates more verification or data-entry work may move rather than remove workload.
The DfE's July 2026 EdTech procurement guidance advises schools to involve their data protection officer early, establish supplier and controller roles, minimise data, examine AI training and international transfers, plan deletion and exit, and monitor supplier changes.
An evidence-informed evaluation checklist
Use the following questions before procurement and again during a trial. They are decision prompts, not a numerical certification score.
| Area | Question | Evidence to collect |
|---|---|---|
| Learning value | What teaching, practice or assessment process should improve? | A baseline and a named learner outcome or process measure. |
| Accuracy | Can staff identify and correct plausible errors? | Representative tasks checked against trusted subject sources. |
| Curriculum fit | Does the output match the relevant curriculum, specification and age group? | Subject-lead review of real examples. |
| Accessibility | Can intended users perceive, operate and understand the service? | Keyboard, screen-reader, contrast, caption and learner testing. |
| Safeguarding | What can learners create, receive, share or encounter? | Risk assessment, supervision plan, filtering and reporting routes. |
| Data protection | What personal data enters the service, where does it go and why? | Data-flow map, contract, retention terms and DPIA where required. |
| Workload | Does the whole process save useful time without lowering quality? | Time-on-task plus checking, correction, training and support time. |
| Cost and implementation | Can the school sustain licences, devices, training and support? | Total-cost estimate, implementation owner and exit plan. |
Where teacher-facing AI may help
Planning and resource drafting
Teachers can use a generative tool to produce a first draft of questions, examples, explanations or parent communications. The output still needs subject, curriculum, safeguarding and accessibility checks. Do not enter identifiable pupil information into a public tool, and do not assume that confident prose is correct. The DfE describes AI output as potentially inaccurate, biased, unsafe, out of date or taken out of context.
For practical examples, see the separate guides to AI in lesson planning and using ChatGPT as a teacher. Those pages cover tasks and prompts; this hub remains the evaluation and navigation owner.
Practice and feedback
Technology may increase the quantity of practice, improve access to worked examples or help surface common errors. The educational value comes from the design of the practice and feedback, not the presence of a screen. Check whether questions retrieve the intended knowledge, whether feedback explains rather than merely reveals an answer, and whether learners can transfer what they practised to an independent task.
AI-generated quizzes require the same accuracy check as any other generated resource. The guide to AI-assisted retrieval questions should be used alongside the evidence base for retrieval practice.
Administration and communication
Drafting routine summaries, schedules or communications may be a lower-risk starting point when no sensitive data is supplied and a responsible staff member reviews the final version. Define which records may be used, who approves the output and where the final document is stored. Do not use automation as the sole or unreviewed decision-maker for admissions, behaviour, safeguarding, SEND support or other decisions with significant effects. Involve the appropriate professional and seek data-protection advice where profiling or automated decision-making is proposed.
When pupils use AI or EdTech directly
Pupil-facing use needs stronger controls because learners may encounter unsuitable output, disclose personal information, misunderstand a generated answer or become dependent on support that completes the thinking for them. Follow product age restrictions, your safeguarding and filtering arrangements, and the current DfE guidance. Explain what the tool may do, what information must not be entered and how learners should report a concerning response.
Design the activity so the learner still performs the important thinking. For example, a pupil might compare an AI explanation with a textbook and identify three errors, or use a tool to generate alternative examples after first solving a problem independently. A task that asks the tool to produce the assessed answer provides little evidence of the pupil's knowledge.
For assessed work, use current awarding-body and centre rules. In England, Ofqual's 2026 resources state that using AI to generate coursework without proper disclosure is cheating and recommend consistent whole-school messaging. The AI and academic integrity guide covers classroom discussion, while official rules remain the authority for individual qualifications.
Accessibility and SEND: test, do not assume
A digital tool may support some learners through captions, text-to-speech, speech input, adjustable display, structured prompts or alternative formats. It may also introduce barriers through poor keyboard access, confusing interfaces, inaccurate simplification, sensory overload or inaccessible documents. Do not assume that a feature will help solely because a learner has a particular disability label.
Test the specific task with the learner and relevant staff. Keep an equivalent route when the tool fails, record any reasonable adjustments and check whether the learner can work more independently rather than simply complete more with hidden adult or automated support. The broader SEND guide provides context; product decisions still require individual evidence and professional judgement.
Data protection, safety and intellectual property
- Map the data: document what is entered, generated, retained, shared and used for product development or analytics.
- Use the right governance: involve the data protection officer, IT lead and designated safeguarding lead where appropriate.
- Minimise information: do not submit names, work, health information or other identifiable pupil data unless the use has an appropriate lawful and approved basis.
- Check contracts and settings: free and paid versions may use data differently; marketing language is not a data-processing agreement.
- Respect rights: check ownership, licensing, attribution, permission and whether a statutory exception applies before uploading protected material or publishing generated outputs.
- Plan for failure: identify who stops use, preserves evidence and communicates with families if the service produces harmful output or exposes data.
The DfE's AI and data-protection guidance for schools was updated in July 2026. Its product safety standards can also help schools question suppliers about educational AI systems.
How to run a small, useful trial
- Define one use case. Name the users, task, intended benefit and prohibited uses.
- Record the baseline. Measure the current learning, quality, time, access or error rate before introducing the tool.
- Screen the product. Review safety, privacy, accessibility, curriculum fit, evidence, cost and technical support.
- Trial on a limited scale. Use representative tasks and preserve a comparison with the existing process.
- Collect more than opinions. Combine user experience with checked outputs, learning evidence, workload, incidents and total cost.
- Decide and document. Adopt, adapt or stop; name the reason, owner, review date and exit arrangements.
The duration should match the decision. A short drafting trial may reveal accuracy and workload quickly; a claim about learner progress requires enough teaching and assessment evidence to support it. A “six-week” or “100-day” plan can be a local project timetable, but it is not a universal evidence-based formula.
How to judge product evidence
Ask whether the evidence concerns the same product version, users, subject, age group and outcome that your school cares about. A study of one adaptive mathematics system does not establish the effect of EdTech in general. A supplier case study can identify questions, but it is weaker than independently collected evidence with a clear comparison and disclosed methods.
Randomised trials can be valuable, but they are not the only decision input. Schools also need implementation feasibility, safeguarding, accessibility, data governance, workload and cost evidence. Avoid converting every research result into “months of progress,” a guaranteed percentage gain or a cost-per-percentile score. Report what was actually measured and preserve uncertainty.
A practical whole-school policy boundary
A useful policy names permitted, restricted and prohibited uses; distinguishes staff-facing from pupil-facing activity; explains data and safeguarding rules; sets expectations for checking and disclosure; and identifies who approves new services. It should connect with existing assessment, acceptable-use, data-protection, safeguarding, curriculum and procurement processes rather than operate as an isolated AI document.
Use the school AI policy guide, AI policy template and whole-school AI strategy for implementation. Review them against the current DfE material rather than treating a template as legal advice.
Choose the next guide
- AI tools for teachers for a task-based tools overview.
- AI for teachers for the wider evidence and professional-practice discussion.
- DfE guidance on AI in schools for an implementation summary linked to the official source.
- AI ethics in education for fairness, agency and accountability questions.
- AI CPD for schools for staff-development planning.
Authoritative guidance and further reading
- Department for Education: generative AI in education
- Department for Education: 2026–27 AI support materials
- Department for Education: generative AI product safety standards
- Department for Education: procuring educational technology
- Ofqual: AI in coursework resources for schools
- Education Endowment Foundation: using digital technology to improve learning
- UNESCO: guidance for generative AI in education and research
